diff options
Diffstat (limited to 'src/gallium/drivers/r600')
24 files changed, 1536 insertions, 1284 deletions
diff --git a/src/gallium/drivers/r600/Android.mk b/src/gallium/drivers/r600/Android.mk new file mode 100644 index 00000000000..994ae07789c --- /dev/null +++ b/src/gallium/drivers/r600/Android.mk @@ -0,0 +1,42 @@ +# Mesa 3-D graphics library +# +# Copyright (C) 2010-2011 Chia-I Wu <[email protected]> +# Copyright (C) 2010-2011 LunarG Inc. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the "Software"), +# to deal in the Software without restriction, including without limitation +# the rights to use, copy, modify, merge, publish, distribute, sublicense, +# and/or sell copies of the Software, and to permit persons to whom the +# Software is furnished to do so, subject to the following conditions: +# +# The above copyright notice and this permission notice shall be included +# in all copies or substantial portions of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL +# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING +# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER +# DEALINGS IN THE SOFTWARE. + +LOCAL_PATH := $(call my-dir) + +# get C_SOURCES +include $(LOCAL_PATH)/Makefile.sources + +include $(CLEAR_VARS) + +LOCAL_SRC_FILES := $(C_SOURCES) + +LOCAL_CFLAGS := -std=c99 + +LOCAL_C_INCLUDES := \ + $(DRM_TOP) \ + $(DRM_TOP)/include/drm + +LOCAL_MODULE := libmesa_pipe_r600 + +include $(GALLIUM_COMMON_MK) +include $(BUILD_STATIC_LIBRARY) diff --git a/src/gallium/drivers/r600/Makefile b/src/gallium/drivers/r600/Makefile index 7e21e3e32b1..0e68fe99345 100644 --- a/src/gallium/drivers/r600/Makefile +++ b/src/gallium/drivers/r600/Makefile @@ -6,20 +6,7 @@ LIBNAME = r600 LIBRARY_INCLUDES = \ $(shell pkg-config libdrm --cflags-only-I) -C_SOURCES = \ - r600_asm.c \ - r600_blit.c \ - r600_buffer.c \ - r600_pipe.c \ - r600_query.c \ - r600_resource.c \ - r600_shader.c \ - r600_state.c \ - r600_texture.c \ - r700_asm.c \ - evergreen_state.c \ - eg_asm.c \ - r600_translate.c \ - r600_state_common.c +# get C_SOURCES +include Makefile.sources include ../../Makefile.template diff --git a/src/gallium/drivers/r600/Makefile.sources b/src/gallium/drivers/r600/Makefile.sources new file mode 100644 index 00000000000..0366394e5a2 --- /dev/null +++ b/src/gallium/drivers/r600/Makefile.sources @@ -0,0 +1,15 @@ +C_SOURCES := \ + r600_asm.c \ + r600_blit.c \ + r600_buffer.c \ + r600_pipe.c \ + r600_query.c \ + r600_resource.c \ + r600_shader.c \ + r600_state.c \ + r600_texture.c \ + r700_asm.c \ + evergreen_state.c \ + eg_asm.c \ + r600_translate.c \ + r600_state_common.c diff --git a/src/gallium/drivers/r600/SConscript b/src/gallium/drivers/r600/SConscript index 19f07b2bef8..be12255e4d0 100644 --- a/src/gallium/drivers/r600/SConscript +++ b/src/gallium/drivers/r600/SConscript @@ -11,22 +11,8 @@ env.Append(CPPPATH = [ r600 = env.ConvenienceLibrary( target = 'r600', - source = [ - 'r600_asm.c', - 'r600_buffer.c', - 'r600_blit.c', - 'r600_pipe.c', - 'r600_query.c', - 'r600_resource.c', - 'r600_shader.c', - 'r600_state.c', - 'r600_state_common.c', - 'r600_texture.c', - 'r600_translate.c', - 'r700_asm.c', - 'evergreen_state.c', - 'eg_asm.c', - ]) + source = env.ParseSourceList('Makefile.sources', 'C_SOURCES') + ) env.Alias('r600', r600) diff --git a/src/gallium/drivers/r600/eg_asm.c b/src/gallium/drivers/r600/eg_asm.c index c95872b0809..ca25b341ffd 100644 --- a/src/gallium/drivers/r600/eg_asm.c +++ b/src/gallium/drivers/r600/eg_asm.c @@ -29,7 +29,7 @@ #include "r600_opcodes.h" #include "evergreend.h" -int eg_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) +int eg_bytecode_cf_build(struct r600_bytecode *bc, struct r600_bytecode_cf *cf) { unsigned id = cf->id; diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c index 4605c833dea..2135b8ac580 100644 --- a/src/gallium/drivers/r600/evergreen_state.c +++ b/src/gallium/drivers/r600/evergreen_state.c @@ -262,22 +262,16 @@ static uint32_t r600_translate_dbformat(enum pipe_format format) case PIPE_FORMAT_Z16_UNORM: return V_028040_Z_16; case PIPE_FORMAT_Z24X8_UNORM: - return V_028040_Z_24; case PIPE_FORMAT_Z24_UNORM_S8_USCALED: return V_028040_Z_24; + case PIPE_FORMAT_Z32_FLOAT: + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: + return V_028040_Z_32_FLOAT; default: return ~0U; } } -static uint32_t r600_translate_stencilformat(enum pipe_format format) -{ - if (format == PIPE_FORMAT_Z24_UNORM_S8_USCALED) - return 1; - else - return 0; -} - static uint32_t r600_translate_colorswap(enum pipe_format format) { switch (format) { @@ -360,6 +354,7 @@ static uint32_t r600_translate_colorswap(enum pipe_format format) case PIPE_FORMAT_R11G11B10_FLOAT: case PIPE_FORMAT_R32_FLOAT: + case PIPE_FORMAT_Z32_FLOAT: case PIPE_FORMAT_R16G16_FLOAT: case PIPE_FORMAT_R16G16_UNORM: return V_028C70_SWAP_STD; @@ -369,6 +364,7 @@ static uint32_t r600_translate_colorswap(enum pipe_format format) case PIPE_FORMAT_R16G16B16A16_UNORM: case PIPE_FORMAT_R16G16B16A16_SNORM: case PIPE_FORMAT_R16G16B16A16_FLOAT: + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: /* 128-bit buffers. */ case PIPE_FORMAT_R32G32B32A32_FLOAT: @@ -453,7 +449,11 @@ static uint32_t r600_translate_colorformat(enum pipe_format format) case PIPE_FORMAT_S8_USCALED_Z24_UNORM: return V_028C70_COLOR_24_8; + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: + return V_028C70_COLOR_X24_8_32_FLOAT; + case PIPE_FORMAT_R32_FLOAT: + case PIPE_FORMAT_Z32_FLOAT: return V_028C70_COLOR_32_FLOAT; case PIPE_FORMAT_R16G16_FLOAT: @@ -541,6 +541,7 @@ static uint32_t r600_colorformat_endian_swap(uint32_t colorformat) case V_028C70_COLOR_32_32_FLOAT: case V_028C70_COLOR_32_32: + case V_028C70_COLOR_X24_8_32_FLOAT: return ENDIAN_8IN32; /* 96-bit buffers. */ @@ -638,10 +639,10 @@ static void evergreen_set_blend_color(struct pipe_context *ctx, return; rstate->id = R600_PIPE_STATE_BLEND_COLOR; - r600_pipe_state_add_reg(rstate, R_028414_CB_BLEND_RED, fui(state->color[0]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028418_CB_BLEND_GREEN, fui(state->color[1]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02841C_CB_BLEND_BLUE, fui(state->color[2]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028420_CB_BLEND_ALPHA, fui(state->color[3]), 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028414_CB_BLEND_RED, fui(state->color[0]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028418_CB_BLEND_GREEN, fui(state->color[1]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02841C_CB_BLEND_BLUE, fui(state->color[2]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028420_CB_BLEND_ALPHA, fui(state->color[3]), 0xFFFFFFFF, NULL, 0); free(rctx->states[R600_PIPE_STATE_BLEND_COLOR]); rctx->states[R600_PIPE_STATE_BLEND_COLOR] = rstate; @@ -686,13 +687,13 @@ static void *evergreen_create_blend_state(struct pipe_context *ctx, blend->cb_target_mask = target_mask; r600_pipe_state_add_reg(rstate, R_028808_CB_COLOR_CONTROL, - color_control, 0xFFFFFFFD, NULL); + color_control, 0xFFFFFFFD, NULL, 0); if (rctx->chip_class != CAYMAN) - r600_pipe_state_add_reg(rstate, R_028C3C_PA_SC_AA_MASK, 0xFFFFFFFF, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028C3C_PA_SC_AA_MASK, 0xFFFFFFFF, 0xFFFFFFFF, NULL, 0); else { - r600_pipe_state_add_reg(rstate, CM_R_028C38_PA_SC_AA_MASK_X0Y0_X1Y0, 0xFFFFFFFF, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, CM_R_028C3C_PA_SC_AA_MASK_X0Y1_X1Y1, 0xFFFFFFFF, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, CM_R_028C38_PA_SC_AA_MASK_X0Y0_X1Y0, 0xFFFFFFFF, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, CM_R_028C3C_PA_SC_AA_MASK_X0Y1_X1Y1, 0xFFFFFFFF, 0xFFFFFFFF, NULL, 0); } for (int i = 0; i < 8; i++) { @@ -723,7 +724,7 @@ static void *evergreen_create_blend_state(struct pipe_context *ctx, } } for (int i = 0; i < 8; i++) { - r600_pipe_state_add_reg(rstate, R_028780_CB_BLEND0_CONTROL + i * 4, blend_cntl[i], 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028780_CB_BLEND0_CONTROL + i * 4, blend_cntl[i], 0xFFFFFFFF, NULL, 0); } return rstate; @@ -791,27 +792,27 @@ static void *evergreen_create_dsa_state(struct pipe_context *ctx, S_02800C_FORCE_HIS_ENABLE0(V_02800C_FORCE_DISABLE) | S_02800C_FORCE_HIS_ENABLE1(V_02800C_FORCE_DISABLE); /* TODO db_render_override depends on query */ - r600_pipe_state_add_reg(rstate, R_028028_DB_STENCIL_CLEAR, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02802C_DB_DEPTH_CLEAR, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028410_SX_ALPHA_TEST_CONTROL, alpha_test_control, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028028_DB_STENCIL_CLEAR, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02802C_DB_DEPTH_CLEAR, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028410_SX_ALPHA_TEST_CONTROL, alpha_test_control, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028430_DB_STENCILREFMASK, stencil_ref_mask, - 0xFFFFFFFF & C_028430_STENCILREF, NULL); + 0xFFFFFFFF & C_028430_STENCILREF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028434_DB_STENCILREFMASK_BF, stencil_ref_mask_bf, - 0xFFFFFFFF & C_028434_STENCILREF_BF, NULL); - r600_pipe_state_add_reg(rstate, R_0286DC_SPI_FOG_CNTL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028800_DB_DEPTH_CONTROL, db_depth_control, 0xFFFFFFFF, NULL); + 0xFFFFFFFF & C_028434_STENCILREF_BF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0286DC_SPI_FOG_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028800_DB_DEPTH_CONTROL, db_depth_control, 0xFFFFFFFF, NULL, 0); /* The DB_SHADER_CONTROL mask is 0xFFFFFFBC since Z_EXPORT_ENABLE, * STENCIL_EXPORT_ENABLE and KILL_ENABLE are controlled by * evergreen_pipe_shader_ps().*/ - r600_pipe_state_add_reg(rstate, R_02880C_DB_SHADER_CONTROL, db_shader_control, 0xFFFFFFBC, NULL); - r600_pipe_state_add_reg(rstate, R_028000_DB_RENDER_CONTROL, db_render_control, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02800C_DB_RENDER_OVERRIDE, db_render_override, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AC0_DB_SRESULTS_COMPARE_STATE0, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AC4_DB_SRESULTS_COMPARE_STATE1, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AC8_DB_PRELOAD_CONTROL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028B70_DB_ALPHA_TO_MASK, 0x0000AA00, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_02880C_DB_SHADER_CONTROL, db_shader_control, 0xFFFFFFBC, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028000_DB_RENDER_CONTROL, db_render_control, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02800C_DB_RENDER_OVERRIDE, db_render_override, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AC0_DB_SRESULTS_COMPARE_STATE0, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AC4_DB_SRESULTS_COMPARE_STATE1, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AC8_DB_PRELOAD_CONTROL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028B70_DB_ALPHA_TO_MASK, 0x0000AA00, 0xFFFFFFFF, NULL, 0); return rstate; } @@ -856,7 +857,7 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx, tmp |= S_0286D4_PNT_SPRITE_TOP_1(1); } } - r600_pipe_state_add_reg(rstate, R_0286D4_SPI_INTERP_CONTROL_0, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_0286D4_SPI_INTERP_CONTROL_0, tmp, 0xFFFFFFFF, NULL, 0); polygon_dual_mode = (state->fill_front != PIPE_POLYGON_MODE_FILL || state->fill_back != PIPE_POLYGON_MODE_FILL); @@ -870,44 +871,44 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx, S_028814_POLY_OFFSET_PARA_ENABLE(state->offset_tri) | S_028814_POLY_MODE(polygon_dual_mode) | S_028814_POLYMODE_FRONT_PTYPE(r600_translate_fill(state->fill_front)) | - S_028814_POLYMODE_BACK_PTYPE(r600_translate_fill(state->fill_back)), 0xFFFFFFFF, NULL); + S_028814_POLYMODE_BACK_PTYPE(r600_translate_fill(state->fill_back)), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_02881C_PA_CL_VS_OUT_CNTL, S_02881C_USE_VTX_POINT_SIZE(state->point_size_per_vertex) | - S_02881C_VS_OUT_MISC_VEC_ENA(state->point_size_per_vertex), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028820_PA_CL_NANINF_CNTL, 0x00000000, 0xFFFFFFFF, NULL); + S_02881C_VS_OUT_MISC_VEC_ENA(state->point_size_per_vertex), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028820_PA_CL_NANINF_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); /* point size 12.4 fixed point */ tmp = (unsigned)(state->point_size * 8.0); - r600_pipe_state_add_reg(rstate, R_028A00_PA_SU_POINT_SIZE, S_028A00_HEIGHT(tmp) | S_028A00_WIDTH(tmp), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A04_PA_SU_POINT_MINMAX, 0x80000000, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028A00_PA_SU_POINT_SIZE, S_028A00_HEIGHT(tmp) | S_028A00_WIDTH(tmp), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A04_PA_SU_POINT_MINMAX, 0x80000000, 0xFFFFFFFF, NULL, 0); tmp = (unsigned)state->line_width * 8; - r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, S_028A08_WIDTH(tmp), 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, S_028A08_WIDTH(tmp), 0xFFFFFFFF, NULL, 0); if (rctx->chip_class == CAYMAN) { - r600_pipe_state_add_reg(rstate, CM_R_028BDC_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, CM_R_028BDC_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, CM_R_028BE4_PA_SU_VTX_CNTL, S_028C08_PIX_CENTER_HALF(state->gl_rasterization_rules), - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, CM_R_028BE8_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, CM_R_028BEC_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, CM_R_028BF0_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, CM_R_028BF4_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, CM_R_028BE8_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, CM_R_028BEC_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, CM_R_028BF0_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, CM_R_028BF4_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); } else { - r600_pipe_state_add_reg(rstate, R_028C00_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028C00_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028C0C_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028C10_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028C14_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028C18_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028C0C_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028C10_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028C14_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028C18_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C08_PA_SU_VTX_CNTL, S_028C08_PIX_CENTER_HALF(state->gl_rasterization_rules), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); } - r600_pipe_state_add_reg(rstate, R_028B7C_PA_SU_POLY_OFFSET_CLAMP, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02820C_PA_SC_CLIPRECT_RULE, clip_rule, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028B7C_PA_SU_POLY_OFFSET_CLAMP, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02820C_PA_SC_CLIPRECT_RULE, clip_rule, 0xFFFFFFFF, NULL, 0); return rstate; } @@ -933,22 +934,22 @@ static void *evergreen_create_sampler_state(struct pipe_context *ctx, S_03C000_MIP_FILTER(r600_tex_mipfilter(state->min_mip_filter)) | S_03C000_MAX_ANISO(r600_tex_aniso_filter(state->max_anisotropy)) | S_03C000_DEPTH_COMPARE_FUNCTION(r600_tex_compare(state->compare_func)) | - S_03C000_BORDER_COLOR_TYPE(uc.ui ? V_03C000_SQ_TEX_BORDER_COLOR_REGISTER : 0), 0xFFFFFFFF, NULL); + S_03C000_BORDER_COLOR_TYPE(uc.ui ? V_03C000_SQ_TEX_BORDER_COLOR_REGISTER : 0), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg_noblock(rstate, R_03C004_SQ_TEX_SAMPLER_WORD1_0, S_03C004_MIN_LOD(S_FIXED(CLAMP(state->min_lod, 0, 15), 8)) | S_03C004_MAX_LOD(S_FIXED(CLAMP(state->max_lod, 0, 15), 8)), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg_noblock(rstate, R_03C008_SQ_TEX_SAMPLER_WORD2_0, S_03C008_LOD_BIAS(S_FIXED(CLAMP(state->lod_bias, -16, 16), 8)) | (state->seamless_cube_map ? 0 : S_03C008_DISABLE_CUBE_WRAP(1)) | S_03C008_TYPE(1), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); if (uc.ui) { - r600_pipe_state_add_reg_noblock(rstate, R_00A404_TD_PS_SAMPLER0_BORDER_RED, fui(state->border_color[0]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg_noblock(rstate, R_00A408_TD_PS_SAMPLER0_BORDER_GREEN, fui(state->border_color[1]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg_noblock(rstate, R_00A40C_TD_PS_SAMPLER0_BORDER_BLUE, fui(state->border_color[2]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg_noblock(rstate, R_00A410_TD_PS_SAMPLER0_BORDER_ALPHA, fui(state->border_color[3]), 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg_noblock(rstate, R_00A404_TD_PS_SAMPLER0_BORDER_RED, fui(state->border_color[0]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg_noblock(rstate, R_00A408_TD_PS_SAMPLER0_BORDER_GREEN, fui(state->border_color[1]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg_noblock(rstate, R_00A40C_TD_PS_SAMPLER0_BORDER_BLUE, fui(state->border_color[2]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg_noblock(rstate, R_00A410_TD_PS_SAMPLER0_BORDER_ALPHA, fui(state->border_color[3]), 0xFFFFFFFF, NULL, 0); } return rstate; } @@ -957,43 +958,39 @@ static struct pipe_sampler_view *evergreen_create_sampler_view(struct pipe_conte struct pipe_resource *texture, const struct pipe_sampler_view *state) { - struct r600_pipe_sampler_view *resource = CALLOC_STRUCT(r600_pipe_sampler_view); + struct r600_pipe_sampler_view *view = CALLOC_STRUCT(r600_pipe_sampler_view); struct r600_pipe_resource_state *rstate; - const struct util_format_description *desc; - struct r600_resource_texture *tmp; + struct r600_resource_texture *tmp = (struct r600_resource_texture*)texture; struct r600_resource *rbuffer; unsigned format, endian; uint32_t word4 = 0, yuv_format = 0, pitch = 0; unsigned char swizzle[4], array_mode = 0, tile_type = 0; struct r600_bo *bo[2]; - if (resource == NULL) + if (view == NULL) return NULL; - rstate = &resource->state; + rstate = &view->state; /* initialize base object */ - resource->base = *state; - resource->base.texture = NULL; + view->base = *state; + view->base.texture = NULL; pipe_reference(NULL, &texture->reference); - resource->base.texture = texture; - resource->base.reference.count = 1; - resource->base.context = ctx; + view->base.texture = texture; + view->base.reference.count = 1; + view->base.context = ctx; swizzle[0] = state->swizzle_r; swizzle[1] = state->swizzle_g; swizzle[2] = state->swizzle_b; swizzle[3] = state->swizzle_a; + format = r600_translate_texformat(ctx->screen, state->format, swizzle, &word4, &yuv_format); if (format == ~0) { format = 0; } - desc = util_format_description(state->format); - if (desc == NULL) { - R600_ERR("unknow format %d\n", state->format); - } - tmp = (struct r600_resource_texture *)texture; + if (tmp->depth && !tmp->is_flushing_texture) { r600_texture_depth_flush(ctx, texture, TRUE); tmp = tmp->flushed_depth_texture; @@ -1016,6 +1013,8 @@ static struct pipe_sampler_view *evergreen_create_sampler_view(struct pipe_conte rstate->bo[0] = bo[0]; rstate->bo[1] = bo[1]; + rstate->bo_usage[0] = RADEON_USAGE_READ; + rstate->bo_usage[1] = RADEON_USAGE_READ; rstate->val[0] = (S_030000_DIM(r600_tex_dim(texture->target)) | S_030000_PITCH((pitch / 8) - 1) | S_030000_NON_DISP_TILING_ORDER(tile_type) | @@ -1023,8 +1022,8 @@ static struct pipe_sampler_view *evergreen_create_sampler_view(struct pipe_conte rstate->val[1] = (S_030004_TEX_HEIGHT(texture->height0 - 1) | S_030004_TEX_DEPTH(texture->depth0 - 1) | S_030004_ARRAY_MODE(array_mode)); - rstate->val[2] = (tmp->offset[0] + r600_bo_offset(bo[0])) >> 8; - rstate->val[3] = (tmp->offset[1] + r600_bo_offset(bo[1])) >> 8; + rstate->val[2] = tmp->offset[0] >> 8; + rstate->val[3] = tmp->offset[1] >> 8; rstate->val[4] = (word4 | S_030010_SRF_MODE_ALL(V_030010_SRF_MODE_ZERO_CLAMP_MINUS_ONE) | S_030010_ENDIAN_SWAP(endian) | @@ -1036,7 +1035,7 @@ static struct pipe_sampler_view *evergreen_create_sampler_view(struct pipe_conte rstate->val[7] = (S_03001C_DATA_FORMAT(format) | S_03001C_TYPE(V_03001C_SQ_TEX_VTX_VALID_TEXTURE)); - return &resource->base; + return &view->base; } static void evergreen_set_vs_sampler_view(struct pipe_context *ctx, unsigned count, @@ -1131,21 +1130,21 @@ static void evergreen_set_clip_state(struct pipe_context *ctx, for (int i = 0; i < state->nr; i++) { r600_pipe_state_add_reg(rstate, R_0285BC_PA_CL_UCP0_X + i * 16, - fui(state->ucp[i][0]), 0xFFFFFFFF, NULL); + fui(state->ucp[i][0]), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0285C0_PA_CL_UCP0_Y + i * 16, - fui(state->ucp[i][1]) , 0xFFFFFFFF, NULL); + fui(state->ucp[i][1]) , 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0285C4_PA_CL_UCP0_Z + i * 16, - fui(state->ucp[i][2]), 0xFFFFFFFF, NULL); + fui(state->ucp[i][2]), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0285C8_PA_CL_UCP0_W + i * 16, - fui(state->ucp[i][3]), 0xFFFFFFFF, NULL); + fui(state->ucp[i][3]), 0xFFFFFFFF, NULL, 0); } r600_pipe_state_add_reg(rstate, R_028810_PA_CL_CLIP_CNTL, S_028810_PS_UCP_MODE(3) | ((1 << state->nr) - 1) | S_028810_ZCLIP_NEAR_DISABLE(state->depth_clamp) | - S_028810_ZCLIP_FAR_DISABLE(state->depth_clamp), 0xFFFFFFFF, NULL); + S_028810_ZCLIP_FAR_DISABLE(state->depth_clamp), 0xFFFFFFFF, NULL, 0); free(rctx->states[R600_PIPE_STATE_CLIP]); rctx->states[R600_PIPE_STATE_CLIP] = rstate; @@ -1176,28 +1175,28 @@ static void evergreen_set_scissor_state(struct pipe_context *ctx, br = S_028244_BR_X(state->maxx) | S_028244_BR_Y(state->maxy); r600_pipe_state_add_reg(rstate, R_028210_PA_SC_CLIPRECT_0_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028214_PA_SC_CLIPRECT_0_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028218_PA_SC_CLIPRECT_1_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_02821C_PA_SC_CLIPRECT_1_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028220_PA_SC_CLIPRECT_2_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028224_PA_SC_CLIPRECT_2_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028228_PA_SC_CLIPRECT_3_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_02822C_PA_SC_CLIPRECT_3_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); free(rctx->states[R600_PIPE_STATE_SCISSOR]); rctx->states[R600_PIPE_STATE_SCISSOR] = rstate; @@ -1219,11 +1218,11 @@ static void evergreen_set_stencil_ref(struct pipe_context *ctx, tmp = S_028430_STENCILREF(state->ref_value[0]); r600_pipe_state_add_reg(rstate, R_028430_DB_STENCILREFMASK, tmp, - ~C_028430_STENCILREF, NULL); + ~C_028430_STENCILREF, NULL, 0); tmp = S_028434_STENCILREF_BF(state->ref_value[1]); r600_pipe_state_add_reg(rstate, R_028434_DB_STENCILREFMASK_BF, tmp, - ~C_028434_STENCILREF_BF, NULL); + ~C_028434_STENCILREF_BF, NULL, 0); free(rctx->states[R600_PIPE_STATE_STENCIL_REF]); rctx->states[R600_PIPE_STATE_STENCIL_REF] = rstate; @@ -1241,15 +1240,15 @@ static void evergreen_set_viewport_state(struct pipe_context *ctx, rctx->viewport = *state; rstate->id = R600_PIPE_STATE_VIEWPORT; - r600_pipe_state_add_reg(rstate, R_0282D0_PA_SC_VPORT_ZMIN_0, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0282D4_PA_SC_VPORT_ZMAX_0, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02843C_PA_CL_VPORT_XSCALE_0, fui(state->scale[0]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028444_PA_CL_VPORT_YSCALE_0, fui(state->scale[1]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02844C_PA_CL_VPORT_ZSCALE_0, fui(state->scale[2]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028440_PA_CL_VPORT_XOFFSET_0, fui(state->translate[0]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028448_PA_CL_VPORT_YOFFSET_0, fui(state->translate[1]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028450_PA_CL_VPORT_ZOFFSET_0, fui(state->translate[2]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028818_PA_CL_VTE_CNTL, 0x0000043F, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_0282D0_PA_SC_VPORT_ZMIN_0, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0282D4_PA_SC_VPORT_ZMAX_0, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02843C_PA_CL_VPORT_XSCALE_0, fui(state->scale[0]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028444_PA_CL_VPORT_YSCALE_0, fui(state->scale[1]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02844C_PA_CL_VPORT_ZSCALE_0, fui(state->scale[2]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028440_PA_CL_VPORT_XOFFSET_0, fui(state->translate[0]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028448_PA_CL_VPORT_YOFFSET_0, fui(state->translate[1]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028450_PA_CL_VPORT_ZOFFSET_0, fui(state->translate[2]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028818_PA_CL_VTE_CNTL, 0x0000043F, 0xFFFFFFFF, NULL, 0); free(rctx->states[R600_PIPE_STATE_VIEWPORT]); rctx->states[R600_PIPE_STATE_VIEWPORT] = rstate; @@ -1354,86 +1353,82 @@ static void evergreen_cb(struct r600_pipe_context *rctx, struct r600_pipe_state /* FIXME handle enabling of CB beyond BASE8 which has different offset */ r600_pipe_state_add_reg(rstate, R_028C60_CB_COLOR0_BASE + cb * 0x3C, - (offset + r600_bo_offset(bo[0])) >> 8, 0xFFFFFFFF, bo[0]); + offset >> 8, 0xFFFFFFFF, bo[0], RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_028C78_CB_COLOR0_DIM + cb * 0x3C, - 0x0, 0xFFFFFFFF, NULL); + 0x0, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C70_CB_COLOR0_INFO + cb * 0x3C, - color_info, 0xFFFFFFFF, bo[0]); + color_info, 0xFFFFFFFF, bo[0], RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_028C64_CB_COLOR0_PITCH + cb * 0x3C, S_028C64_PITCH_TILE_MAX(pitch), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C68_CB_COLOR0_SLICE + cb * 0x3C, S_028C68_SLICE_TILE_MAX(slice), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C6C_CB_COLOR0_VIEW + cb * 0x3C, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C74_CB_COLOR0_ATTRIB + cb * 0x3C, S_028C74_NON_DISP_TILING_ORDER(tile_type), - 0xFFFFFFFF, bo[0]); + 0xFFFFFFFF, bo[0], RADEON_USAGE_READWRITE); } static void evergreen_db(struct r600_pipe_context *rctx, struct r600_pipe_state *rstate, - const struct pipe_framebuffer_state *state) + const struct pipe_framebuffer_state *state) { struct r600_resource_texture *rtex; - struct r600_resource *rbuffer; struct r600_surface *surf; - unsigned level; - unsigned pitch, slice, format, stencil_format; + unsigned level, first_layer; + unsigned pitch, slice, format; unsigned offset; if (state->zsbuf == NULL) return; - level = state->zsbuf->u.tex.level; - surf = (struct r600_surface *)state->zsbuf; - rtex = (struct r600_resource_texture*)state->zsbuf->texture; - - rbuffer = &rtex->resource; + rtex = (struct r600_resource_texture*)surf->base.texture; - /* XXX quite sure for dx10+ hw don't need any offset hacks */ - offset = r600_texture_get_offset((struct r600_resource_texture *)state->zsbuf->texture, - level, state->zsbuf->u.tex.first_layer); + level = surf->base.u.tex.level; + first_layer = surf->base.u.tex.first_layer; + offset = r600_texture_get_offset(rtex, level, first_layer); pitch = rtex->pitch_in_blocks[level] / 8 - 1; slice = rtex->pitch_in_blocks[level] * surf->aligned_height / 64 - 1; - format = r600_translate_dbformat(state->zsbuf->texture->format); - stencil_format = r600_translate_stencilformat(state->zsbuf->texture->format); + format = r600_translate_dbformat(rtex->real_format); r600_pipe_state_add_reg(rstate, R_028048_DB_Z_READ_BASE, - (offset + r600_bo_offset(rbuffer->bo)) >> 8, 0xFFFFFFFF, rbuffer->bo); + offset >> 8, 0xFFFFFFFF, rtex->resource.bo, RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_028050_DB_Z_WRITE_BASE, - (offset + r600_bo_offset(rbuffer->bo)) >> 8, 0xFFFFFFFF, rbuffer->bo); + offset >> 8, 0xFFFFFFFF, rtex->resource.bo, RADEON_USAGE_READWRITE); + r600_pipe_state_add_reg(rstate, R_028008_DB_DEPTH_VIEW, 0x00000000, 0xFFFFFFFF, NULL, 0); - if (stencil_format) { - uint32_t stencil_offset; + if (rtex->stencil) { + uint32_t stencil_offset = + r600_texture_get_offset(rtex->stencil, level, first_layer); - stencil_offset = ((surf->aligned_height * rtex->pitch_in_bytes[level]) + 255) & ~255; r600_pipe_state_add_reg(rstate, R_02804C_DB_STENCIL_READ_BASE, - (offset + stencil_offset + r600_bo_offset(rbuffer->bo)) >> 8, 0xFFFFFFFF, rbuffer->bo); + stencil_offset >> 8, 0xFFFFFFFF, rtex->stencil->resource.bo, RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_028054_DB_STENCIL_WRITE_BASE, - (offset + stencil_offset + r600_bo_offset(rbuffer->bo)) >> 8, 0xFFFFFFFF, rbuffer->bo); + stencil_offset >> 8, 0xFFFFFFFF, rtex->stencil->resource.bo, RADEON_USAGE_READWRITE); + r600_pipe_state_add_reg(rstate, R_028044_DB_STENCIL_INFO, + 1, 0xFFFFFFFF, rtex->stencil->resource.bo, RADEON_USAGE_READWRITE); + } else { + r600_pipe_state_add_reg(rstate, R_028044_DB_STENCIL_INFO, + 0, 0xFFFFFFFF, NULL, RADEON_USAGE_READWRITE); } - r600_pipe_state_add_reg(rstate, R_028008_DB_DEPTH_VIEW, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028044_DB_STENCIL_INFO, - S_028044_FORMAT(stencil_format), 0xFFFFFFFF, rbuffer->bo); - r600_pipe_state_add_reg(rstate, R_028040_DB_Z_INFO, S_028040_ARRAY_MODE(rtex->array_mode[level]) | S_028040_FORMAT(format), - 0xFFFFFFFF, rbuffer->bo); + 0xFFFFFFFF, rtex->resource.bo, RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_028058_DB_DEPTH_SIZE, S_028058_PITCH_TILE_MAX(pitch), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_02805C_DB_DEPTH_SLICE, S_02805C_SLICE_TILE_MAX(slice), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); } static void evergreen_set_framebuffer_state(struct pipe_context *ctx, @@ -1492,49 +1487,49 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_028240_PA_SC_GENERIC_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028244_PA_SC_GENERIC_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028250_PA_SC_VPORT_SCISSOR_0_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028254_PA_SC_VPORT_SCISSOR_0_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028030_PA_SC_SCREEN_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028034_PA_SC_SCREEN_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028204_PA_SC_WINDOW_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028208_PA_SC_WINDOW_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028200_PA_SC_WINDOW_OFFSET, 0x00000000, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028230_PA_SC_EDGERULE, 0xAAAAAAAA, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028238_CB_TARGET_MASK, - 0x00000000, target_mask, NULL); + 0x00000000, target_mask, NULL, 0); r600_pipe_state_add_reg(rstate, R_02823C_CB_SHADER_MASK, - shader_mask, 0xFFFFFFFF, NULL); + shader_mask, 0xFFFFFFFF, NULL, 0); if (rctx->chip_class == CAYMAN) { r600_pipe_state_add_reg(rstate, CM_R_028BE0_PA_SC_AA_CONFIG, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); } else { r600_pipe_state_add_reg(rstate, R_028C04_PA_SC_AA_CONFIG, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C1C_PA_SC_AA_SAMPLE_LOCS_MCTX, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); } free(rctx->states[R600_PIPE_STATE_FRAMEBUFFER]); @@ -1609,78 +1604,78 @@ static void cayman_init_config(struct r600_pipe_context *rctx) tmp = 0x00000000; tmp |= S_008C00_EXPORT_SRC_C(1); - r600_pipe_state_add_reg(rstate, R_008C00_SQ_CONFIG, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C00_SQ_CONFIG, tmp, 0xFFFFFFFF, NULL, 0); /* always set the temp clauses */ - r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, S_008C04_NUM_CLAUSE_TEMP_GPRS(4), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, 0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2, 0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, (1 << 8), 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_028A48_PA_SC_MODE_CNTL_0, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL_1, 0x0, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A14_VGT_HOS_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A20_VGT_HOS_REUSE_DEPTH, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A24_VGT_GROUP_PRIM_TYPE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A28_VGT_GROUP_FIRST_DECR, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A2C_VGT_GROUP_DECR, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A30_VGT_GROUP_VECT_0_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A34_VGT_GROUP_VECT_1_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A40_VGT_GS_MODE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028B94_VGT_STRMOUT_CONFIG, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AB4_VGT_REUSE_OFF, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AB8_VGT_VTX_CNT_EN, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_028380_SQ_VTX_SEMANTIC_0, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028384_SQ_VTX_SEMANTIC_1, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028388_SQ_VTX_SEMANTIC_2, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02838C_SQ_VTX_SEMANTIC_3, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028390_SQ_VTX_SEMANTIC_4, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028394_SQ_VTX_SEMANTIC_5, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028398_SQ_VTX_SEMANTIC_6, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02839C_SQ_VTX_SEMANTIC_7, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283A0_SQ_VTX_SEMANTIC_8, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283A4_SQ_VTX_SEMANTIC_9, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283A8_SQ_VTX_SEMANTIC_10, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283AC_SQ_VTX_SEMANTIC_11, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283B0_SQ_VTX_SEMANTIC_12, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283B4_SQ_VTX_SEMANTIC_13, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283B8_SQ_VTX_SEMANTIC_14, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283BC_SQ_VTX_SEMANTIC_15, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283C0_SQ_VTX_SEMANTIC_16, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283C4_SQ_VTX_SEMANTIC_17, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283C8_SQ_VTX_SEMANTIC_18, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283CC_SQ_VTX_SEMANTIC_19, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283D0_SQ_VTX_SEMANTIC_20, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283D4_SQ_VTX_SEMANTIC_21, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283D8_SQ_VTX_SEMANTIC_22, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283DC_SQ_VTX_SEMANTIC_23, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283E0_SQ_VTX_SEMANTIC_24, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283E4_SQ_VTX_SEMANTIC_25, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283E8_SQ_VTX_SEMANTIC_26, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283EC_SQ_VTX_SEMANTIC_27, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283F0_SQ_VTX_SEMANTIC_28, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283F4_SQ_VTX_SEMANTIC_29, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283F8_SQ_VTX_SEMANTIC_30, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283FC_SQ_VTX_SEMANTIC_31, 0x0, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_028810_PA_CL_CLIP_CNTL, 0x0, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, CM_R_028BD4_PA_SC_CENTROID_PRIORITY_0, 0x76543210, 0xffffffff, 0); - r600_pipe_state_add_reg(rstate, CM_R_028BD8_PA_SC_CENTROID_PRIORITY_1, 0xfedcba98, 0xffffffff, 0); - - r600_pipe_state_add_reg(rstate, CM_R_0288E8_SQ_LDS_ALLOC, 0, 0xffffffff, NULL); - r600_pipe_state_add_reg(rstate, R_0288EC_SQ_LDS_ALLOC_PS, 0, 0xffffffff, NULL); - - r600_pipe_state_add_reg(rstate, CM_R_028804_DB_EQAA, 0x110000, 0xffffffff, NULL); + r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, S_008C04_NUM_CLAUSE_TEMP_GPRS(4), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, 0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2, 0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, (1 << 8), 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_028A48_PA_SC_MODE_CNTL_0, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL_1, 0x0, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A14_VGT_HOS_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A20_VGT_HOS_REUSE_DEPTH, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A24_VGT_GROUP_PRIM_TYPE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A28_VGT_GROUP_FIRST_DECR, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A2C_VGT_GROUP_DECR, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A30_VGT_GROUP_VECT_0_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A34_VGT_GROUP_VECT_1_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A40_VGT_GS_MODE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028B94_VGT_STRMOUT_CONFIG, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AB4_VGT_REUSE_OFF, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AB8_VGT_VTX_CNT_EN, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_028380_SQ_VTX_SEMANTIC_0, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028384_SQ_VTX_SEMANTIC_1, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028388_SQ_VTX_SEMANTIC_2, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02838C_SQ_VTX_SEMANTIC_3, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028390_SQ_VTX_SEMANTIC_4, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028394_SQ_VTX_SEMANTIC_5, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028398_SQ_VTX_SEMANTIC_6, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02839C_SQ_VTX_SEMANTIC_7, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283A0_SQ_VTX_SEMANTIC_8, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283A4_SQ_VTX_SEMANTIC_9, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283A8_SQ_VTX_SEMANTIC_10, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283AC_SQ_VTX_SEMANTIC_11, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283B0_SQ_VTX_SEMANTIC_12, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283B4_SQ_VTX_SEMANTIC_13, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283B8_SQ_VTX_SEMANTIC_14, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283BC_SQ_VTX_SEMANTIC_15, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283C0_SQ_VTX_SEMANTIC_16, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283C4_SQ_VTX_SEMANTIC_17, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283C8_SQ_VTX_SEMANTIC_18, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283CC_SQ_VTX_SEMANTIC_19, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283D0_SQ_VTX_SEMANTIC_20, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283D4_SQ_VTX_SEMANTIC_21, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283D8_SQ_VTX_SEMANTIC_22, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283DC_SQ_VTX_SEMANTIC_23, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283E0_SQ_VTX_SEMANTIC_24, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283E4_SQ_VTX_SEMANTIC_25, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283E8_SQ_VTX_SEMANTIC_26, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283EC_SQ_VTX_SEMANTIC_27, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283F0_SQ_VTX_SEMANTIC_28, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283F4_SQ_VTX_SEMANTIC_29, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283F8_SQ_VTX_SEMANTIC_30, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283FC_SQ_VTX_SEMANTIC_31, 0x0, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_028810_PA_CL_CLIP_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, CM_R_028BD4_PA_SC_CENTROID_PRIORITY_0, 0x76543210, 0xffffffff, NULL, 0); + r600_pipe_state_add_reg(rstate, CM_R_028BD8_PA_SC_CENTROID_PRIORITY_1, 0xfedcba98, 0xffffffff, NULL, 0); + + r600_pipe_state_add_reg(rstate, CM_R_0288E8_SQ_LDS_ALLOC, 0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0288EC_SQ_LDS_ALLOC_PS, 0, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, CM_R_028804_DB_EQAA, 0x110000, 0xFFFFFFFF, NULL, 0); r600_context_pipe_state_set(&rctx->ctx, rstate); } @@ -1964,39 +1959,39 @@ void evergreen_init_config(struct r600_pipe_context *rctx) tmp |= S_008C00_VS_PRIO(vs_prio); tmp |= S_008C00_GS_PRIO(gs_prio); tmp |= S_008C00_ES_PRIO(es_prio); - r600_pipe_state_add_reg(rstate, R_008C00_SQ_CONFIG, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C00_SQ_CONFIG, tmp, 0xFFFFFFFF, NULL, 0); /* enable dynamic GPR resource management */ if (r600_get_minor_version(rctx->radeon) >= 7) { /* always set temp clauses */ r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, - S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, 0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2, 0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, (1 << 8), 0xFFFFFFFF, NULL); + S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_008C10_SQ_GLOBAL_GPR_RESOURCE_MGMT_1, 0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_008C14_SQ_GLOBAL_GPR_RESOURCE_MGMT_2, 0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, (1 << 8), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028838_SQ_DYN_GPR_RESOURCE_LIMIT_1, S_028838_PS_GPRS(0x1e) | S_028838_VS_GPRS(0x1e) | S_028838_GS_GPRS(0x1e) | S_028838_ES_GPRS(0x1e) | S_028838_HS_GPRS(0x1e) | - S_028838_LS_GPRS(0x1e), 0xFFFFFFFF, NULL); /* workaround for hw issues with dyn gpr - must set all limits to 240 instead of 0, 0x1e == 240 / 8*/ + S_028838_LS_GPRS(0x1e), 0xFFFFFFFF, NULL, 0); /* workaround for hw issues with dyn gpr - must set all limits to 240 instead of 0, 0x1e == 240 / 8*/ } else { tmp = 0; tmp |= S_008C04_NUM_PS_GPRS(num_ps_gprs); tmp |= S_008C04_NUM_VS_GPRS(num_vs_gprs); tmp |= S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs); - r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL, 0); tmp = 0; tmp |= S_008C08_NUM_GS_GPRS(num_gs_gprs); tmp |= S_008C08_NUM_ES_GPRS(num_es_gprs); - r600_pipe_state_add_reg(rstate, R_008C08_SQ_GPR_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C08_SQ_GPR_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL, 0); tmp = 0; tmp |= S_008C0C_NUM_HS_GPRS(num_hs_gprs); tmp |= S_008C0C_NUM_HS_GPRS(num_ls_gprs); - r600_pipe_state_add_reg(rstate, R_008C0C_SQ_GPR_RESOURCE_MGMT_3, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C0C_SQ_GPR_RESOURCE_MGMT_3, tmp, 0xFFFFFFFF, NULL, 0); } tmp = 0; @@ -2004,109 +1999,109 @@ void evergreen_init_config(struct r600_pipe_context *rctx) tmp |= S_008C18_NUM_VS_THREADS(num_vs_threads); tmp |= S_008C18_NUM_GS_THREADS(num_gs_threads); tmp |= S_008C18_NUM_ES_THREADS(num_es_threads); - r600_pipe_state_add_reg(rstate, R_008C18_SQ_THREAD_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C18_SQ_THREAD_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL, 0); tmp = 0; tmp |= S_008C1C_NUM_HS_THREADS(num_hs_threads); tmp |= S_008C1C_NUM_LS_THREADS(num_ls_threads); - r600_pipe_state_add_reg(rstate, R_008C1C_SQ_THREAD_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C1C_SQ_THREAD_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL, 0); tmp = 0; tmp |= S_008C20_NUM_PS_STACK_ENTRIES(num_ps_stack_entries); tmp |= S_008C20_NUM_VS_STACK_ENTRIES(num_vs_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C20_SQ_STACK_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C20_SQ_STACK_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL, 0); tmp = 0; tmp |= S_008C24_NUM_GS_STACK_ENTRIES(num_gs_stack_entries); tmp |= S_008C24_NUM_ES_STACK_ENTRIES(num_es_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C24_SQ_STACK_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C24_SQ_STACK_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL, 0); tmp = 0; tmp |= S_008C28_NUM_HS_STACK_ENTRIES(num_hs_stack_entries); tmp |= S_008C28_NUM_LS_STACK_ENTRIES(num_ls_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C28_SQ_STACK_RESOURCE_MGMT_3, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C28_SQ_STACK_RESOURCE_MGMT_3, tmp, 0xFFFFFFFF, NULL, 0); tmp = 0; tmp |= S_008E2C_NUM_PS_LDS(0x1000); tmp |= S_008E2C_NUM_LS_LDS(0x1000); - r600_pipe_state_add_reg(rstate, R_008E2C_SQ_LDS_RESOURCE_MGMT, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008E2C_SQ_LDS_RESOURCE_MGMT, tmp, 0xFFFFFFFF, NULL, 0); - r600_pipe_state_add_reg(rstate, R_009100_SPI_CONFIG_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_00913C_SPI_CONFIG_CNTL_1, S_00913C_VTX_DONE_DELAY(4), 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_009100_SPI_CONFIG_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_00913C_SPI_CONFIG_CNTL_1, S_00913C_VTX_DONE_DELAY(4), 0xFFFFFFFF, NULL, 0); #if 0 - r600_pipe_state_add_reg(rstate, R_028350_SX_MISC, 0x0, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028350_SX_MISC, 0x0, 0xFFFFFFFF, NULL, 0); - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x0, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x0, 0xFFFFFFFF, NULL, 0); #endif - r600_pipe_state_add_reg(rstate, R_028A48_PA_SC_MODE_CNTL_0, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL_1, 0x0, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_028900_SQ_ESGS_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028904_SQ_GSVS_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028908_SQ_ESTMP_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02890C_SQ_GSTMP_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028910_SQ_VSTMP_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028914_SQ_PSTMP_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_02891C_SQ_GS_VERT_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028920_SQ_GS_VERT_ITEMSIZE_1, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028924_SQ_GS_VERT_ITEMSIZE_2, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028928_SQ_GS_VERT_ITEMSIZE_3, 0x0, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A14_VGT_HOS_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A20_VGT_HOS_REUSE_DEPTH, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A24_VGT_GROUP_PRIM_TYPE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A28_VGT_GROUP_FIRST_DECR, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A2C_VGT_GROUP_DECR, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A30_VGT_GROUP_VECT_0_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A34_VGT_GROUP_VECT_1_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A40_VGT_GS_MODE, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028B94_VGT_STRMOUT_CONFIG, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AB4_VGT_REUSE_OFF, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AB8_VGT_VTX_CNT_EN, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_028380_SQ_VTX_SEMANTIC_0, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028384_SQ_VTX_SEMANTIC_1, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028388_SQ_VTX_SEMANTIC_2, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02838C_SQ_VTX_SEMANTIC_3, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028390_SQ_VTX_SEMANTIC_4, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028394_SQ_VTX_SEMANTIC_5, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028398_SQ_VTX_SEMANTIC_6, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02839C_SQ_VTX_SEMANTIC_7, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283A0_SQ_VTX_SEMANTIC_8, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283A4_SQ_VTX_SEMANTIC_9, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283A8_SQ_VTX_SEMANTIC_10, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283AC_SQ_VTX_SEMANTIC_11, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283B0_SQ_VTX_SEMANTIC_12, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283B4_SQ_VTX_SEMANTIC_13, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283B8_SQ_VTX_SEMANTIC_14, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283BC_SQ_VTX_SEMANTIC_15, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283C0_SQ_VTX_SEMANTIC_16, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283C4_SQ_VTX_SEMANTIC_17, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283C8_SQ_VTX_SEMANTIC_18, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283CC_SQ_VTX_SEMANTIC_19, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283D0_SQ_VTX_SEMANTIC_20, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283D4_SQ_VTX_SEMANTIC_21, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283D8_SQ_VTX_SEMANTIC_22, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283DC_SQ_VTX_SEMANTIC_23, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283E0_SQ_VTX_SEMANTIC_24, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283E4_SQ_VTX_SEMANTIC_25, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283E8_SQ_VTX_SEMANTIC_26, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283EC_SQ_VTX_SEMANTIC_27, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283F0_SQ_VTX_SEMANTIC_28, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283F4_SQ_VTX_SEMANTIC_29, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283F8_SQ_VTX_SEMANTIC_30, 0x0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0283FC_SQ_VTX_SEMANTIC_31, 0x0, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_028810_PA_CL_CLIP_CNTL, 0x0, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028A48_PA_SC_MODE_CNTL_0, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL_1, 0x0, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_028900_SQ_ESGS_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028904_SQ_GSVS_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028908_SQ_ESTMP_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02890C_SQ_GSTMP_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028910_SQ_VSTMP_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028914_SQ_PSTMP_RING_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_02891C_SQ_GS_VERT_ITEMSIZE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028920_SQ_GS_VERT_ITEMSIZE_1, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028924_SQ_GS_VERT_ITEMSIZE_2, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028928_SQ_GS_VERT_ITEMSIZE_3, 0x0, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A14_VGT_HOS_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A20_VGT_HOS_REUSE_DEPTH, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A24_VGT_GROUP_PRIM_TYPE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A28_VGT_GROUP_FIRST_DECR, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A2C_VGT_GROUP_DECR, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A30_VGT_GROUP_VECT_0_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A34_VGT_GROUP_VECT_1_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A40_VGT_GS_MODE, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028B94_VGT_STRMOUT_CONFIG, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AB4_VGT_REUSE_OFF, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AB8_VGT_VTX_CNT_EN, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_028380_SQ_VTX_SEMANTIC_0, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028384_SQ_VTX_SEMANTIC_1, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028388_SQ_VTX_SEMANTIC_2, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02838C_SQ_VTX_SEMANTIC_3, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028390_SQ_VTX_SEMANTIC_4, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028394_SQ_VTX_SEMANTIC_5, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028398_SQ_VTX_SEMANTIC_6, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02839C_SQ_VTX_SEMANTIC_7, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283A0_SQ_VTX_SEMANTIC_8, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283A4_SQ_VTX_SEMANTIC_9, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283A8_SQ_VTX_SEMANTIC_10, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283AC_SQ_VTX_SEMANTIC_11, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283B0_SQ_VTX_SEMANTIC_12, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283B4_SQ_VTX_SEMANTIC_13, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283B8_SQ_VTX_SEMANTIC_14, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283BC_SQ_VTX_SEMANTIC_15, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283C0_SQ_VTX_SEMANTIC_16, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283C4_SQ_VTX_SEMANTIC_17, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283C8_SQ_VTX_SEMANTIC_18, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283CC_SQ_VTX_SEMANTIC_19, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283D0_SQ_VTX_SEMANTIC_20, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283D4_SQ_VTX_SEMANTIC_21, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283D8_SQ_VTX_SEMANTIC_22, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283DC_SQ_VTX_SEMANTIC_23, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283E0_SQ_VTX_SEMANTIC_24, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283E4_SQ_VTX_SEMANTIC_25, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283E8_SQ_VTX_SEMANTIC_26, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283EC_SQ_VTX_SEMANTIC_27, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283F0_SQ_VTX_SEMANTIC_28, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283F4_SQ_VTX_SEMANTIC_29, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283F8_SQ_VTX_SEMANTIC_30, 0x0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0283FC_SQ_VTX_SEMANTIC_31, 0x0, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_028810_PA_CL_CLIP_CNTL, 0x0, 0xFFFFFFFF, NULL, 0); r600_context_pipe_state_set(&rctx->ctx, rstate); } @@ -2128,6 +2123,7 @@ void evergreen_polygon_offset_update(struct r600_pipe_context *rctx) offset_units *= 2.0f; break; case PIPE_FORMAT_Z32_FLOAT: + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: depth = -23; offset_units *= 1.0f; offset_db_fmt_cntl |= S_028B78_POLY_OFFSET_DB_IS_FLOAT_FMT(1); @@ -2143,19 +2139,19 @@ void evergreen_polygon_offset_update(struct r600_pipe_context *rctx) offset_db_fmt_cntl |= S_028B78_POLY_OFFSET_NEG_NUM_DB_BITS(depth); r600_pipe_state_add_reg(&state, R_028B80_PA_SU_POLY_OFFSET_FRONT_SCALE, - fui(rctx->rasterizer->offset_scale), 0xFFFFFFFF, NULL); + fui(rctx->rasterizer->offset_scale), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&state, R_028B84_PA_SU_POLY_OFFSET_FRONT_OFFSET, - fui(offset_units), 0xFFFFFFFF, NULL); + fui(offset_units), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&state, R_028B88_PA_SU_POLY_OFFSET_BACK_SCALE, - fui(rctx->rasterizer->offset_scale), 0xFFFFFFFF, NULL); + fui(rctx->rasterizer->offset_scale), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&state, R_028B8C_PA_SU_POLY_OFFSET_BACK_OFFSET, - fui(offset_units), 0xFFFFFFFF, NULL); + fui(offset_units), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&state, R_028B78_PA_SU_POLY_OFFSET_DB_FMT_CNTL, - offset_db_fmt_cntl, 0xFFFFFFFF, NULL); + offset_db_fmt_cntl, 0xFFFFFFFF, NULL, 0); r600_context_pipe_state_set(&rctx->ctx, &state); } } @@ -2252,32 +2248,32 @@ void evergreen_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader S_0286E0_LINEAR_CENTROID_ENA(have_centroid); r600_pipe_state_add_reg(rstate, R_0286CC_SPI_PS_IN_CONTROL_0, - spi_ps_in_control_0, 0xFFFFFFFF, NULL); + spi_ps_in_control_0, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0286D0_SPI_PS_IN_CONTROL_1, - spi_ps_in_control_1, 0xFFFFFFFF, NULL); + spi_ps_in_control_1, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0286E4_SPI_PS_IN_CONTROL_2, - 0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0286D8_SPI_INPUT_Z, spi_input_z, 0xFFFFFFFF, NULL); + 0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0286D8_SPI_INPUT_Z, spi_input_z, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0286E0_SPI_BARYC_CNTL, spi_baryc_cntl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028840_SQ_PGM_START_PS, - (r600_bo_offset(shader->bo)) >> 8, 0xFFFFFFFF, shader->bo); + 0, 0xFFFFFFFF, shader->bo, RADEON_USAGE_READ); r600_pipe_state_add_reg(rstate, R_028844_SQ_PGM_RESOURCES_PS, S_028844_NUM_GPRS(rshader->bc.ngpr) | S_028844_PRIME_CACHE_ON_DRAW(1) | S_028844_STACK_SIZE(rshader->bc.nstack), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028848_SQ_PGM_RESOURCES_2_PS, - 0x0, 0xFFFFFFFF, NULL); + 0x0, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_02884C_SQ_PGM_EXPORTS_PS, - exports_ps, 0xFFFFFFFF, NULL); + exports_ps, 0xFFFFFFFF, NULL, 0); /* FIXME: Evergreen doesn't seem to support MULTIWRITE_ENABLE. */ /* only set some bits here, the other bits are set in the dsa state */ r600_pipe_state_add_reg(rstate, @@ -2286,10 +2282,10 @@ void evergreen_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader S_02880C_Z_EXPORT_ENABLE(1) | S_02880C_STENCIL_EXPORT_ENABLE(1) | S_02880C_KILL_ENABLE(1), - NULL); + NULL, 0); r600_pipe_state_add_reg(rstate, R_03A200_SQ_LOOP_CONST_0, 0x01000FFF, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); } void evergreen_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shader) @@ -2298,7 +2294,7 @@ void evergreen_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader struct r600_pipe_state *rstate = &shader->rstate; struct r600_shader *rshader = &shader->shader; unsigned spi_vs_out_id[10]; - unsigned i, tmp; + unsigned i, tmp, nparams; /* clear previous register */ rstate->nregs = 0; @@ -2314,28 +2310,36 @@ void evergreen_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader for (i = 0; i < 10; i++) { r600_pipe_state_add_reg(rstate, R_02861C_SPI_VS_OUT_ID_0 + i * 4, - spi_vs_out_id[i], 0xFFFFFFFF, NULL); + spi_vs_out_id[i], 0xFFFFFFFF, NULL, 0); } + /* Certain attributes (position, psize, etc.) don't count as params. + * VS is required to export at least one param and r600_shader_from_tgsi() + * takes care of adding a dummy export. + */ + nparams = rshader->noutput - rshader->npos; + if (nparams < 1) + nparams = 1; + r600_pipe_state_add_reg(rstate, R_0286C4_SPI_VS_OUT_CONFIG, - S_0286C4_VS_EXPORT_COUNT(rshader->noutput - 2), - 0xFFFFFFFF, NULL); + S_0286C4_VS_EXPORT_COUNT(nparams - 1), + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028860_SQ_PGM_RESOURCES_VS, S_028860_NUM_GPRS(rshader->bc.ngpr) | S_028860_STACK_SIZE(rshader->bc.nstack), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028864_SQ_PGM_RESOURCES_2_VS, - 0x0, 0xFFFFFFFF, NULL); + 0x0, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_02885C_SQ_PGM_START_VS, - (r600_bo_offset(shader->bo)) >> 8, 0xFFFFFFFF, shader->bo); + 0, 0xFFFFFFFF, shader->bo, RADEON_USAGE_READ); r600_pipe_state_add_reg(rstate, R_03A200_SQ_LOOP_CONST_0 + (32 * 4), 0x01000FFF, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); } void evergreen_fetch_shader(struct pipe_context *ctx, @@ -2346,10 +2350,10 @@ void evergreen_fetch_shader(struct pipe_context *ctx, rstate->id = R600_PIPE_STATE_FETCH_SHADER; rstate->nregs = 0; r600_pipe_state_add_reg(rstate, R_0288A8_SQ_PGM_RESOURCES_FS, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0288A4_SQ_PGM_START_FS, - (r600_bo_offset(ve->fetch_shader)) >> 8, - 0xFFFFFFFF, ve->fetch_shader); + 0, + 0xFFFFFFFF, ve->fetch_shader, RADEON_USAGE_READ); } void *evergreen_create_db_flush_dsa(struct r600_pipe_context *rctx) @@ -2363,7 +2367,7 @@ void *evergreen_create_db_flush_dsa(struct r600_pipe_context *rctx) r600_pipe_state_add_reg(rstate, R_02880C_DB_SHADER_CONTROL, 0x0, - S_02880C_DUAL_EXPORT_ENABLE(1), NULL); + S_02880C_DUAL_EXPORT_ENABLE(1), NULL, 0); r600_pipe_state_add_reg(rstate, R_028000_DB_RENDER_CONTROL, S_028000_DEPTH_COPY_ENABLE(1) | @@ -2371,7 +2375,7 @@ void *evergreen_create_db_flush_dsa(struct r600_pipe_context *rctx) S_028000_COPY_CENTROID(1), S_028000_DEPTH_COPY_ENABLE(1) | S_028000_STENCIL_COPY_ENABLE(1) | - S_028000_COPY_CENTROID(1), NULL); + S_028000_COPY_CENTROID(1), NULL, 0); return rstate; } @@ -2397,9 +2401,11 @@ void evergreen_pipe_init_buffer_resource(struct r600_pipe_context *rctx, void evergreen_pipe_mod_buffer_resource(struct r600_pipe_resource_state *rstate, struct r600_resource *rbuffer, - unsigned offset, unsigned stride) + unsigned offset, unsigned stride, + enum radeon_bo_usage usage) { rstate->bo[0] = rbuffer->bo; + rstate->bo_usage[0] = usage; rstate->val[0] = offset; rstate->val[1] = rbuffer->bo_size - offset - 1; rstate->val[2] = S_030008_ENDIAN_SWAP(r600_endian_swap(32)) | diff --git a/src/gallium/drivers/r600/evergreend.h b/src/gallium/drivers/r600/evergreend.h index 96dbd4da91b..9a8c353e4ee 100644 --- a/src/gallium/drivers/r600/evergreend.h +++ b/src/gallium/drivers/r600/evergreend.h @@ -1697,6 +1697,10 @@ #define R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL 0x00028A3C #define R_028A48_PA_SC_MODE_CNTL_0 0x00028A48 #define R_028A4C_PA_SC_MODE_CNTL_1 0x00028A4C +#define R_028A94_VGT_MULTI_PRIM_IB_RESET_EN 0x00028A94 +#define S_028A94_RESET_EN(x) (((x) & 0x1) << 0) +#define G_028A94_RESET_EN(x) (((x) >> 0) & 0x1) +#define C_028A94_RESET_EN 0xFFFFFFFE #define R_028AB4_VGT_REUSE_OFF 0x00028AB4 #define R_028AB8_VGT_VTX_CNT_EN 0x00028AB8 #define R_028ABC_DB_HTILE_SURFACE 0x00028ABC diff --git a/src/gallium/drivers/r600/r600.h b/src/gallium/drivers/r600/r600.h index 61adc7ed988..f24146edcf1 100644 --- a/src/gallium/drivers/r600/r600.h +++ b/src/gallium/drivers/r600/r600.h @@ -26,14 +26,8 @@ #ifndef R600_H #define R600_H -#include <assert.h> -#include <stdint.h> -#include <stdio.h> -#include <util/u_double_list.h> -#include <util/u_inlines.h> -#include <pipe/p_compiler.h> - -#define RADEON_CTX_MAX_PM4 (64 * 1024 / 4) +#include "../../winsys/radeon/drm/radeon_winsys.h" +#include "util/u_double_list.h" #define R600_ERR(fmt, args...) \ fprintf(stderr, "EE %s:%d %s - "fmt, __FILE__, __LINE__, __func__, ##args) @@ -94,31 +88,32 @@ struct r600_tiling_info *r600_get_tiling_info(struct radeon *radeon); unsigned r600_get_clock_crystal_freq(struct radeon *radeon); unsigned r600_get_minor_version(struct radeon *radeon); unsigned r600_get_num_backends(struct radeon *radeon); +unsigned r600_get_num_tile_pipes(struct radeon *radeon); +unsigned r600_get_backend_map(struct radeon *radeon); /* r600_bo.c */ struct r600_bo; +struct radeon_winsys_cs; + struct r600_bo *r600_bo(struct radeon *radeon, unsigned size, unsigned alignment, unsigned binding, unsigned usage); -struct r600_bo *r600_bo_handle(struct radeon *radeon, - unsigned handle, unsigned *array_mode); -void *r600_bo_map(struct radeon *radeon, struct r600_bo *bo, unsigned usage, void *ctx); +struct r600_bo *r600_bo_handle(struct radeon *radeon, struct winsys_handle *whandle, + unsigned *stride, unsigned *array_mode); +void *r600_bo_map(struct radeon *radeon, struct r600_bo *bo, struct radeon_winsys_cs *cs, unsigned usage); void r600_bo_unmap(struct radeon *radeon, struct r600_bo *bo); boolean r600_bo_get_winsys_handle(struct radeon *radeon, struct r600_bo *pb_bo, - unsigned stride, struct winsys_handle *whandle); -static INLINE unsigned r600_bo_offset(struct r600_bo *bo) -{ - return 0; -} -void r600_bo_destroy(struct radeon *radeon, struct r600_bo *bo); + unsigned stride, struct winsys_handle *whandle); + +void r600_bo_destroy(struct r600_bo *bo); /* this relies on the pipe_reference being the first member of r600_bo */ -static INLINE void r600_bo_reference(struct radeon *radeon, struct r600_bo **dst, struct r600_bo *src) +static INLINE void r600_bo_reference(struct r600_bo **dst, struct r600_bo *src) { struct r600_bo *old = *dst; if (pipe_reference((struct pipe_reference *)(*dst), (struct pipe_reference *)src)) { - r600_bo_destroy(radeon, old); + r600_bo_destroy(old); } *dst = src; } @@ -145,6 +140,7 @@ struct r600_pipe_reg { u32 mask; struct r600_block *block; struct r600_bo *bo; + enum radeon_bo_usage bo_usage; u32 id; }; @@ -157,7 +153,8 @@ struct r600_pipe_state { struct r600_pipe_resource_state { unsigned id; u32 val[8]; - struct r600_bo *bo[2]; + struct r600_bo *bo[2]; + enum radeon_bo_usage bo_usage[2]; /* XXX set these */ }; #define R600_BLOCK_STATUS_ENABLED (1 << 0) @@ -168,6 +165,7 @@ struct r600_pipe_resource_state { struct r600_block_reloc { struct r600_bo *bo; + enum radeon_bo_usage bo_usage; unsigned flush_flags; unsigned flush_mask; unsigned bo_pm4_index; @@ -195,18 +193,6 @@ struct r600_range { }; /* - * relocation - */ -#pragma pack(1) -struct r600_reloc { - uint32_t handle; - uint32_t read_domain; - uint32_t write_domain; - uint32_t flags; -}; -#pragma pack() - -/* * query */ struct r600_query { @@ -243,6 +229,8 @@ struct r600_query { struct r600_context { struct radeon *radeon; + struct radeon_winsys_cs *cs; + struct r600_range *range; unsigned nblocks; struct r600_block **blocks; @@ -250,18 +238,19 @@ struct r600_context { struct list_head resource_dirty; struct list_head enable_list; unsigned pm4_ndwords; - unsigned pm4_cdwords; unsigned pm4_dirty_cdwords; unsigned ctx_pm4_ndwords; unsigned init_dwords; - unsigned nreloc; + unsigned creloc; - struct r600_reloc *reloc; - struct radeon_bo **bo; + struct r600_bo **bo; + u32 *pm4; + unsigned pm4_cdwords; + struct list_head query_list; unsigned num_query_running; - struct list_head fenced_bo; + unsigned backend_mask; unsigned max_db; /* for OQ */ unsigned num_dest_buffers; unsigned flags; @@ -282,6 +271,7 @@ struct r600_draw { struct r600_bo *indices; }; +void r600_get_backend_mask(struct r600_context *ctx); int r600_context_init(struct r600_context *ctx, struct radeon *radeon); void r600_context_fini(struct r600_context *ctx); void r600_context_pipe_state_set(struct r600_context *ctx, struct r600_pipe_state *state); @@ -290,8 +280,7 @@ void r600_context_pipe_state_set_vs_resource(struct r600_context *ctx, struct r6 void r600_context_pipe_state_set_fs_resource(struct r600_context *ctx, struct r600_pipe_resource_state *state, unsigned rid); void r600_context_pipe_state_set_ps_sampler(struct r600_context *ctx, struct r600_pipe_state *state, unsigned id); void r600_context_pipe_state_set_vs_sampler(struct r600_context *ctx, struct r600_pipe_state *state, unsigned id); -void r600_context_flush(struct r600_context *ctx); -void r600_context_dump_bof(struct r600_context *ctx, const char *file); +void r600_context_flush(struct r600_context *ctx, unsigned flags); void r600_context_draw(struct r600_context *ctx, const struct r600_draw *draw); struct r600_query *r600_context_query_create(struct r600_context *ctx, unsigned query_type); @@ -319,18 +308,21 @@ void evergreen_context_pipe_state_set_fs_resource(struct r600_context *ctx, stru void evergreen_context_pipe_state_set_ps_sampler(struct r600_context *ctx, struct r600_pipe_state *state, unsigned id); void evergreen_context_pipe_state_set_vs_sampler(struct r600_context *ctx, struct r600_pipe_state *state, unsigned id); -struct radeon *radeon_decref(struct radeon *radeon); +struct radeon *radeon_destroy(struct radeon *radeon); void _r600_pipe_state_add_reg(struct r600_context *ctx, struct r600_pipe_state *state, u32 offset, u32 value, u32 mask, u32 range_id, u32 block_id, - struct r600_bo *bo); + struct r600_bo *bo, + enum radeon_bo_usage usage); void r600_pipe_state_add_reg_noblock(struct r600_pipe_state *state, u32 offset, u32 value, u32 mask, - struct r600_bo *bo); -#define r600_pipe_state_add_reg(state, offset, value, mask, bo) _r600_pipe_state_add_reg(&rctx->ctx, state, offset, value, mask, CTX_RANGE_ID(offset), CTX_BLOCK_ID(offset), bo) + struct r600_bo *bo, + enum radeon_bo_usage usage); + +#define r600_pipe_state_add_reg(state, offset, value, mask, bo, usage) _r600_pipe_state_add_reg(&rctx->ctx, state, offset, value, mask, CTX_RANGE_ID(offset), CTX_BLOCK_ID(offset), bo, usage) static inline void r600_pipe_state_mod_reg(struct r600_pipe_state *state, u32 value) @@ -340,10 +332,12 @@ static inline void r600_pipe_state_mod_reg(struct r600_pipe_state *state, } static inline void r600_pipe_state_mod_reg_bo(struct r600_pipe_state *state, - u32 value, struct r600_bo *bo) + u32 value, struct r600_bo *bo, + enum radeon_bo_usage usage) { state->regs[state->nregs].value = value; state->regs[state->nregs].bo = bo; + state->regs[state->nregs].bo_usage = usage; state->nregs++; } diff --git a/src/gallium/drivers/r600/r600_asm.c b/src/gallium/drivers/r600/r600_asm.c index 5fae2b00c8b..27febdf9d03 100644 --- a/src/gallium/drivers/r600/r600_asm.c +++ b/src/gallium/drivers/r600/r600_asm.c @@ -36,7 +36,7 @@ #define NUM_OF_CYCLES 3 #define NUM_OF_COMPONENTS 4 -static inline unsigned int r600_bc_get_num_operands(struct r600_bc *bc, struct r600_bc_alu *alu) +static inline unsigned int r600_bytecode_get_num_operands(struct r600_bytecode *bc, struct r600_bytecode_alu *alu) { if(alu->is_op3) return 3; @@ -88,6 +88,7 @@ static inline unsigned int r600_bc_get_num_operands(struct r600_bc *bc, struct r case V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_INT_TO_FLT: case V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN: case V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS: + case V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RNDNE: return 1; default: R600_ERR( "Need instruction operand number for 0x%x.\n", alu->inst); @@ -140,6 +141,7 @@ static inline unsigned int r600_bc_get_num_operands(struct r600_bc *bc, struct r case EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_INT_TO_FLT: case EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN: case EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS: + case EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RNDNE: return 1; default: R600_ERR( "Need instruction operand number for 0x%x.\n", alu->inst); @@ -150,11 +152,11 @@ static inline unsigned int r600_bc_get_num_operands(struct r600_bc *bc, struct r return 3; } -int r700_bc_alu_build(struct r600_bc *bc, struct r600_bc_alu *alu, unsigned id); +int r700_bytecode_alu_build(struct r600_bytecode *bc, struct r600_bytecode_alu *alu, unsigned id); -static struct r600_bc_cf *r600_bc_cf(void) +static struct r600_bytecode_cf *r600_bytecode_cf(void) { - struct r600_bc_cf *cf = CALLOC_STRUCT(r600_bc_cf); + struct r600_bytecode_cf *cf = CALLOC_STRUCT(r600_bytecode_cf); if (cf == NULL) return NULL; @@ -165,9 +167,9 @@ static struct r600_bc_cf *r600_bc_cf(void) return cf; } -static struct r600_bc_alu *r600_bc_alu(void) +static struct r600_bytecode_alu *r600_bytecode_alu(void) { - struct r600_bc_alu *alu = CALLOC_STRUCT(r600_bc_alu); + struct r600_bytecode_alu *alu = CALLOC_STRUCT(r600_bytecode_alu); if (alu == NULL) return NULL; @@ -175,9 +177,9 @@ static struct r600_bc_alu *r600_bc_alu(void) return alu; } -static struct r600_bc_vtx *r600_bc_vtx(void) +static struct r600_bytecode_vtx *r600_bytecode_vtx(void) { - struct r600_bc_vtx *vtx = CALLOC_STRUCT(r600_bc_vtx); + struct r600_bytecode_vtx *vtx = CALLOC_STRUCT(r600_bytecode_vtx); if (vtx == NULL) return NULL; @@ -185,9 +187,9 @@ static struct r600_bc_vtx *r600_bc_vtx(void) return vtx; } -static struct r600_bc_tex *r600_bc_tex(void) +static struct r600_bytecode_tex *r600_bytecode_tex(void) { - struct r600_bc_tex *tex = CALLOC_STRUCT(r600_bc_tex); + struct r600_bytecode_tex *tex = CALLOC_STRUCT(r600_bytecode_tex); if (tex == NULL) return NULL; @@ -195,15 +197,15 @@ static struct r600_bc_tex *r600_bc_tex(void) return tex; } -void r600_bc_init(struct r600_bc *bc, enum chip_class chip_class) +void r600_bytecode_init(struct r600_bytecode *bc, enum chip_class chip_class) { LIST_INITHEAD(&bc->cf); bc->chip_class = chip_class; } -static int r600_bc_add_cf(struct r600_bc *bc) +static int r600_bytecode_add_cf(struct r600_bytecode *bc) { - struct r600_bc_cf *cf = r600_bc_cf(); + struct r600_bytecode_cf *cf = r600_bytecode_cf(); if (cf == NULL) return -ENOMEM; @@ -217,7 +219,7 @@ static int r600_bc_add_cf(struct r600_bc *bc) return 0; } -int r600_bc_add_output(struct r600_bc *bc, const struct r600_bc_output *output) +int r600_bytecode_add_output(struct r600_bytecode *bc, const struct r600_bytecode_output *output) { int r; @@ -252,16 +254,16 @@ int r600_bc_add_output(struct r600_bc *bc, const struct r600_bc_output *output) } } - r = r600_bc_add_cf(bc); + r = r600_bytecode_add_cf(bc); if (r) return r; bc->cf_last->inst = output->inst; - memcpy(&bc->cf_last->output, output, sizeof(struct r600_bc_output)); + memcpy(&bc->cf_last->output, output, sizeof(struct r600_bytecode_output)); return 0; } /* alu instructions that can ony exits once per group */ -static int is_alu_once_inst(struct r600_bc *bc, struct r600_bc_alu *alu) +static int is_alu_once_inst(struct r600_bytecode *bc, struct r600_bytecode_alu *alu) { switch (bc->chip_class) { case R600: @@ -342,7 +344,7 @@ static int is_alu_once_inst(struct r600_bc *bc, struct r600_bc_alu *alu) } } -static int is_alu_reduction_inst(struct r600_bc *bc, struct r600_bc_alu *alu) +static int is_alu_reduction_inst(struct r600_bytecode *bc, struct r600_bytecode_alu *alu) { switch (bc->chip_class) { case R600: @@ -363,7 +365,7 @@ static int is_alu_reduction_inst(struct r600_bc *bc, struct r600_bc_alu *alu) } } -static int is_alu_cube_inst(struct r600_bc *bc, struct r600_bc_alu *alu) +static int is_alu_cube_inst(struct r600_bytecode *bc, struct r600_bytecode_alu *alu) { switch (bc->chip_class) { case R600: @@ -378,7 +380,7 @@ static int is_alu_cube_inst(struct r600_bc *bc, struct r600_bc_alu *alu) } } -static int is_alu_mova_inst(struct r600_bc *bc, struct r600_bc_alu *alu) +static int is_alu_mova_inst(struct r600_bytecode *bc, struct r600_bytecode_alu *alu) { switch (bc->chip_class) { case R600: @@ -396,7 +398,7 @@ static int is_alu_mova_inst(struct r600_bc *bc, struct r600_bc_alu *alu) } /* alu instructions that can only execute on the vector unit */ -static int is_alu_vec_unit_inst(struct r600_bc *bc, struct r600_bc_alu *alu) +static int is_alu_vec_unit_inst(struct r600_bytecode *bc, struct r600_bytecode_alu *alu) { return is_alu_reduction_inst(bc, alu) || is_alu_mova_inst(bc, alu) || @@ -405,7 +407,7 @@ static int is_alu_vec_unit_inst(struct r600_bc *bc, struct r600_bc_alu *alu) } /* alu instructions that can only execute on the trans unit */ -static int is_alu_trans_unit_inst(struct r600_bc *bc, struct r600_bc_alu *alu) +static int is_alu_trans_unit_inst(struct r600_bytecode *bc, struct r600_bytecode_alu *alu) { switch (bc->chip_class) { case R600: @@ -476,23 +478,23 @@ static int is_alu_trans_unit_inst(struct r600_bc *bc, struct r600_bc_alu *alu) } /* alu instructions that can execute on any unit */ -static int is_alu_any_unit_inst(struct r600_bc *bc, struct r600_bc_alu *alu) +static int is_alu_any_unit_inst(struct r600_bytecode *bc, struct r600_bytecode_alu *alu) { return !is_alu_vec_unit_inst(bc, alu) && !is_alu_trans_unit_inst(bc, alu); } -static int assign_alu_units(struct r600_bc *bc, struct r600_bc_alu *alu_first, - struct r600_bc_alu *assignment[5]) +static int assign_alu_units(struct r600_bytecode *bc, struct r600_bytecode_alu *alu_first, + struct r600_bytecode_alu *assignment[5]) { - struct r600_bc_alu *alu; + struct r600_bytecode_alu *alu; unsigned i, chan, trans; int max_slots = bc->chip_class == CAYMAN ? 4 : 5; for (i = 0; i < max_slots; i++) assignment[i] = NULL; - for (alu = alu_first; alu; alu = LIST_ENTRY(struct r600_bc_alu, alu->list.next, list)) { + for (alu = alu_first; alu; alu = LIST_ENTRY(struct r600_bytecode_alu, alu->list.next, list)) { chan = alu->dst.chan; if (max_slots == 4) trans = 0; @@ -571,7 +573,7 @@ static int reserve_gpr(struct alu_bank_swizzle *bs, unsigned sel, unsigned chan, return 0; } -static int reserve_cfile(struct r600_bc *bc, struct alu_bank_swizzle *bs, unsigned sel, unsigned chan) +static int reserve_cfile(struct r600_bytecode *bc, struct alu_bank_swizzle *bs, unsigned sel, unsigned chan) { int res, num_res = 4; if (bc->chip_class >= R700) { @@ -613,12 +615,12 @@ static int is_const(int sel) sel <= V_SQ_ALU_SRC_LITERAL); } -static int check_vector(struct r600_bc *bc, struct r600_bc_alu *alu, +static int check_vector(struct r600_bytecode *bc, struct r600_bytecode_alu *alu, struct alu_bank_swizzle *bs, int bank_swizzle) { int r, src, num_src, sel, elem, cycle; - num_src = r600_bc_get_num_operands(bc, alu); + num_src = r600_bytecode_get_num_operands(bc, alu); for (src = 0; src < num_src; src++) { sel = alu->src[src].sel; elem = alu->src[src].chan; @@ -643,12 +645,12 @@ static int check_vector(struct r600_bc *bc, struct r600_bc_alu *alu, return 0; } -static int check_scalar(struct r600_bc *bc, struct r600_bc_alu *alu, +static int check_scalar(struct r600_bytecode *bc, struct r600_bytecode_alu *alu, struct alu_bank_swizzle *bs, int bank_swizzle) { int r, src, num_src, const_count, sel, elem, cycle; - num_src = r600_bc_get_num_operands(bc, alu); + num_src = r600_bytecode_get_num_operands(bc, alu); for (const_count = 0, src = 0; src < num_src; ++src) { sel = alu->src[src].sel; elem = alu->src[src].chan; @@ -689,20 +691,24 @@ static int check_scalar(struct r600_bc *bc, struct r600_bc_alu *alu, return 0; } -static int check_and_set_bank_swizzle(struct r600_bc *bc, - struct r600_bc_alu *slots[5]) +static int check_and_set_bank_swizzle(struct r600_bytecode *bc, + struct r600_bytecode_alu *slots[5]) { struct alu_bank_swizzle bs; int bank_swizzle[5]; - int i, r = 0, forced = 0; + int i, r = 0, forced = 1; boolean scalar_only = bc->chip_class == CAYMAN ? false : true; int max_slots = bc->chip_class == CAYMAN ? 4 : 5; for (i = 0; i < max_slots; i++) { - if (slots[i] && slots[i]->bank_swizzle_force) { - slots[i]->bank_swizzle = slots[i]->bank_swizzle_force; - forced = 1; + if (slots[i]) { + if (slots[i]->bank_swizzle_force) { + slots[i]->bank_swizzle = slots[i]->bank_swizzle_force; + } else { + forced = 0; + } } + if (i < 4 && slots[i]) scalar_only = false; } @@ -712,7 +718,11 @@ static int check_and_set_bank_swizzle(struct r600_bc *bc, /* Just check every possible combination of bank swizzle. * Not very efficent, but works on the first try in most of the cases. */ for (i = 0; i < 4; i++) - bank_swizzle[i] = SQ_ALU_VEC_012; + if (!slots[i] || !slots[i]->bank_swizzle_force) + bank_swizzle[i] = SQ_ALU_VEC_012; + else + bank_swizzle[i] = slots[i]->bank_swizzle; + bank_swizzle[4] = SQ_ALU_SCL_210; while(bank_swizzle[4] <= SQ_ALU_SCL_221) { @@ -749,11 +759,13 @@ static int check_and_set_bank_swizzle(struct r600_bc *bc, bank_swizzle[4]++; } else { for (i = 0; i < max_slots; i++) { - bank_swizzle[i]++; - if (bank_swizzle[i] <= SQ_ALU_VEC_210) - break; - else - bank_swizzle[i] = SQ_ALU_VEC_012; + if (!slots[i] || !slots[i]->bank_swizzle_force) { + bank_swizzle[i]++; + if (bank_swizzle[i] <= SQ_ALU_VEC_210) + break; + else + bank_swizzle[i] = SQ_ALU_VEC_012; + } } } } @@ -762,10 +774,10 @@ static int check_and_set_bank_swizzle(struct r600_bc *bc, return -1; } -static int replace_gpr_with_pv_ps(struct r600_bc *bc, - struct r600_bc_alu *slots[5], struct r600_bc_alu *alu_prev) +static int replace_gpr_with_pv_ps(struct r600_bytecode *bc, + struct r600_bytecode_alu *slots[5], struct r600_bytecode_alu *alu_prev) { - struct r600_bc_alu *prev[5]; + struct r600_bytecode_alu *prev[5]; int gpr[5], chan[5]; int i, j, r, src, num_src; int max_slots = bc->chip_class == CAYMAN ? 4 : 5; @@ -775,7 +787,7 @@ static int replace_gpr_with_pv_ps(struct r600_bc *bc, return r; for (i = 0; i < max_slots; ++i) { - if(prev[i] && prev[i]->dst.write && !prev[i]->dst.rel) { + if (prev[i] && (prev[i]->dst.write || prev[i]->is_op3) && !prev[i]->dst.rel) { gpr[i] = prev[i]->dst.sel; /* cube writes more than PV.X */ if (!is_alu_cube_inst(bc, prev[i]) && is_alu_reduction_inst(bc, prev[i])) @@ -787,11 +799,11 @@ static int replace_gpr_with_pv_ps(struct r600_bc *bc, } for (i = 0; i < max_slots; ++i) { - struct r600_bc_alu *alu = slots[i]; + struct r600_bytecode_alu *alu = slots[i]; if(!alu) continue; - num_src = r600_bc_get_num_operands(bc, alu); + num_src = r600_bytecode_get_num_operands(bc, alu); for (src = 0; src < num_src; ++src) { if (!is_gpr(alu->src[src].sel) || alu->src[src].rel) continue; @@ -819,7 +831,7 @@ static int replace_gpr_with_pv_ps(struct r600_bc *bc, return 0; } -void r600_bc_special_constants(u32 value, unsigned *sel, unsigned *neg) +void r600_bytecode_special_constants(u32 value, unsigned *sel, unsigned *neg) { switch(value) { case 0: @@ -852,10 +864,10 @@ void r600_bc_special_constants(u32 value, unsigned *sel, unsigned *neg) } /* compute how many literal are needed */ -static int r600_bc_alu_nliterals(struct r600_bc *bc, struct r600_bc_alu *alu, +static int r600_bytecode_alu_nliterals(struct r600_bytecode *bc, struct r600_bytecode_alu *alu, uint32_t literal[4], unsigned *nliteral) { - unsigned num_src = r600_bc_get_num_operands(bc, alu); + unsigned num_src = r600_bytecode_get_num_operands(bc, alu); unsigned i, j; for (i = 0; i < num_src; ++i) { @@ -878,11 +890,11 @@ static int r600_bc_alu_nliterals(struct r600_bc *bc, struct r600_bc_alu *alu, return 0; } -static void r600_bc_alu_adjust_literals(struct r600_bc *bc, - struct r600_bc_alu *alu, +static void r600_bytecode_alu_adjust_literals(struct r600_bytecode *bc, + struct r600_bytecode_alu *alu, uint32_t literal[4], unsigned nliteral) { - unsigned num_src = r600_bc_get_num_operands(bc, alu); + unsigned num_src = r600_bytecode_get_num_operands(bc, alu); unsigned i, j; for (i = 0; i < num_src; ++i) { @@ -898,11 +910,11 @@ static void r600_bc_alu_adjust_literals(struct r600_bc *bc, } } -static int merge_inst_groups(struct r600_bc *bc, struct r600_bc_alu *slots[5], - struct r600_bc_alu *alu_prev) +static int merge_inst_groups(struct r600_bytecode *bc, struct r600_bytecode_alu *slots[5], + struct r600_bytecode_alu *alu_prev) { - struct r600_bc_alu *prev[5]; - struct r600_bc_alu *result[5] = { NULL }; + struct r600_bytecode_alu *prev[5]; + struct r600_bytecode_alu *result[5] = { NULL }; uint32_t literal[4], prev_literal[4]; unsigned nliteral = 0, prev_nliteral = 0; @@ -917,13 +929,13 @@ static int merge_inst_groups(struct r600_bc *bc, struct r600_bc_alu *slots[5], return r; for (i = 0; i < max_slots; ++i) { - struct r600_bc_alu *alu; + struct r600_bytecode_alu *alu; /* check number of literals */ if (prev[i]) { - if (r600_bc_alu_nliterals(bc, prev[i], literal, &nliteral)) + if (r600_bytecode_alu_nliterals(bc, prev[i], literal, &nliteral)) return 0; - if (r600_bc_alu_nliterals(bc, prev[i], prev_literal, &prev_nliteral)) + if (r600_bytecode_alu_nliterals(bc, prev[i], prev_literal, &prev_nliteral)) return 0; if (is_alu_mova_inst(bc, prev[i])) { if (have_rel) @@ -932,7 +944,7 @@ static int merge_inst_groups(struct r600_bc *bc, struct r600_bc_alu *slots[5], } num_once_inst += is_alu_once_inst(bc, prev[i]); } - if (slots[i] && r600_bc_alu_nliterals(bc, slots[i], literal, &nliteral)) + if (slots[i] && r600_bytecode_alu_nliterals(bc, slots[i], literal, &nliteral)) return 0; /* Let's check used slots. */ @@ -968,7 +980,7 @@ static int merge_inst_groups(struct r600_bc *bc, struct r600_bc_alu *slots[5], } /* Let's check source gprs */ - num_src = r600_bc_get_num_operands(bc, alu); + num_src = r600_bytecode_get_num_operands(bc, alu); for (src = 0; src < num_src; ++src) { if (alu->src[src].rel) { if (have_mova) @@ -1018,7 +1030,7 @@ static int merge_inst_groups(struct r600_bc *bc, struct r600_bc_alu *slots[5], } /* determine new last instruction */ - LIST_ENTRY(struct r600_bc_alu, bc->cf_last->alu.prev, list)->last = 1; + LIST_ENTRY(struct r600_bytecode_alu, bc->cf_last->alu.prev, list)->last = 1; /* determine new first instruction */ for (i = 0; i < max_slots; ++i) { @@ -1038,9 +1050,9 @@ static int merge_inst_groups(struct r600_bc *bc, struct r600_bc_alu *slots[5], * probably do slightly better by recognizing that we actually have two * consecutive lines of 16 constants, but the resulting code would also be * somewhat more complicated. */ -static int r600_bc_alloc_kcache_lines(struct r600_bc *bc, struct r600_bc_alu *alu, int type) +static int r600_bytecode_alloc_kcache_lines(struct r600_bytecode *bc, struct r600_bytecode_alu *alu, int type) { - struct r600_bc_kcache *kcache = bc->cf_last->kcache; + struct r600_bytecode_kcache *kcache = bc->cf_last->kcache; unsigned int required_lines; unsigned int free_lines = 0; unsigned int cache_line[3]; @@ -1093,7 +1105,7 @@ static int r600_bc_alloc_kcache_lines(struct r600_bc *bc, struct r600_bc_alu *al /* Start a new ALU clause if needed. */ if (required_lines > free_lines) { - if ((r = r600_bc_add_cf(bc))) { + if ((r = r600_bytecode_add_cf(bc))) { return r; } bc->cf_last->inst = (type << 3); @@ -1148,15 +1160,15 @@ static int r600_bc_alloc_kcache_lines(struct r600_bc *bc, struct r600_bc_alu *al return 0; } -int r600_bc_add_alu_type(struct r600_bc *bc, const struct r600_bc_alu *alu, int type) +int r600_bytecode_add_alu_type(struct r600_bytecode *bc, const struct r600_bytecode_alu *alu, int type) { - struct r600_bc_alu *nalu = r600_bc_alu(); - struct r600_bc_alu *lalu; + struct r600_bytecode_alu *nalu = r600_bytecode_alu(); + struct r600_bytecode_alu *lalu; int i, r; if (nalu == NULL) return -ENOMEM; - memcpy(nalu, alu, sizeof(struct r600_bc_alu)); + memcpy(nalu, alu, sizeof(struct r600_bytecode_alu)); if (bc->cf_last != NULL && bc->cf_last->inst != (type << 3)) { /* check if we could add it anyway */ @@ -1174,7 +1186,7 @@ int r600_bc_add_alu_type(struct r600_bc *bc, const struct r600_bc_alu *alu, int /* cf can contains only alu or only vtx or only tex */ if (bc->cf_last == NULL || bc->force_add_cf) { - r = r600_bc_add_cf(bc); + r = r600_bytecode_add_cf(bc); if (r) { free(nalu); return r; @@ -1184,7 +1196,7 @@ int r600_bc_add_alu_type(struct r600_bc *bc, const struct r600_bc_alu *alu, int /* Setup the kcache for this ALU instruction. This will start a new * ALU clause if needed. */ - if ((r = r600_bc_alloc_kcache_lines(bc, nalu, type))) { + if ((r = r600_bytecode_alloc_kcache_lines(bc, nalu, type))) { free(nalu); return r; } @@ -1198,7 +1210,7 @@ int r600_bc_add_alu_type(struct r600_bc *bc, const struct r600_bc_alu *alu, int bc->ngpr = nalu->src[i].sel + 1; } if (nalu->src[i].sel == V_SQ_ALU_SRC_LITERAL) - r600_bc_special_constants(nalu->src[i].value, + r600_bytecode_special_constants(nalu->src[i].value, &nalu->src[i].sel, &nalu->src[i].neg); } if (nalu->dst.sel >= bc->ngpr) { @@ -1213,7 +1225,7 @@ int r600_bc_add_alu_type(struct r600_bc *bc, const struct r600_bc_alu *alu, int if (nalu->last) { uint32_t literal[4]; unsigned nliteral; - struct r600_bc_alu *slots[5]; + struct r600_bytecode_alu *slots[5]; int max_slots = bc->chip_class == CAYMAN ? 4 : 5; r = assign_alu_units(bc, bc->cf_last->curr_bs_head, slots); if (r) @@ -1237,7 +1249,7 @@ int r600_bc_add_alu_type(struct r600_bc *bc, const struct r600_bc_alu *alu, int for (i = 0, nliteral = 0; i < max_slots; i++) { if (slots[i]) { - r = r600_bc_alu_nliterals(bc, slots[i], literal, &nliteral); + r = r600_bytecode_alu_nliterals(bc, slots[i], literal, &nliteral); if (r) return r; } @@ -1257,12 +1269,12 @@ int r600_bc_add_alu_type(struct r600_bc *bc, const struct r600_bc_alu *alu, int return 0; } -int r600_bc_add_alu(struct r600_bc *bc, const struct r600_bc_alu *alu) +int r600_bytecode_add_alu(struct r600_bytecode *bc, const struct r600_bytecode_alu *alu) { - return r600_bc_add_alu_type(bc, alu, BC_INST(bc, V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU)); + return r600_bytecode_add_alu_type(bc, alu, BC_INST(bc, V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU)); } -static unsigned r600_bc_num_tex_and_vtx_instructions(const struct r600_bc *bc) +static unsigned r600_bytecode_num_tex_and_vtx_instructions(const struct r600_bytecode *bc) { switch (bc->chip_class) { case R600: @@ -1281,7 +1293,7 @@ static unsigned r600_bc_num_tex_and_vtx_instructions(const struct r600_bc *bc) } } -static inline boolean last_inst_was_vtx_fetch(struct r600_bc *bc) +static inline boolean last_inst_was_vtx_fetch(struct r600_bytecode *bc) { if (bc->chip_class == CAYMAN) { if (bc->cf_last->inst != CM_V_SQ_CF_WORD1_SQ_CF_INST_TC) @@ -1294,20 +1306,20 @@ static inline boolean last_inst_was_vtx_fetch(struct r600_bc *bc) return FALSE; } -int r600_bc_add_vtx(struct r600_bc *bc, const struct r600_bc_vtx *vtx) +int r600_bytecode_add_vtx(struct r600_bytecode *bc, const struct r600_bytecode_vtx *vtx) { - struct r600_bc_vtx *nvtx = r600_bc_vtx(); + struct r600_bytecode_vtx *nvtx = r600_bytecode_vtx(); int r; if (nvtx == NULL) return -ENOMEM; - memcpy(nvtx, vtx, sizeof(struct r600_bc_vtx)); + memcpy(nvtx, vtx, sizeof(struct r600_bytecode_vtx)); /* cf can contains only alu or only vtx or only tex */ if (bc->cf_last == NULL || last_inst_was_vtx_fetch(bc) || bc->force_add_cf) { - r = r600_bc_add_cf(bc); + r = r600_bytecode_add_cf(bc); if (r) { free(nvtx); return r; @@ -1321,24 +1333,24 @@ int r600_bc_add_vtx(struct r600_bc *bc, const struct r600_bc_vtx *vtx) /* each fetch use 4 dwords */ bc->cf_last->ndw += 4; bc->ndw += 4; - if ((bc->cf_last->ndw / 4) >= r600_bc_num_tex_and_vtx_instructions(bc)) + if ((bc->cf_last->ndw / 4) >= r600_bytecode_num_tex_and_vtx_instructions(bc)) bc->force_add_cf = 1; return 0; } -int r600_bc_add_tex(struct r600_bc *bc, const struct r600_bc_tex *tex) +int r600_bytecode_add_tex(struct r600_bytecode *bc, const struct r600_bytecode_tex *tex) { - struct r600_bc_tex *ntex = r600_bc_tex(); + struct r600_bytecode_tex *ntex = r600_bytecode_tex(); int r; if (ntex == NULL) return -ENOMEM; - memcpy(ntex, tex, sizeof(struct r600_bc_tex)); + memcpy(ntex, tex, sizeof(struct r600_bytecode_tex)); /* we can't fetch data und use it as texture lookup address in the same TEX clause */ if (bc->cf_last != NULL && bc->cf_last->inst == V_SQ_CF_WORD1_SQ_CF_INST_TEX) { - struct r600_bc_tex *ttex; + struct r600_bytecode_tex *ttex; LIST_FOR_EACH_ENTRY(ttex, &bc->cf_last->tex, list) { if (ttex->dst_gpr == ntex->src_gpr) { bc->force_add_cf = 1; @@ -1354,7 +1366,7 @@ int r600_bc_add_tex(struct r600_bc *bc, const struct r600_bc_tex *tex) if (bc->cf_last == NULL || bc->cf_last->inst != V_SQ_CF_WORD1_SQ_CF_INST_TEX || bc->force_add_cf) { - r = r600_bc_add_cf(bc); + r = r600_bytecode_add_cf(bc); if (r) { free(ntex); return r; @@ -1371,15 +1383,15 @@ int r600_bc_add_tex(struct r600_bc *bc, const struct r600_bc_tex *tex) /* each texture fetch use 4 dwords */ bc->cf_last->ndw += 4; bc->ndw += 4; - if ((bc->cf_last->ndw / 4) >= r600_bc_num_tex_and_vtx_instructions(bc)) + if ((bc->cf_last->ndw / 4) >= r600_bytecode_num_tex_and_vtx_instructions(bc)) bc->force_add_cf = 1; return 0; } -int r600_bc_add_cfinst(struct r600_bc *bc, int inst) +int r600_bytecode_add_cfinst(struct r600_bytecode *bc, int inst) { int r; - r = r600_bc_add_cf(bc); + r = r600_bytecode_add_cf(bc); if (r) return r; @@ -1388,13 +1400,13 @@ int r600_bc_add_cfinst(struct r600_bc *bc, int inst) return 0; } -int cm_bc_add_cf_end(struct r600_bc *bc) +int cm_bytecode_add_cf_end(struct r600_bytecode *bc) { - return r600_bc_add_cfinst(bc, CM_V_SQ_CF_WORD1_SQ_CF_INST_END); + return r600_bytecode_add_cfinst(bc, CM_V_SQ_CF_WORD1_SQ_CF_INST_END); } /* common to all 3 families */ -static int r600_bc_vtx_build(struct r600_bc *bc, struct r600_bc_vtx *vtx, unsigned id) +static int r600_bytecode_vtx_build(struct r600_bytecode *bc, struct r600_bytecode_vtx *vtx, unsigned id) { bc->bytecode[id] = S_SQ_VTX_WORD0_BUFFER_ID(vtx->buffer_id) | S_SQ_VTX_WORD0_FETCH_TYPE(vtx->fetch_type) | @@ -1423,7 +1435,7 @@ static int r600_bc_vtx_build(struct r600_bc *bc, struct r600_bc_vtx *vtx, unsign } /* common to all 3 families */ -static int r600_bc_tex_build(struct r600_bc *bc, struct r600_bc_tex *tex, unsigned id) +static int r600_bytecode_tex_build(struct r600_bytecode *bc, struct r600_bytecode_tex *tex, unsigned id) { bc->bytecode[id++] = S_SQ_TEX_WORD0_TEX_INST(tex->inst) | S_SQ_TEX_WORD0_RESOURCE_ID(tex->resource_id) | @@ -1453,7 +1465,7 @@ static int r600_bc_tex_build(struct r600_bc *bc, struct r600_bc_tex *tex, unsign } /* r600 only, r700/eg bits in r700_asm.c */ -static int r600_bc_alu_build(struct r600_bc *bc, struct r600_bc_alu *alu, unsigned id) +static int r600_bytecode_alu_build(struct r600_bytecode *bc, struct r600_bytecode_alu *alu, unsigned id) { /* don't replace gpr by pv or ps for destination register */ bc->bytecode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) | @@ -1494,7 +1506,7 @@ static int r600_bc_alu_build(struct r600_bc *bc, struct r600_bc_alu *alu, unsign return 0; } -static void r600_bc_cf_vtx_build(uint32_t *bytecode, const struct r600_bc_cf *cf) +static void r600_bytecode_cf_vtx_build(uint32_t *bytecode, const struct r600_bytecode_cf *cf) { *bytecode++ = S_SQ_CF_WORD0_ADDR(cf->addr >> 1); *bytecode++ = S_SQ_CF_WORD1_CF_INST(cf->inst) | @@ -1503,7 +1515,7 @@ static void r600_bc_cf_vtx_build(uint32_t *bytecode, const struct r600_bc_cf *cf } /* common for r600/r700 - eg in eg_asm.c */ -static int r600_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) +static int r600_bytecode_cf_build(struct r600_bytecode *bc, struct r600_bytecode_cf *cf) { unsigned id = cf->id; @@ -1529,9 +1541,9 @@ static int r600_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) case V_SQ_CF_WORD1_SQ_CF_INST_VTX: case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC: if (bc->chip_class == R700) - r700_bc_cf_vtx_build(&bc->bytecode[id], cf); + r700_bytecode_cf_vtx_build(&bc->bytecode[id], cf); else - r600_bc_cf_vtx_build(&bc->bytecode[id], cf); + r600_bytecode_cf_vtx_build(&bc->bytecode[id], cf); break; case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT: case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE: @@ -1571,12 +1583,12 @@ static int r600_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) return 0; } -int r600_bc_build(struct r600_bc *bc) +int r600_bytecode_build(struct r600_bytecode *bc) { - struct r600_bc_cf *cf; - struct r600_bc_alu *alu; - struct r600_bc_vtx *vtx; - struct r600_bc_tex *tex; + struct r600_bytecode_cf *cf; + struct r600_bytecode_alu *alu; + struct r600_bytecode_vtx *vtx; + struct r600_bytecode_tex *tex; uint32_t literal[4]; unsigned nliteral; unsigned addr; @@ -1636,9 +1648,9 @@ int r600_bc_build(struct r600_bc *bc) LIST_FOR_EACH_ENTRY(cf, &bc->cf, list) { addr = cf->addr; if (bc->chip_class >= EVERGREEN) - r = eg_bc_cf_build(bc, cf); + r = eg_bytecode_cf_build(bc, cf); else - r = r600_bc_cf_build(bc, cf); + r = r600_bytecode_cf_build(bc, cf); if (r) return r; switch (cf->inst) { @@ -1649,18 +1661,18 @@ int r600_bc_build(struct r600_bc *bc) nliteral = 0; memset(literal, 0, sizeof(literal)); LIST_FOR_EACH_ENTRY(alu, &cf->alu, list) { - r = r600_bc_alu_nliterals(bc, alu, literal, &nliteral); + r = r600_bytecode_alu_nliterals(bc, alu, literal, &nliteral); if (r) return r; - r600_bc_alu_adjust_literals(bc, alu, literal, nliteral); + r600_bytecode_alu_adjust_literals(bc, alu, literal, nliteral); switch(bc->chip_class) { case R600: - r = r600_bc_alu_build(bc, alu, addr); + r = r600_bytecode_alu_build(bc, alu, addr); break; case R700: case EVERGREEN: /* eg alu is same encoding as r700 */ case CAYMAN: /* eg alu is same encoding as r700 */ - r = r700_bc_alu_build(bc, alu, addr); + r = r700_bytecode_alu_build(bc, alu, addr); break; default: R600_ERR("unknown chip class %d.\n", bc->chip_class); @@ -1681,7 +1693,7 @@ int r600_bc_build(struct r600_bc *bc) case V_SQ_CF_WORD1_SQ_CF_INST_VTX: case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC: LIST_FOR_EACH_ENTRY(vtx, &cf->vtx, list) { - r = r600_bc_vtx_build(bc, vtx, addr); + r = r600_bytecode_vtx_build(bc, vtx, addr); if (r) return r; addr += 4; @@ -1690,14 +1702,14 @@ int r600_bc_build(struct r600_bc *bc) case V_SQ_CF_WORD1_SQ_CF_INST_TEX: if (bc->chip_class == CAYMAN) { LIST_FOR_EACH_ENTRY(vtx, &cf->vtx, list) { - r = r600_bc_vtx_build(bc, vtx, addr); + r = r600_bytecode_vtx_build(bc, vtx, addr); if (r) return r; addr += 4; } } LIST_FOR_EACH_ENTRY(tex, &cf->tex, list) { - r = r600_bc_tex_build(bc, tex, addr); + r = r600_bytecode_tex_build(bc, tex, addr); if (r) return r; addr += 4; @@ -1726,17 +1738,17 @@ int r600_bc_build(struct r600_bc *bc) return 0; } -void r600_bc_clear(struct r600_bc *bc) +void r600_bytecode_clear(struct r600_bytecode *bc) { - struct r600_bc_cf *cf = NULL, *next_cf; + struct r600_bytecode_cf *cf = NULL, *next_cf; free(bc->bytecode); bc->bytecode = NULL; LIST_FOR_EACH_ENTRY_SAFE(cf, next_cf, &bc->cf, list) { - struct r600_bc_alu *alu = NULL, *next_alu; - struct r600_bc_tex *tex = NULL, *next_tex; - struct r600_bc_tex *vtx = NULL, *next_vtx; + struct r600_bytecode_alu *alu = NULL, *next_alu; + struct r600_bytecode_tex *tex = NULL, *next_tex; + struct r600_bytecode_tex *vtx = NULL, *next_vtx; LIST_FOR_EACH_ENTRY_SAFE(alu, next_alu, &cf->alu, list) { free(alu); @@ -1762,12 +1774,12 @@ void r600_bc_clear(struct r600_bc *bc) LIST_INITHEAD(&cf->list); } -void r600_bc_dump(struct r600_bc *bc) +void r600_bytecode_dump(struct r600_bytecode *bc) { - struct r600_bc_cf *cf = NULL; - struct r600_bc_alu *alu = NULL; - struct r600_bc_vtx *vtx = NULL; - struct r600_bc_tex *tex = NULL; + struct r600_bytecode_cf *cf = NULL; + struct r600_bytecode_alu *alu = NULL; + struct r600_bytecode_vtx *vtx = NULL; + struct r600_bytecode_tex *tex = NULL; unsigned i, id; uint32_t literal[4]; @@ -1866,7 +1878,7 @@ void r600_bc_dump(struct r600_bc *bc) id = cf->addr; nliteral = 0; LIST_FOR_EACH_ENTRY(alu, &cf->alu, list) { - r600_bc_alu_nliterals(bc, alu, literal, &nliteral); + r600_bytecode_alu_nliterals(bc, alu, literal, &nliteral); fprintf(stderr, "%04d %08X ", id, bc->bytecode[id]); fprintf(stderr, "SRC0(SEL:%d ", alu->src[0].sel); @@ -2120,8 +2132,8 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru { static int dump_shaders = -1; - struct r600_bc bc; - struct r600_bc_vtx vtx; + struct r600_bytecode bc; + struct r600_bytecode_vtx vtx; struct pipe_vertex_element *elements = ve->elements; const struct util_format_description *desc; unsigned fetch_resource_start = rctx->chip_class >= EVERGREEN ? 0 : 160; @@ -2142,11 +2154,11 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru } memset(&bc, 0, sizeof(bc)); - r600_bc_init(&bc, rctx->chip_class); + r600_bytecode_init(&bc, rctx->chip_class); for (i = 0; i < ve->count; i++) { if (elements[i].instance_divisor > 1) { - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; memset(&alu, 0, sizeof(alu)); alu.inst = BC_INST(&bc, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MULHI_UINT); @@ -2161,8 +2173,8 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru alu.dst.write = 1; alu.last = 1; - if ((r = r600_bc_add_alu(&bc, &alu))) { - r600_bc_clear(&bc); + if ((r = r600_bytecode_add_alu(&bc, &alu))) { + r600_bytecode_clear(&bc); return r; } } @@ -2173,7 +2185,7 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru r600_vertex_data_type(ve->elements[i].src_format, &format, &num_format, &format_comp, &endian); desc = util_format_description(ve->elements[i].src_format); if (desc == NULL) { - r600_bc_clear(&bc); + r600_bytecode_clear(&bc); R600_ERR("unknown format %d\n", ve->elements[i].src_format); return -EINVAL; } @@ -2198,16 +2210,16 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru vtx.offset = elements[i].src_offset; vtx.endian = endian; - if ((r = r600_bc_add_vtx(&bc, &vtx))) { - r600_bc_clear(&bc); + if ((r = r600_bytecode_add_vtx(&bc, &vtx))) { + r600_bytecode_clear(&bc); return r; } } - r600_bc_add_cfinst(&bc, BC_INST(&bc, V_SQ_CF_WORD1_SQ_CF_INST_RETURN)); + r600_bytecode_add_cfinst(&bc, BC_INST(&bc, V_SQ_CF_WORD1_SQ_CF_INST_RETURN)); - if ((r = r600_bc_build(&bc))) { - r600_bc_clear(&bc); + if ((r = r600_bytecode_build(&bc))) { + r600_bytecode_clear(&bc); return r; } @@ -2216,7 +2228,7 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru if (dump_shaders) { fprintf(stderr, "--------------------------------------------------------------\n"); - r600_bc_dump(&bc); + r600_bytecode_dump(&bc); fprintf(stderr, "______________________________________________________________\n"); } @@ -2225,14 +2237,14 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru /* use PIPE_BIND_VERTEX_BUFFER so we use the cache buffer manager */ ve->fetch_shader = r600_bo(rctx->radeon, ve->fs_size, 256, PIPE_BIND_VERTEX_BUFFER, PIPE_USAGE_IMMUTABLE); if (ve->fetch_shader == NULL) { - r600_bc_clear(&bc); + r600_bytecode_clear(&bc); return -ENOMEM; } - bytecode = r600_bo_map(rctx->radeon, ve->fetch_shader, 0, NULL); + bytecode = r600_bo_map(rctx->radeon, ve->fetch_shader, rctx->ctx.cs, PIPE_TRANSFER_WRITE); if (bytecode == NULL) { - r600_bc_clear(&bc); - r600_bo_reference(rctx->radeon, &ve->fetch_shader, NULL); + r600_bytecode_clear(&bc); + r600_bo_reference(&ve->fetch_shader, NULL); return -ENOMEM; } @@ -2245,7 +2257,7 @@ int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, stru } r600_bo_unmap(rctx->radeon, ve->fetch_shader); - r600_bc_clear(&bc); + r600_bytecode_clear(&bc); if (rctx->chip_class >= EVERGREEN) evergreen_fetch_shader(&rctx->context, ve); diff --git a/src/gallium/drivers/r600/r600_asm.h b/src/gallium/drivers/r600/r600_asm.h index cbdaacf7178..61caa4b915e 100644 --- a/src/gallium/drivers/r600/r600_asm.h +++ b/src/gallium/drivers/r600/r600_asm.h @@ -23,12 +23,10 @@ #ifndef R600_ASM_H #define R600_ASM_H -#include "util/u_double_list.h" - struct r600_vertex_element; struct r600_pipe_context; -struct r600_bc_alu_src { +struct r600_bytecode_alu_src { unsigned sel; unsigned chan; unsigned neg; @@ -37,7 +35,7 @@ struct r600_bc_alu_src { uint32_t value; }; -struct r600_bc_alu_dst { +struct r600_bytecode_alu_dst { unsigned sel; unsigned chan; unsigned clamp; @@ -45,10 +43,10 @@ struct r600_bc_alu_dst { unsigned rel; }; -struct r600_bc_alu { +struct r600_bytecode_alu { struct list_head list; - struct r600_bc_alu_src src[3]; - struct r600_bc_alu_dst dst; + struct r600_bytecode_alu_src src[3]; + struct r600_bytecode_alu_dst dst; unsigned inst; unsigned last; unsigned is_op3; @@ -58,7 +56,7 @@ struct r600_bc_alu { unsigned omod; }; -struct r600_bc_tex { +struct r600_bytecode_tex { struct list_head list; unsigned inst; unsigned resource_id; @@ -85,7 +83,7 @@ struct r600_bc_tex { unsigned src_sel_w; }; -struct r600_bc_vtx { +struct r600_bytecode_vtx { struct list_head list; unsigned inst; unsigned fetch_type; @@ -107,7 +105,7 @@ struct r600_bc_vtx { unsigned endian; }; -struct r600_bc_output { +struct r600_bytecode_output { unsigned array_base; unsigned type; unsigned end_of_program; @@ -122,13 +120,13 @@ struct r600_bc_output { unsigned barrier; }; -struct r600_bc_kcache { +struct r600_bytecode_kcache { unsigned bank; unsigned mode; unsigned addr; }; -struct r600_bc_cf { +struct r600_bytecode_cf { struct list_head list; unsigned inst; unsigned addr; @@ -137,15 +135,15 @@ struct r600_bc_cf { unsigned cond; unsigned pop_count; unsigned cf_addr; /* control flow addr */ - struct r600_bc_kcache kcache[2]; + struct r600_bytecode_kcache kcache[2]; unsigned r6xx_uses_waterfall; struct list_head alu; struct list_head tex; struct list_head vtx; - struct r600_bc_output output; - struct r600_bc_alu *curr_bs_head; - struct r600_bc_alu *prev_bs_head; - struct r600_bc_alu *prev2_bs_head; + struct r600_bytecode_output output; + struct r600_bytecode_alu *curr_bs_head; + struct r600_bytecode_alu *prev_bs_head; + struct r600_bytecode_alu *prev2_bs_head; }; #define FC_NONE 0 @@ -157,8 +155,8 @@ struct r600_bc_cf { struct r600_cf_stack_entry { int type; - struct r600_bc_cf *start; - struct r600_bc_cf **mid; /* used to store the else point */ + struct r600_bytecode_cf *start; + struct r600_bytecode_cf **mid; /* used to store the else point */ int num_mid; }; @@ -170,11 +168,11 @@ struct r600_cf_callstack { int max; }; -struct r600_bc { +struct r600_bytecode { enum chip_class chip_class; int type; struct list_head cf; - struct r600_bc_cf *cf_last; + struct r600_bytecode_cf *cf_last; unsigned ndw; unsigned ncf; unsigned ngpr; @@ -189,27 +187,27 @@ struct r600_bc { }; /* eg_asm.c */ -int eg_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf); +int eg_bytecode_cf_build(struct r600_bytecode *bc, struct r600_bytecode_cf *cf); /* r600_asm.c */ -void r600_bc_init(struct r600_bc *bc, enum chip_class chip_class); -void r600_bc_clear(struct r600_bc *bc); -int r600_bc_add_alu(struct r600_bc *bc, const struct r600_bc_alu *alu); -int r600_bc_add_vtx(struct r600_bc *bc, const struct r600_bc_vtx *vtx); -int r600_bc_add_tex(struct r600_bc *bc, const struct r600_bc_tex *tex); -int r600_bc_add_output(struct r600_bc *bc, const struct r600_bc_output *output); -int r600_bc_build(struct r600_bc *bc); -int r600_bc_add_cfinst(struct r600_bc *bc, int inst); -int r600_bc_add_alu_type(struct r600_bc *bc, const struct r600_bc_alu *alu, int type); -void r600_bc_special_constants(u32 value, unsigned *sel, unsigned *neg); -void r600_bc_dump(struct r600_bc *bc); - -int cm_bc_add_cf_end(struct r600_bc *bc); +void r600_bytecode_init(struct r600_bytecode *bc, enum chip_class chip_class); +void r600_bytecode_clear(struct r600_bytecode *bc); +int r600_bytecode_add_alu(struct r600_bytecode *bc, const struct r600_bytecode_alu *alu); +int r600_bytecode_add_vtx(struct r600_bytecode *bc, const struct r600_bytecode_vtx *vtx); +int r600_bytecode_add_tex(struct r600_bytecode *bc, const struct r600_bytecode_tex *tex); +int r600_bytecode_add_output(struct r600_bytecode *bc, const struct r600_bytecode_output *output); +int r600_bytecode_build(struct r600_bytecode *bc); +int r600_bytecode_add_cfinst(struct r600_bytecode *bc, int inst); +int r600_bytecode_add_alu_type(struct r600_bytecode *bc, const struct r600_bytecode_alu *alu, int type); +void r600_bytecode_special_constants(u32 value, unsigned *sel, unsigned *neg); +void r600_bytecode_dump(struct r600_bytecode *bc); + +int cm_bytecode_add_cf_end(struct r600_bytecode *bc); int r600_vertex_elements_build_fetch_shader(struct r600_pipe_context *rctx, struct r600_vertex_element *ve); /* r700_asm.c */ -void r700_bc_cf_vtx_build(uint32_t *bytecode, const struct r600_bc_cf *cf); -int r700_bc_alu_build(struct r600_bc *bc, struct r600_bc_alu *alu, unsigned id); +void r700_bytecode_cf_vtx_build(uint32_t *bytecode, const struct r600_bytecode_cf *cf); +int r700_bytecode_alu_build(struct r600_bytecode *bc, struct r600_bytecode_alu *alu, unsigned id); #endif diff --git a/src/gallium/drivers/r600/r600_blit.c b/src/gallium/drivers/r600/r600_blit.c index 35e68b6e222..2f7e871448a 100644 --- a/src/gallium/drivers/r600/r600_blit.c +++ b/src/gallium/drivers/r600/r600_blit.c @@ -111,7 +111,7 @@ void r600_blit_uncompress_depth(struct pipe_context *ctx, struct r600_resource_t if (!texture->dirty_db) return; - surf_tmpl.format = texture->resource.b.b.b.format; + surf_tmpl.format = texture->real_format; surf_tmpl.u.tex.level = level; surf_tmpl.u.tex.first_layer = 0; surf_tmpl.u.tex.last_layer = 0; @@ -119,7 +119,7 @@ void r600_blit_uncompress_depth(struct pipe_context *ctx, struct r600_resource_t zsurf = ctx->create_surface(ctx, &texture->resource.b.b.b, &surf_tmpl); - surf_tmpl.format = ((struct pipe_resource*)texture->flushed_depth_texture)->format; + surf_tmpl.format = texture->flushed_depth_texture->real_format; surf_tmpl.usage = PIPE_BIND_RENDER_TARGET; cbsurf = ctx->create_surface(ctx, (struct pipe_resource*)texture->flushed_depth_texture, &surf_tmpl); @@ -233,8 +233,8 @@ static void r600_hw_copy_region(struct pipe_context *ctx, struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; r600_blitter_begin(ctx, R600_COPY); - util_blitter_copy_region(rctx->blitter, dst, dst_level, dstx, dsty, dstz, - src, src_level, src_box, TRUE); + util_blitter_copy_texture(rctx->blitter, dst, dst_level, dstx, dsty, dstz, + src, src_level, src_box, TRUE); r600_blitter_end(ctx); } @@ -249,7 +249,7 @@ static void r600_compressed_to_blittable(struct pipe_resource *tex, struct texture_orig_info *orig) { struct r600_resource_texture *rtex = (struct r600_resource_texture*)tex; - unsigned pixsize = util_format_get_blocksize(tex->format); + unsigned pixsize = util_format_get_blocksize(rtex->real_format); int new_format; int new_height, new_width; @@ -269,7 +269,6 @@ static void r600_compressed_to_blittable(struct pipe_resource *tex, tex->width0 = new_width; tex->height0 = new_height; tex->format = new_format; - } static void r600_reset_blittable_to_compressed(struct pipe_resource *tex, diff --git a/src/gallium/drivers/r600/r600_buffer.c b/src/gallium/drivers/r600/r600_buffer.c index 72f352df3c3..ca2415adb28 100644 --- a/src/gallium/drivers/r600/r600_buffer.c +++ b/src/gallium/drivers/r600/r600_buffer.c @@ -33,11 +33,6 @@ #include <util/u_memory.h> #include "util/u_upload_mgr.h" -#include "state_tracker/drm_driver.h" - -#include <xf86drm.h> -#include "radeon_drm.h" - #include "r600.h" #include "r600_pipe.h" @@ -48,7 +43,7 @@ static void r600_buffer_destroy(struct pipe_screen *screen, struct r600_resource_buffer *rbuffer = r600_buffer(buf); if (rbuffer->r.bo) { - r600_bo_reference((struct radeon*)screen->winsys, &rbuffer->r.bo, NULL); + r600_bo_reference(&rbuffer->r.bo, NULL); } rbuffer->r.bo = NULL; util_slab_free(&rscreen->pool_buffers, rbuffer); @@ -81,12 +76,13 @@ static void *r600_buffer_transfer_map(struct pipe_context *pipe, struct pipe_transfer *transfer) { struct r600_resource_buffer *rbuffer = r600_buffer(transfer->resource); + struct r600_pipe_context *rctx = (struct r600_pipe_context*)pipe; uint8_t *data; if (rbuffer->r.b.user_ptr) return (uint8_t*)rbuffer->r.b.user_ptr + transfer->box.x; - data = r600_bo_map((struct radeon*)pipe->winsys, rbuffer->r.bo, transfer->usage, pipe); + data = r600_bo_map(rctx->screen->radeon, rbuffer->r.bo, rctx->ctx.cs, transfer->usage); if (!data) return NULL; @@ -97,12 +93,13 @@ static void r600_buffer_transfer_unmap(struct pipe_context *pipe, struct pipe_transfer *transfer) { struct r600_resource_buffer *rbuffer = r600_buffer(transfer->resource); + struct r600_pipe_context *rctx = (struct r600_pipe_context*)pipe; if (rbuffer->r.b.user_ptr) return; if (rbuffer->r.bo) - r600_bo_unmap((struct radeon*)pipe->winsys, rbuffer->r.bo); + r600_bo_unmap(rctx->screen->radeon, rbuffer->r.bo); } static void r600_buffer_transfer_flush_region(struct pipe_context *pipe, @@ -127,20 +124,20 @@ static void r600_buffer_transfer_inline_write(struct pipe_context *pipe, unsigned stride, unsigned layer_stride) { - struct radeon *ws = (struct radeon*)pipe->winsys; + struct r600_pipe_context *rctx = (struct r600_pipe_context*)pipe; + struct radeon *radeon = rctx->screen->radeon; struct r600_resource_buffer *rbuffer = r600_buffer(resource); uint8_t *map = NULL; assert(rbuffer->r.b.user_ptr == NULL); - map = r600_bo_map(ws, rbuffer->r.bo, - PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD | usage, - pipe); + map = r600_bo_map(radeon, rbuffer->r.bo, rctx->ctx.cs, + PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD | usage); memcpy(map + box->x, data, box->width); if (rbuffer->r.bo) - r600_bo_unmap(ws, rbuffer->r.bo); + r600_bo_unmap(radeon, rbuffer->r.bo); } static const struct u_resource_vtbl r600_buffer_vtbl = @@ -175,7 +172,7 @@ struct pipe_resource *r600_buffer_create(struct pipe_screen *screen, rbuffer->r.size = rbuffer->r.b.b.b.width0; rbuffer->r.bo_size = rbuffer->r.size; - bo = r600_bo((struct radeon*)screen->winsys, + bo = r600_bo(rscreen->radeon, rbuffer->r.b.b.b.width0, alignment, rbuffer->r.b.b.b.bind, rbuffer->r.b.b.b.usage); @@ -219,18 +216,18 @@ struct pipe_resource *r600_user_buffer_create(struct pipe_screen *screen, struct pipe_resource *r600_buffer_from_handle(struct pipe_screen *screen, struct winsys_handle *whandle) { - struct radeon *rw = (struct radeon*)screen->winsys; + struct radeon *rw = ((struct r600_screen*)screen)->radeon; struct r600_resource *rbuffer; struct r600_bo *bo = NULL; - bo = r600_bo_handle(rw, whandle->handle, NULL); + bo = r600_bo_handle(rw, whandle, NULL, NULL); if (bo == NULL) { return NULL; } rbuffer = CALLOC_STRUCT(r600_resource); if (rbuffer == NULL) { - r600_bo_reference(rw, &bo, NULL); + r600_bo_reference(&bo, NULL); return NULL; } diff --git a/src/gallium/drivers/r600/r600_formats.h b/src/gallium/drivers/r600/r600_formats.h index 1c1089d89d2..b822cba9293 100644 --- a/src/gallium/drivers/r600/r600_formats.h +++ b/src/gallium/drivers/r600/r600_formats.h @@ -99,7 +99,6 @@ static INLINE bool r600_is_vertex_format_supported(enum pipe_format format) /* No fixed, no double. */ if (desc->layout != UTIL_FORMAT_LAYOUT_PLAIN || - desc->channel[i].type == UTIL_FORMAT_TYPE_FIXED || (desc->channel[i].size == 64 && desc->channel[i].type == UTIL_FORMAT_TYPE_FLOAT)) return false; diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c index 615f0688eb3..ceaebbb4431 100644 --- a/src/gallium/drivers/r600/r600_pipe.c +++ b/src/gallium/drivers/r600/r600_pipe.c @@ -47,12 +47,14 @@ #include "r600_resource.h" #include "r600_shader.h" #include "r600_pipe.h" +#include "../../winsys/r600/drm/r600_drm_public.h" /* * pipe_context */ static struct r600_fence *r600_create_fence(struct r600_pipe_context *ctx) { + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_fence *fence = NULL; if (!ctx->fences.bo) { @@ -62,7 +64,8 @@ static struct r600_fence *r600_create_fence(struct r600_pipe_context *ctx) R600_ERR("r600: failed to create bo for fence objects\n"); return NULL; } - ctx->fences.data = r600_bo_map(ctx->radeon, ctx->fences.bo, PIPE_TRANSFER_UNSYNCHRONIZED, NULL); + ctx->fences.data = r600_bo_map(ctx->radeon, ctx->fences.bo, rctx->ctx.cs, + PIPE_TRANSFER_UNSYNCHRONIZED | PIPE_TRANSFER_WRITE); } if (!LIST_IS_EMPTY(&ctx->fences.pool)) { @@ -113,29 +116,28 @@ static struct r600_fence *r600_create_fence(struct r600_pipe_context *ctx) return fence; } -static void r600_flush(struct pipe_context *ctx, - struct pipe_fence_handle **fence) + +void r600_flush(struct pipe_context *ctx, struct pipe_fence_handle **fence, + unsigned flags) { struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_fence **rfence = (struct r600_fence**)fence; -#if 0 - static int dc = 0; - char dname[256]; -#endif - if (rfence) *rfence = r600_create_fence(rctx); -#if 0 - sprintf(dname, "gallium-%08d.bof", dc); - if (dc < 20) { - r600_context_dump_bof(&rctx->ctx, dname); - R600_ERR("dumped %s\n", dname); - } - dc++; -#endif - r600_context_flush(&rctx->ctx); + r600_context_flush(&rctx->ctx, flags); +} + +static void r600_flush_from_st(struct pipe_context *ctx, + struct pipe_fence_handle **fence) +{ + r600_flush(ctx, fence, 0); +} + +static void r600_flush_from_winsys(void *ctx, unsigned flags) +{ + r600_flush((struct pipe_context*)ctx, NULL, flags); } static void r600_update_num_contexts(struct r600_screen *rscreen, int diff) @@ -184,7 +186,7 @@ static void r600_destroy_context(struct pipe_context *context) } r600_bo_unmap(rctx->radeon, rctx->fences.bo); - r600_bo_reference(rctx->radeon, &rctx->fences.bo, NULL); + r600_bo_reference(&rctx->fences.bo, NULL); } r600_update_num_contexts(rctx->screen, -1); @@ -206,7 +208,7 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void rctx->context.screen = screen; rctx->context.priv = priv; rctx->context.destroy = r600_destroy_context; - rctx->context.flush = r600_flush; + rctx->context.flush = r600_flush_from_st; /* Easy accessing of screen/winsys. */ rctx->screen = rscreen; @@ -256,6 +258,8 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void return NULL; } + rctx->screen->ws->cs_set_flush_callback(rctx->ctx.cs, r600_flush_from_winsys, rctx); + util_slab_create(&rctx->pool_transfers, sizeof(struct pipe_transfer), 64, UTIL_SLAB_SINGLETHREADED); @@ -269,6 +273,7 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void r600_destroy_context(&rctx->context); return NULL; } + rctx->vbuf_mgr->caps.format_fixed32 = 0; rctx->blitter = util_blitter_create(&rctx->context); if (rctx->blitter == NULL) { @@ -355,6 +360,7 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param) case PIPE_CAP_SM3: case PIPE_CAP_SEAMLESS_CUBE_MAP: case PIPE_CAP_FRAGMENT_COLOR_CLAMP_CONTROL: + case PIPE_CAP_PRIMITIVE_RESTART: return 1; /* Supported except the original R600. */ @@ -369,7 +375,6 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param) /* Unsupported features. */ case PIPE_CAP_STREAM_OUTPUT: - case PIPE_CAP_PRIMITIVE_RESTART: case PIPE_CAP_TGSI_INSTANCEID: case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT: case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER: @@ -481,6 +486,8 @@ static int r600_get_shader_param(struct pipe_screen* pscreen, unsigned shader, e return 1; case PIPE_SHADER_CAP_SUBROUTINES: return 0; + case PIPE_SHADER_CAP_INTEGERS: + return 0; default: return 0; } @@ -498,6 +505,8 @@ static int r600_get_video_param(struct pipe_screen *screen, case PIPE_VIDEO_CAP_MAX_WIDTH: case PIPE_VIDEO_CAP_MAX_HEIGHT: return vl_video_buffer_max_size(screen); + case PIPE_VIDEO_CAP_NUM_BUFFERS_DESIRED: + return vl_num_buffers_desired(screen, profile); default: return 0; } @@ -510,7 +519,8 @@ static void r600_destroy_screen(struct pipe_screen* pscreen) if (rscreen == NULL) return; - radeon_decref(rscreen->radeon); + radeon_destroy(rscreen->radeon); + rscreen->ws->destroy(rscreen->ws); util_slab_destroy(&rscreen->pool_buffers); pipe_mutex_destroy(rscreen->mutex_num_contexts); @@ -574,17 +584,19 @@ static boolean r600_fence_finish(struct pipe_screen *pscreen, return TRUE; } -struct pipe_screen *r600_screen_create(struct radeon *radeon) +struct pipe_screen *r600_screen_create(struct radeon_winsys *ws) { struct r600_screen *rscreen; + struct radeon *radeon = radeon_create(ws); rscreen = CALLOC_STRUCT(r600_screen); if (rscreen == NULL) { return NULL; } + rscreen->ws = ws; rscreen->radeon = radeon; - rscreen->screen.winsys = (struct pipe_winsys*)radeon; + rscreen->screen.winsys = (struct pipe_winsys*)ws; rscreen->screen.destroy = r600_destroy_screen; rscreen->screen.get_name = r600_get_name; rscreen->screen.get_vendor = r600_get_vendor; diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h index 6f399ed43b0..2747f54079c 100644 --- a/src/gallium/drivers/r600/r600_pipe.h +++ b/src/gallium/drivers/r600/r600_pipe.h @@ -26,6 +26,8 @@ #ifndef R600_PIPE_H #define R600_PIPE_H +#include "../../winsys/radeon/drm/radeon_winsys.h" + #include <pipe/p_state.h> #include <pipe/p_screen.h> #include <pipe/p_context.h> @@ -72,6 +74,7 @@ enum r600_pipe_state_id { struct r600_screen { struct pipe_screen screen; + struct radeon_winsys *ws; struct radeon *radeon; struct r600_tiling_info *tiling_info; struct util_slab_mempool pool_buffers; @@ -183,7 +186,7 @@ struct r600_pipe_context { struct r600_pipe_state *states[R600_PIPE_NSTATES]; struct r600_context ctx; struct r600_vertex_element *vertex_elements; - struct r600_pipe_resource_state fs_resource[PIPE_MAX_ATTRIBS]; + struct r600_pipe_resource_state fs_resource[PIPE_MAX_ATTRIBS]; struct pipe_framebuffer_state framebuffer; struct pipe_index_buffer index_buffer; unsigned cb_target_mask; @@ -247,7 +250,8 @@ void evergreen_pipe_init_buffer_resource(struct r600_pipe_context *rctx, struct r600_pipe_resource_state *rstate); void evergreen_pipe_mod_buffer_resource(struct r600_pipe_resource_state *rstate, struct r600_resource *rbuffer, - unsigned offset, unsigned stride); + unsigned offset, unsigned stride, + enum radeon_bo_usage usage); boolean evergreen_is_format_supported(struct pipe_screen *screen, enum pipe_format format, enum pipe_texture_target target, @@ -270,6 +274,11 @@ struct pipe_resource *r600_buffer_from_handle(struct pipe_screen *screen, struct winsys_handle *whandle); void r600_upload_index_buffer(struct r600_pipe_context *rctx, struct r600_drawl *draw); + +/* r600_pipe.c */ +void r600_flush(struct pipe_context *ctx, struct pipe_fence_handle **fence, + unsigned flags); + /* r600_query.c */ void r600_init_query_functions(struct r600_pipe_context *rctx); @@ -294,7 +303,8 @@ void r600_pipe_init_buffer_resource(struct r600_pipe_context *rctx, struct r600_pipe_resource_state *rstate); void r600_pipe_mod_buffer_resource(struct r600_pipe_resource_state *rstate, struct r600_resource *rbuffer, - unsigned offset, unsigned stride); + unsigned offset, unsigned stride, + enum radeon_bo_usage usage); void r600_adjust_gprs(struct r600_pipe_context *rctx); boolean r600_is_format_supported(struct pipe_screen *screen, enum pipe_format format, diff --git a/src/gallium/drivers/r600/r600_public.h b/src/gallium/drivers/r600/r600_public.h index f1970201e89..e4fe23a87b7 100644 --- a/src/gallium/drivers/r600/r600_public.h +++ b/src/gallium/drivers/r600/r600_public.h @@ -23,6 +23,8 @@ #ifndef R600_PUBLIC_H #define R600_PUBLIC_H -struct pipe_screen *r600_screen_create(struct radeon *radeon); +struct radeon_winsys; + +struct pipe_screen *r600_screen_create(struct radeon_winsys *ws); #endif diff --git a/src/gallium/drivers/r600/r600_resource.h b/src/gallium/drivers/r600/r600_resource.h index 836e7491f1f..d9d29db7968 100644 --- a/src/gallium/drivers/r600/r600_resource.h +++ b/src/gallium/drivers/r600/r600_resource.h @@ -52,6 +52,12 @@ struct r600_resource { struct r600_resource_texture { struct r600_resource resource; + + /* If this resource is a depth-stencil buffer on evergreen, this contains + * the depth part of the format. There is a separate stencil resource + * for the stencil buffer below. */ + enum pipe_format real_format; + unsigned offset[PIPE_MAX_TEXTURE_LEVELS]; unsigned pitch_in_bytes[PIPE_MAX_TEXTURE_LEVELS]; /* transfer */ unsigned pitch_in_blocks[PIPE_MAX_TEXTURE_LEVELS]; /* texture resource */ @@ -62,6 +68,7 @@ struct r600_resource_texture { unsigned tile_type; unsigned depth; unsigned dirty_db; + struct r600_resource_texture *stencil; /* Stencil is in a separate buffer on Evergreen. */ struct r600_resource_texture *flushed_depth_texture; boolean is_flushing_texture; diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c index 3e21ad1fdc6..c37bb729ce3 100644 --- a/src/gallium/drivers/r600/r600_shader.c +++ b/src/gallium/drivers/r600/r600_shader.c @@ -86,7 +86,7 @@ static int r600_pipe_shader(struct pipe_context *ctx, struct r600_pipe_shader *s if (shader->bo == NULL) { return -ENOMEM; } - ptr = (uint32_t*)r600_bo_map(rctx->radeon, shader->bo, 0, NULL); + ptr = (uint32_t*)r600_bo_map(rctx->radeon, shader->bo, rctx->ctx.cs, PIPE_TRANSFER_WRITE); if (R600_BIG_ENDIAN) { for (i = 0; i < rshader->bc.ndw; ++i) { ptr[i] = bswap_32(rshader->bc.bytecode[i]); @@ -140,13 +140,13 @@ int r600_pipe_shader_create(struct pipe_context *ctx, struct r600_pipe_shader *s R600_ERR("translation from TGSI failed !\n"); return r; } - r = r600_bc_build(&shader->shader.bc); + r = r600_bytecode_build(&shader->shader.bc); if (r) { R600_ERR("building bytecode failed !\n"); return r; } if (dump_shaders) { - r600_bc_dump(&shader->shader.bc); + r600_bytecode_dump(&shader->shader.bc); fprintf(stderr, "______________________________________________________________\n"); } return r600_pipe_shader(ctx, shader); @@ -154,10 +154,8 @@ int r600_pipe_shader_create(struct pipe_context *ctx, struct r600_pipe_shader *s void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *shader) { - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - r600_bo_reference(rctx->radeon, &shader->bo, NULL); - r600_bc_clear(&shader->shader.bc); + r600_bo_reference(&shader->bo, NULL); + r600_bytecode_clear(&shader->shader.bc); memset(&shader->shader,0,sizeof(struct r600_shader)); } @@ -185,7 +183,7 @@ struct r600_shader_ctx { unsigned temp_reg; unsigned ar_reg; struct r600_shader_tgsi_instruction *inst_info; - struct r600_bc *bc; + struct r600_bytecode *bc; struct r600_shader *shader; struct r600_shader_src src[4]; u32 *literals; @@ -246,7 +244,7 @@ static int tgsi_is_supported(struct r600_shader_ctx *ctx) static int evergreen_interp_alu(struct r600_shader_ctx *ctx, int input) { int i, r; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int gpr = 0, base_chan = 0; int ij_index = 0; @@ -272,7 +270,7 @@ static int evergreen_interp_alu(struct r600_shader_ctx *ctx, int input) base_chan = (2 * (ij_index % 2)) + 1; for (i = 0; i < 8; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); if (i < 4) alu.inst = EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INTERP_ZW; @@ -294,7 +292,7 @@ static int evergreen_interp_alu(struct r600_shader_ctx *ctx, int input) alu.bank_swizzle_force = SQ_ALU_VEC_210; if ((i % 4) == 3) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -332,6 +330,12 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx) ctx->shader->output[i].sid = d->Semantic.Index; ctx->shader->output[i].gpr = ctx->file_offset[TGSI_FILE_OUTPUT] + i; ctx->shader->output[i].interpolate = d->Declaration.Interpolate; + if (ctx->type == TGSI_PROCESSOR_VERTEX) { + /* these don't count as vertex param exports */ + if ((ctx->shader->output[i].name == TGSI_SEMANTIC_POSITION) || + (ctx->shader->output[i].name == TGSI_SEMANTIC_PSIZE)) + ctx->shader->npos++; + } break; case TGSI_FILE_CONSTANT: case TGSI_FILE_TEMPORARY: @@ -341,8 +345,8 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx) case TGSI_FILE_SYSTEM_VALUE: if (d->Semantic.Name == TGSI_SEMANTIC_INSTANCEID) { - struct r600_bc_alu alu; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + struct r600_bytecode_alu alu; + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_INT_TO_FLT); alu.src[0].sel = 0; @@ -353,7 +357,7 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - if ((r = r600_bc_add_alu(ctx->bc, &alu))) + if ((r = r600_bytecode_add_alu(ctx->bc, &alu))) return r; break; } @@ -436,7 +440,7 @@ static void tgsi_src(struct r600_shader_ctx *ctx, (tgsi_src->Register.SwizzleX == tgsi_src->Register.SwizzleW)) { index = tgsi_src->Register.Index * 4 + tgsi_src->Register.SwizzleX; - r600_bc_special_constants(ctx->literals[index], &r600_src->sel, &r600_src->neg); + r600_bytecode_special_constants(ctx->literals[index], &r600_src->sel, &r600_src->neg); if (r600_src->sel != V_SQ_ALU_SRC_LITERAL) return; } @@ -460,12 +464,12 @@ static void tgsi_src(struct r600_shader_ctx *ctx, static int tgsi_fetch_rel_const(struct r600_shader_ctx *ctx, unsigned int offset, unsigned int dst_reg) { - struct r600_bc_vtx vtx; + struct r600_bytecode_vtx vtx; unsigned int ar_reg; int r; if (offset) { - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; memset(&alu, 0, sizeof(alu)); @@ -479,7 +483,7 @@ static int tgsi_fetch_rel_const(struct r600_shader_ctx *ctx, unsigned int offset alu.dst.write = 1; alu.last = 1; - if ((r = r600_bc_add_alu(ctx->bc, &alu))) + if ((r = r600_bytecode_add_alu(ctx->bc, &alu))) return r; ar_reg = dst_reg; @@ -502,7 +506,7 @@ static int tgsi_fetch_rel_const(struct r600_shader_ctx *ctx, unsigned int offset vtx.srf_mode_all = 1; /* SRF_MODE_NO_ZERO */ vtx.endian = r600_endian_swap(32); - if ((r = r600_bc_add_vtx(ctx->bc, &vtx))) + if ((r = r600_bytecode_add_vtx(ctx->bc, &vtx))) return r; return 0; @@ -511,7 +515,7 @@ static int tgsi_fetch_rel_const(struct r600_shader_ctx *ctx, unsigned int offset static int tgsi_split_constant(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, j, k, nconst, r; for (i = 0, nconst = 0; i < inst->Instruction.NumSrcRegs; i++) { @@ -536,7 +540,7 @@ static int tgsi_split_constant(struct r600_shader_ctx *ctx) } else if (j > 0) { int treg = r600_get_temp(ctx); for (k = 0; k < 4; k++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = ctx->src[i].sel; alu.src[0].chan = k; @@ -546,7 +550,7 @@ static int tgsi_split_constant(struct r600_shader_ctx *ctx) alu.dst.write = 1; if (k == 3) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -562,7 +566,7 @@ static int tgsi_split_constant(struct r600_shader_ctx *ctx) static int tgsi_split_literal_constant(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, j, k, nliteral, r; for (i = 0, nliteral = 0; i < inst->Instruction.NumSrcRegs; i++) { @@ -574,7 +578,7 @@ static int tgsi_split_literal_constant(struct r600_shader_ctx *ctx) if (j > 0 && ctx->src[i].sel == V_SQ_ALU_SRC_LITERAL) { int treg = r600_get_temp(ctx); for (k = 0; k < 4; k++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = ctx->src[i].sel; alu.src[0].chan = k; @@ -584,7 +588,7 @@ static int tgsi_split_literal_constant(struct r600_shader_ctx *ctx) alu.dst.write = 1; if (k == 3) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -602,14 +606,14 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi struct tgsi_full_immediate *immediate; struct tgsi_full_property *property; struct r600_shader_ctx ctx; - struct r600_bc_output output[32]; + struct r600_bytecode_output output[32]; unsigned output_done, noutput; unsigned opcode; int i, j, r = 0, pos0; ctx.bc = &shader->bc; ctx.shader = shader; - r600_bc_init(ctx.bc, rctx->chip_class); + r600_bytecode_init(ctx.bc, rctx->chip_class); ctx.tokens = tokens; tgsi_scan_shader(tokens, &ctx.info); tgsi_parse_init(&ctx.parse, tokens); @@ -649,18 +653,18 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi if (ctx.type == TGSI_PROCESSOR_VERTEX) { ctx.file_offset[TGSI_FILE_INPUT] = 1; if (ctx.bc->chip_class >= EVERGREEN) { - r600_bc_add_cfinst(ctx.bc, EG_V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS); + r600_bytecode_add_cfinst(ctx.bc, EG_V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS); } else { - r600_bc_add_cfinst(ctx.bc, V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS); + r600_bytecode_add_cfinst(ctx.bc, V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS); } } if (ctx.type == TGSI_PROCESSOR_FRAGMENT && ctx.bc->chip_class >= EVERGREEN) { ctx.file_offset[TGSI_FILE_INPUT] = evergreen_gpr_count(&ctx); } ctx.file_offset[TGSI_FILE_OUTPUT] = ctx.file_offset[TGSI_FILE_INPUT] + - ctx.info.file_count[TGSI_FILE_INPUT]; + ctx.info.file_max[TGSI_FILE_INPUT] + 1; ctx.file_offset[TGSI_FILE_TEMPORARY] = ctx.file_offset[TGSI_FILE_OUTPUT] + - ctx.info.file_count[TGSI_FILE_OUTPUT]; + ctx.info.file_max[TGSI_FILE_OUTPUT] + 1; /* Outside the GPR range. This will be translated to one of the * kcache banks later. */ @@ -668,7 +672,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi ctx.file_offset[TGSI_FILE_IMMEDIATE] = V_SQ_ALU_SRC_LITERAL; ctx.ar_reg = ctx.file_offset[TGSI_FILE_TEMPORARY] + - ctx.info.file_count[TGSI_FILE_TEMPORARY]; + ctx.info.file_max[TGSI_FILE_TEMPORARY] + 1; ctx.temp_reg = ctx.ar_reg + 1; ctx.nliterals = 0; @@ -742,8 +746,8 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi int j; for (j = 0; j < 4; j++) { - struct r600_bc_alu alu; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + struct r600_bytecode_alu alu; + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); /* MOV_SAT R, R */ alu.inst = BC_INST(ctx.bc, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); @@ -757,7 +761,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi if (j == 3) { alu.last = 1; } - r = r600_bc_add_alu(ctx.bc, &alu); + r = r600_bytecode_add_alu(ctx.bc, &alu); if (r) return r; } @@ -768,7 +772,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi /* export output */ j = 0; for (i = 0, pos0 = 0; i < noutput; i++) { - memset(&output[i], 0, sizeof(struct r600_bc_output)); + memset(&output[i], 0, sizeof(struct r600_bytecode_output)); output[i + j].gpr = shader->output[i].gpr; output[i + j].elem_size = 3; output[i + j].swizzle_x = 0; @@ -801,7 +805,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi output[i + j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL; if (shader->fs_write_all && (rctx->chip_class >= EVERGREEN)) { for (j = 1; j < shader->nr_cbufs; j++) { - memset(&output[i + j], 0, sizeof(struct r600_bc_output)); + memset(&output[i + j], 0, sizeof(struct r600_bytecode_output)); output[i + j].gpr = shader->output[i].gpr; output[i + j].elem_size = 3; output[i + j].swizzle_x = 0; @@ -850,7 +854,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi } } if (!pos0) { - memset(&output[i], 0, sizeof(struct r600_bc_output)); + memset(&output[i], 0, sizeof(struct r600_bytecode_output)); output[i].gpr = 0; output[i].elem_size = 3; output[i].swizzle_x = 0; @@ -867,7 +871,7 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi } /* add fake pixel export */ if (ctx.type == TGSI_PROCESSOR_FRAGMENT && !noutput) { - memset(&output[0], 0, sizeof(struct r600_bc_output)); + memset(&output[0], 0, sizeof(struct r600_bytecode_output)); output[0].gpr = 0; output[0].elem_size = 3; output[0].swizzle_x = 7; @@ -895,13 +899,13 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi } /* add output to bytecode */ for (i = 0; i < noutput; i++) { - r = r600_bc_add_output(ctx.bc, &output[i]); + r = r600_bytecode_add_output(ctx.bc, &output[i]); if (r) goto out_err; } /* add program end */ if (ctx.bc->chip_class == CAYMAN) - cm_bc_add_cf_end(ctx.bc); + cm_bytecode_add_cf_end(ctx.bc); free(ctx.literals); tgsi_parse_free(&ctx.parse); @@ -924,7 +928,7 @@ static int tgsi_end(struct r600_shader_ctx *ctx) return 0; } -static void r600_bc_src(struct r600_bc_alu_src *bc_src, +static void r600_bytecode_src(struct r600_bytecode_alu_src *bc_src, const struct r600_shader_src *shader_src, unsigned chan) { @@ -936,13 +940,13 @@ static void r600_bc_src(struct r600_bc_alu_src *bc_src, bc_src->value = shader_src->value[bc_src->chan]; } -static void r600_bc_src_set_abs(struct r600_bc_alu_src *bc_src) +static void r600_bytecode_src_set_abs(struct r600_bytecode_alu_src *bc_src) { bc_src->abs = 1; bc_src->neg = 0; } -static void r600_bc_src_toggle_neg(struct r600_bc_alu_src *bc_src) +static void r600_bytecode_src_toggle_neg(struct r600_bytecode_alu_src *bc_src) { bc_src->neg = !bc_src->neg; } @@ -950,7 +954,7 @@ static void r600_bc_src_toggle_neg(struct r600_bc_alu_src *bc_src) static void tgsi_dst(struct r600_shader_ctx *ctx, const struct tgsi_full_dst_register *tgsi_dst, unsigned swizzle, - struct r600_bc_alu_dst *r600_dst) + struct r600_bytecode_alu_dst *r600_dst) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; @@ -980,7 +984,7 @@ static int tgsi_last_instruction(unsigned writemask) static int tgsi_op2_s(struct r600_shader_ctx *ctx, int swap) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, j, r; int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask); @@ -988,25 +992,25 @@ static int tgsi_op2_s(struct r600_shader_ctx *ctx, int swap) if (!(inst->Dst[0].Register.WriteMask & (1 << i))) continue; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); alu.inst = ctx->inst_info->r600_opcode; if (!swap) { for (j = 0; j < inst->Instruction.NumSrcRegs; j++) { - r600_bc_src(&alu.src[j], &ctx->src[j], i); + r600_bytecode_src(&alu.src[j], &ctx->src[j], i); } } else { - r600_bc_src(&alu.src[0], &ctx->src[1], i); - r600_bc_src(&alu.src[1], &ctx->src[0], i); + r600_bytecode_src(&alu.src[0], &ctx->src[1], i); + r600_bytecode_src(&alu.src[1], &ctx->src[0], i); } /* handle some special cases */ switch (ctx->inst_info->tgsi_opcode) { case TGSI_OPCODE_SUB: - r600_bc_src_toggle_neg(&alu.src[1]); + r600_bytecode_src_toggle_neg(&alu.src[1]); break; case TGSI_OPCODE_ABS: - r600_bc_src_set_abs(&alu.src[0]); + r600_bytecode_src_set_abs(&alu.src[0]); break; default: break; @@ -1014,7 +1018,7 @@ static int tgsi_op2_s(struct r600_shader_ctx *ctx, int swap) if (i == lasti) { alu.last = 1; } - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1035,21 +1039,21 @@ static int cayman_emit_float_instr(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; int i, j, r; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int last_slot = (inst->Dst[0].Register.WriteMask & 0x8) ? 4 : 3; for (i = 0 ; i < last_slot; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = ctx->inst_info->r600_opcode; for (j = 0; j < inst->Instruction.NumSrcRegs; j++) { - r600_bc_src(&alu.src[j], &ctx->src[j], 0); + r600_bytecode_src(&alu.src[j], &ctx->src[j], 0); } tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); alu.dst.write = (inst->Dst[0].Register.WriteMask >> i) & 1; if (i == last_slot - 1) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1068,9 +1072,9 @@ static int tgsi_setup_trig(struct r600_shader_ctx *ctx) static float neg_pi = -3.1415926535; int r; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD); alu.is_op3 = 1; @@ -1078,7 +1082,7 @@ static int tgsi_setup_trig(struct r600_shader_ctx *ctx) alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.src[1].sel = V_SQ_ALU_SRC_LITERAL; alu.src[1].chan = 0; @@ -1086,11 +1090,11 @@ static int tgsi_setup_trig(struct r600_shader_ctx *ctx) alu.src[2].sel = V_SQ_ALU_SRC_0_5; alu.src[2].chan = 0; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT); alu.dst.chan = 0; @@ -1100,11 +1104,11 @@ static int tgsi_setup_trig(struct r600_shader_ctx *ctx) alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 0; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD); alu.is_op3 = 1; @@ -1130,7 +1134,7 @@ static int tgsi_setup_trig(struct r600_shader_ctx *ctx) } alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; return 0; @@ -1139,7 +1143,7 @@ static int tgsi_setup_trig(struct r600_shader_ctx *ctx) static int cayman_trig(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int last_slot = (inst->Dst[0].Register.WriteMask & 0x8) ? 4 : 3; int i, r; @@ -1149,7 +1153,7 @@ static int cayman_trig(struct r600_shader_ctx *ctx) for (i = 0; i < last_slot; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = ctx->inst_info->r600_opcode; alu.dst.chan = i; @@ -1160,7 +1164,7 @@ static int cayman_trig(struct r600_shader_ctx *ctx) alu.src[0].chan = 0; if (i == last_slot - 1) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1170,7 +1174,7 @@ static int cayman_trig(struct r600_shader_ctx *ctx) static int tgsi_trig(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r; int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask); @@ -1178,7 +1182,7 @@ static int tgsi_trig(struct r600_shader_ctx *ctx) if (r) return r; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = ctx->inst_info->r600_opcode; alu.dst.chan = 0; alu.dst.sel = ctx->temp_reg; @@ -1187,7 +1191,7 @@ static int tgsi_trig(struct r600_shader_ctx *ctx) alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 0; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; @@ -1196,14 +1200,14 @@ static int tgsi_trig(struct r600_shader_ctx *ctx) if (!(inst->Dst[0].Register.WriteMask & (1 << i))) continue; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = ctx->temp_reg; tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); if (i == lasti) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1213,7 +1217,7 @@ static int tgsi_trig(struct r600_shader_ctx *ctx) static int tgsi_scs(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r; /* We'll only need the trig stuff if we are going to write to the @@ -1229,7 +1233,7 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) { if (ctx->bc->chip_class == CAYMAN) { for (i = 0 ; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS); tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); @@ -1241,19 +1245,19 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) alu.src[0].chan = 0; if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS); tgsi_dst(ctx, &inst->Dst[0], 0, &alu.dst); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 0; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1263,7 +1267,7 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) { if (ctx->bc->chip_class == CAYMAN) { for (i = 0 ; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN); tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); if (i == 1) @@ -1274,19 +1278,19 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) alu.src[0].chan = 0; if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN); tgsi_dst(ctx, &inst->Dst[0], 1, &alu.dst); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 0; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1294,7 +1298,7 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) /* dst.z = 0.0; */ if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); @@ -1305,14 +1309,14 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } /* dst.w = 1.0; */ if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); @@ -1323,7 +1327,7 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1333,11 +1337,11 @@ static int tgsi_scs(struct r600_shader_ctx *ctx) static int tgsi_kill(struct r600_shader_ctx *ctx) { - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r; for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = ctx->inst_info->r600_opcode; alu.dst.chan = i; @@ -1348,12 +1352,12 @@ static int tgsi_kill(struct r600_shader_ctx *ctx) alu.src[1].sel = V_SQ_ALU_SRC_1; alu.src[1].neg = 1; } else { - r600_bc_src(&alu.src[1], &ctx->src[0], i); + r600_bytecode_src(&alu.src[1], &ctx->src[0], i); } if (i == 3) { alu.last = 1; } - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1367,13 +1371,13 @@ static int tgsi_kill(struct r600_shader_ctx *ctx) static int tgsi_lit(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int r; /* tmp.x = max(src.y, 0.0) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX); - r600_bc_src(&alu.src[0], &ctx->src[0], 1); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 1); alu.src[1].sel = V_SQ_ALU_SRC_0; /*0.0*/ alu.src[1].chan = 1; @@ -1382,7 +1386,7 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; @@ -1395,7 +1399,7 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { /* tmp.z = log(tmp.x) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 0; @@ -1407,13 +1411,13 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) } else alu.dst.write = 0; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { /* tmp.z = log(tmp.x) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 0; @@ -1421,7 +1425,7 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) alu.dst.chan = 2; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1430,25 +1434,25 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) sel = alu.dst.sel; /* tmp.x = amd MUL_LIT(tmp.z, src.w, src.x ) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT); alu.src[0].sel = sel; alu.src[0].chan = chan; - r600_bc_src(&alu.src[1], &ctx->src[0], 3); - r600_bc_src(&alu.src[2], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[1], &ctx->src[0], 3); + r600_bytecode_src(&alu.src[2], &ctx->src[0], 0); alu.dst.sel = ctx->temp_reg; alu.dst.chan = 0; alu.dst.write = 1; alu.is_op3 = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { /* dst.z = exp(tmp.x) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 0; @@ -1458,56 +1462,56 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) alu.last = 1; } else alu.dst.write = 0; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { /* dst.z = exp(tmp.x) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 0; tgsi_dst(ctx, &inst->Dst[0], 2, &alu.dst); alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } /* dst.x, <- 1.0 */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = V_SQ_ALU_SRC_1; /*1.0*/ alu.src[0].chan = 0; tgsi_dst(ctx, &inst->Dst[0], 0, &alu.dst); alu.dst.write = (inst->Dst[0].Register.WriteMask >> 0) & 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; /* dst.y = max(src.x, 0.0) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.src[1].sel = V_SQ_ALU_SRC_0; /*0.0*/ alu.src[1].chan = 0; tgsi_dst(ctx, &inst->Dst[0], 1, &alu.dst); alu.dst.write = (inst->Dst[0].Register.WriteMask >> 1) & 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; /* dst.w, <- 1.0 */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = V_SQ_ALU_SRC_1; alu.src[0].chan = 0; tgsi_dst(ctx, &inst->Dst[0], 3, &alu.dst); alu.dst.write = (inst->Dst[0].Register.WriteMask >> 3) & 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; @@ -1517,10 +1521,10 @@ static int tgsi_lit(struct r600_shader_ctx *ctx) static int tgsi_rsq(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); /* FIXME: * For state trackers other than OpenGL, we'll want to use @@ -1529,13 +1533,13 @@ static int tgsi_rsq(struct r600_shader_ctx *ctx) alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_CLAMPED); for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { - r600_bc_src(&alu.src[i], &ctx->src[i], 0); - r600_bc_src_set_abs(&alu.src[i]); + r600_bytecode_src(&alu.src[i], &ctx->src[i], 0); + r600_bytecode_src_set_abs(&alu.src[i]); } alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; /* replicate result */ @@ -1545,11 +1549,11 @@ static int tgsi_rsq(struct r600_shader_ctx *ctx) static int tgsi_helper_tempx_replicate(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r; for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.src[0].sel = ctx->temp_reg; alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.dst.chan = i; @@ -1557,7 +1561,7 @@ static int tgsi_helper_tempx_replicate(struct r600_shader_ctx *ctx) alu.dst.write = (inst->Dst[0].Register.WriteMask >> i) & 1; if (i == 3) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1567,18 +1571,18 @@ static int tgsi_helper_tempx_replicate(struct r600_shader_ctx *ctx) static int tgsi_trans_srcx_replicate(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = ctx->inst_info->r600_opcode; for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { - r600_bc_src(&alu.src[i], &ctx->src[i], 0); + r600_bytecode_src(&alu.src[i], &ctx->src[i], 0); } alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; /* replicate result */ @@ -1589,38 +1593,38 @@ static int cayman_pow(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; int i, r; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int last_slot = (inst->Dst[0].Register.WriteMask & 0x8) ? 4 : 3; for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; alu.dst.write = 1; if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } /* b * LOG2(a) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL); - r600_bc_src(&alu.src[0], &ctx->src[1], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[1], 0); alu.src[1].sel = ctx->temp_reg; alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; for (i = 0; i < last_slot; i++) { /* POW(a,b) = EXP2(b * LOG2(a))*/ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); alu.src[0].sel = ctx->temp_reg; @@ -1628,7 +1632,7 @@ static int cayman_pow(struct r600_shader_ctx *ctx) alu.dst.write = (inst->Dst[0].Register.WriteMask >> i) & 1; if (i == last_slot - 1) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1637,38 +1641,38 @@ static int cayman_pow(struct r600_shader_ctx *ctx) static int tgsi_pow(struct r600_shader_ctx *ctx) { - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int r; /* LOG2(a) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; /* b * LOG2(a) */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL); - r600_bc_src(&alu.src[0], &ctx->src[1], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[1], 0); alu.src[1].sel = ctx->temp_reg; alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; /* POW(a,b) = EXP2(b * LOG2(a))*/ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); alu.src[0].sel = ctx->temp_reg; alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; return tgsi_helper_tempx_replicate(ctx); @@ -1677,32 +1681,32 @@ static int tgsi_pow(struct r600_shader_ctx *ctx) static int tgsi_ssg(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r; /* tmp = (src > 0 ? 1 : src) */ for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT); alu.is_op3 = 1; alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; - r600_bc_src(&alu.src[0], &ctx->src[0], i); + r600_bytecode_src(&alu.src[0], &ctx->src[0], i); alu.src[1].sel = V_SQ_ALU_SRC_1; - r600_bc_src(&alu.src[2], &ctx->src[0], i); + r600_bytecode_src(&alu.src[2], &ctx->src[0], i); if (i == 3) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } /* dst = (-tmp > 0 ? -1 : tmp) */ for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT); alu.is_op3 = 1; tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); @@ -1719,7 +1723,7 @@ static int tgsi_ssg(struct r600_shader_ctx *ctx) if (i == 3) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1728,11 +1732,11 @@ static int tgsi_ssg(struct r600_shader_ctx *ctx) static int tgsi_helper_copy(struct r600_shader_ctx *ctx, struct tgsi_full_instruction *inst) { - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r; for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); if (!(inst->Dst[0].Register.WriteMask & (1 << i))) { alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP); alu.dst.chan = i; @@ -1745,7 +1749,7 @@ static int tgsi_helper_copy(struct r600_shader_ctx *ctx, struct tgsi_full_instru if (i == 3) { alu.last = 1; } - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1755,7 +1759,7 @@ static int tgsi_helper_copy(struct r600_shader_ctx *ctx, struct tgsi_full_instru static int tgsi_op3(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, j, r; int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask); @@ -1763,10 +1767,10 @@ static int tgsi_op3(struct r600_shader_ctx *ctx) if (!(inst->Dst[0].Register.WriteMask & (1 << i))) continue; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = ctx->inst_info->r600_opcode; for (j = 0; j < inst->Instruction.NumSrcRegs; j++) { - r600_bc_src(&alu.src[j], &ctx->src[j], i); + r600_bytecode_src(&alu.src[j], &ctx->src[j], i); } tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); @@ -1776,7 +1780,7 @@ static int tgsi_op3(struct r600_shader_ctx *ctx) if (i == lasti) { alu.last = 1; } - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1786,14 +1790,14 @@ static int tgsi_op3(struct r600_shader_ctx *ctx) static int tgsi_dp(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, j, r; for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = ctx->inst_info->r600_opcode; for (j = 0; j < inst->Instruction.NumSrcRegs; j++) { - r600_bc_src(&alu.src[j], &ctx->src[j], i); + r600_bytecode_src(&alu.src[j], &ctx->src[j], i); } tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); @@ -1826,7 +1830,7 @@ static int tgsi_dp(struct r600_shader_ctx *ctx) if (i == 3) { alu.last = 1; } - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1853,8 +1857,8 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) { static float one_point_five = 1.5f; struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_tex tex; - struct r600_bc_alu alu; + struct r600_bytecode_tex tex; + struct r600_bytecode_alu alu; unsigned src_gpr; int r, i, j; int opcode; @@ -1872,7 +1876,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) for (i = 1; i < 3; i++) { /* set gradients h/v */ - memset(&tex, 0, sizeof(struct r600_bc_tex)); + memset(&tex, 0, sizeof(struct r600_bytecode_tex)); tex.inst = (i == 1) ? SQ_TEX_INST_SET_GRADIENTS_H : SQ_TEX_INST_SET_GRADIENTS_V; tex.sampler_id = tgsi_tex_get_src_gpr(ctx, sampler_src_reg); @@ -1886,15 +1890,15 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) tex.src_sel_w = 3; for (j = 0; j < 4; j++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); - r600_bc_src(&alu.src[0], &ctx->src[i], j); + r600_bytecode_src(&alu.src[0], &ctx->src[i], j); alu.dst.sel = tex.src_gpr; alu.dst.chan = j; if (j == 3) alu.last = 1; alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -1915,7 +1919,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) tex.coord_type_z = 1; tex.coord_type_w = 1; } - r = r600_bc_add_tex(ctx->bc, &tex); + r = r600_bytecode_add_tex(ctx->bc, &tex); if (r) return r; } @@ -1925,9 +1929,9 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) if (ctx->bc->chip_class == CAYMAN) { out_chan = 2; for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 3); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 3); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; @@ -1935,40 +1939,40 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) alu.last = 1; if (out_chan == i) alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { out_chan = 3; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 3); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 3); alu.dst.sel = ctx->temp_reg; alu.dst.chan = out_chan; alu.last = 1; alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = out_chan; - r600_bc_src(&alu.src[1], &ctx->src[0], i); + r600_bytecode_src(&alu.src[1], &ctx->src[0], i); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = V_SQ_ALU_SRC_1; alu.src[0].chan = 0; @@ -1976,7 +1980,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) alu.dst.chan = 3; alu.last = 1; alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; src_loaded = TRUE; @@ -1989,16 +1993,16 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) /* tmp1.xyzw = CUBE(R0.zzxy, R0.yxzz) */ for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE); - r600_bc_src(&alu.src[0], &ctx->src[0], src0_swizzle[i]); - r600_bc_src(&alu.src[1], &ctx->src[0], src1_swizzle[i]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], src0_swizzle[i]); + r600_bytecode_src(&alu.src[1], &ctx->src[0], src1_swizzle[i]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; if (i == 3) alu.last = 1; alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2006,7 +2010,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) /* tmp1.z = RCP_e(|tmp1.z|) */ if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 2; @@ -2017,12 +2021,12 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) alu.dst.write = 1; if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 2; @@ -2031,7 +2035,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) alu.dst.chan = 2; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2040,7 +2044,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) * MULADD R0.y, R0.y, PS1, (0x3FC00000, 1.5f).x * muladd has no writemask, have to use another temp */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD); alu.is_op3 = 1; @@ -2057,11 +2061,11 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) alu.dst.chan = 0; alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD); alu.is_op3 = 1; @@ -2079,7 +2083,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; @@ -2089,15 +2093,15 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) if (src_requires_loading && !src_loaded) { for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); - r600_bc_src(&alu.src[0], &ctx->src[0], i); + r600_bytecode_src(&alu.src[0], &ctx->src[0], i); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; if (i == 3) alu.last = 1; alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2120,7 +2124,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) } } - memset(&tex, 0, sizeof(struct r600_bc_tex)); + memset(&tex, 0, sizeof(struct r600_bytecode_tex)); tex.inst = opcode; tex.sampler_id = tgsi_tex_get_src_gpr(ctx, sampler_src_reg); @@ -2167,7 +2171,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) if (inst->Texture.Texture == TGSI_TEXTURE_SHADOW1D || inst->Texture.Texture == TGSI_TEXTURE_SHADOW2D) tex.src_sel_w = tex.src_sel_z; - r = r600_bc_add_tex(ctx->bc, &tex); + r = r600_bytecode_add_tex(ctx->bc, &tex); if (r) return r; @@ -2178,7 +2182,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) static int tgsi_lrp(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask); unsigned i; int r; @@ -2189,17 +2193,17 @@ static int tgsi_lrp(struct r600_shader_ctx *ctx) if (!(inst->Dst[0].Register.WriteMask & (1 << i))) continue; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD); - r600_bc_src(&alu.src[0], &ctx->src[1], i); - r600_bc_src(&alu.src[1], &ctx->src[2], i); + r600_bytecode_src(&alu.src[0], &ctx->src[1], i); + r600_bytecode_src(&alu.src[1], &ctx->src[2], i); alu.omod = 3; tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); alu.dst.chan = i; if (i == lasti) { alu.last = 1; } - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2211,19 +2215,19 @@ static int tgsi_lrp(struct r600_shader_ctx *ctx) if (!(inst->Dst[0].Register.WriteMask & (1 << i))) continue; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD); alu.src[0].sel = V_SQ_ALU_SRC_1; alu.src[0].chan = 0; - r600_bc_src(&alu.src[1], &ctx->src[0], i); - r600_bc_src_toggle_neg(&alu.src[1]); + r600_bytecode_src(&alu.src[1], &ctx->src[0], i); + r600_bytecode_src_toggle_neg(&alu.src[1]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; if (i == lasti) { alu.last = 1; } alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2233,18 +2237,18 @@ static int tgsi_lrp(struct r600_shader_ctx *ctx) if (!(inst->Dst[0].Register.WriteMask & (1 << i))) continue; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = i; - r600_bc_src(&alu.src[1], &ctx->src[2], i); + r600_bytecode_src(&alu.src[1], &ctx->src[2], i); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; if (i == lasti) { alu.last = 1; } alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2254,11 +2258,11 @@ static int tgsi_lrp(struct r600_shader_ctx *ctx) if (!(inst->Dst[0].Register.WriteMask & (1 << i))) continue; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD); alu.is_op3 = 1; - r600_bc_src(&alu.src[0], &ctx->src[0], i); - r600_bc_src(&alu.src[1], &ctx->src[1], i); + r600_bytecode_src(&alu.src[0], &ctx->src[0], i); + r600_bytecode_src(&alu.src[1], &ctx->src[1], i); alu.src[2].sel = ctx->temp_reg; alu.src[2].chan = i; @@ -2267,7 +2271,7 @@ static int tgsi_lrp(struct r600_shader_ctx *ctx) if (i == lasti) { alu.last = 1; } - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2277,7 +2281,7 @@ static int tgsi_lrp(struct r600_shader_ctx *ctx) static int tgsi_cmp(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r; int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask); @@ -2285,18 +2289,18 @@ static int tgsi_cmp(struct r600_shader_ctx *ctx) if (!(inst->Dst[0].Register.WriteMask & (1 << i))) continue; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGE); - r600_bc_src(&alu.src[0], &ctx->src[0], i); - r600_bc_src(&alu.src[1], &ctx->src[2], i); - r600_bc_src(&alu.src[2], &ctx->src[1], i); + r600_bytecode_src(&alu.src[0], &ctx->src[0], i); + r600_bytecode_src(&alu.src[1], &ctx->src[2], i); + r600_bytecode_src(&alu.src[2], &ctx->src[1], i); tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); alu.dst.chan = i; alu.dst.write = 1; alu.is_op3 = 1; if (i == lasti) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2308,7 +2312,7 @@ static int tgsi_xpd(struct r600_shader_ctx *ctx) struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; static const unsigned int src0_swizzle[] = {2, 0, 1}; static const unsigned int src1_swizzle[] = {1, 2, 0}; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; uint32_t use_temp = 0; int i, r; @@ -2316,11 +2320,11 @@ static int tgsi_xpd(struct r600_shader_ctx *ctx) use_temp = 1; for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL); if (i < 3) { - r600_bc_src(&alu.src[0], &ctx->src[0], src0_swizzle[i]); - r600_bc_src(&alu.src[1], &ctx->src[1], src1_swizzle[i]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], src0_swizzle[i]); + r600_bytecode_src(&alu.src[1], &ctx->src[1], src1_swizzle[i]); } else { alu.src[0].sel = V_SQ_ALU_SRC_0; alu.src[0].chan = i; @@ -2334,18 +2338,18 @@ static int tgsi_xpd(struct r600_shader_ctx *ctx) if (i == 3) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD); if (i < 3) { - r600_bc_src(&alu.src[0], &ctx->src[0], src1_swizzle[i]); - r600_bc_src(&alu.src[1], &ctx->src[1], src0_swizzle[i]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], src1_swizzle[i]); + r600_bytecode_src(&alu.src[1], &ctx->src[1], src0_swizzle[i]); } else { alu.src[0].sel = V_SQ_ALU_SRC_0; alu.src[0].chan = i; @@ -2366,7 +2370,7 @@ static int tgsi_xpd(struct r600_shader_ctx *ctx) alu.is_op3 = 1; if (i == 3) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2378,22 +2382,22 @@ static int tgsi_xpd(struct r600_shader_ctx *ctx) static int tgsi_exp(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int r; int i; /* result.x = 2^floor(src); */ if (inst->Dst[0].Register.WriteMask & 1) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.dst.sel = ctx->temp_reg; alu.dst.chan = 0; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; @@ -2409,7 +2413,7 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) alu.dst.write = 1; if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2422,7 +2426,7 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) alu.dst.chan = 0; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2430,10 +2434,10 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) /* result.y = tmp - floor(tmp); */ if ((inst->Dst[0].Register.WriteMask >> 1) & 1) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.dst.sel = ctx->temp_reg; #if 0 @@ -2446,7 +2450,7 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2455,9 +2459,9 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) if ((inst->Dst[0].Register.WriteMask >> 2) & 0x1) { if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; @@ -2466,14 +2470,14 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) alu.last = 1; } - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; @@ -2481,7 +2485,7 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2489,7 +2493,7 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) /* result.w = 1.0;*/ if ((inst->Dst[0].Register.WriteMask >> 3) & 0x1) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = V_SQ_ALU_SRC_1; @@ -2499,7 +2503,7 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) alu.dst.chan = 3; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2509,7 +2513,7 @@ static int tgsi_exp(struct r600_shader_ctx *ctx) static int tgsi_log(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int r; int i; @@ -2517,11 +2521,11 @@ static int tgsi_log(struct r600_shader_ctx *ctx) if (inst->Dst[0].Register.WriteMask & 1) { if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); - r600_bc_src_set_abs(&alu.src[0]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; @@ -2529,23 +2533,23 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.dst.write = 1; if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); - r600_bc_src_set_abs(&alu.src[0]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = 0; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2559,7 +2563,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2569,11 +2573,11 @@ static int tgsi_log(struct r600_shader_ctx *ctx) if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); - r600_bc_src_set_abs(&alu.src[0]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; @@ -2582,28 +2586,28 @@ static int tgsi_log(struct r600_shader_ctx *ctx) if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); - r600_bc_src_set_abs(&alu.src[0]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.chan = 1; alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR); alu.src[0].sel = ctx->temp_reg; @@ -2614,13 +2618,13 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 1; @@ -2632,12 +2636,12 @@ static int tgsi_log(struct r600_shader_ctx *ctx) if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 1; @@ -2647,14 +2651,14 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 1; @@ -2666,12 +2670,12 @@ static int tgsi_log(struct r600_shader_ctx *ctx) if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE); alu.src[0].sel = ctx->temp_reg; alu.src[0].chan = 1; @@ -2681,17 +2685,17 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); - r600_bc_src_set_abs(&alu.src[0]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src_set_abs(&alu.src[0]); alu.src[1].sel = ctx->temp_reg; alu.src[1].chan = 1; @@ -2701,7 +2705,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2710,11 +2714,11 @@ static int tgsi_log(struct r600_shader_ctx *ctx) if ((inst->Dst[0].Register.WriteMask >> 2) & 1) { if (ctx->bc->chip_class == CAYMAN) { for (i = 0; i < 3; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); - r600_bc_src_set_abs(&alu.src[0]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; if (i == 2) @@ -2723,23 +2727,23 @@ static int tgsi_log(struct r600_shader_ctx *ctx) if (i == 2) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } } else { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE); - r600_bc_src(&alu.src[0], &ctx->src[0], 0); - r600_bc_src_set_abs(&alu.src[0]); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src_set_abs(&alu.src[0]); alu.dst.sel = ctx->temp_reg; alu.dst.write = 1; alu.dst.chan = 2; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2747,7 +2751,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) /* result.w = 1.0; */ if ((inst->Dst[0].Register.WriteMask >> 3) & 1) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = V_SQ_ALU_SRC_1; @@ -2758,7 +2762,7 @@ static int tgsi_log(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2769,10 +2773,10 @@ static int tgsi_log(struct r600_shader_ctx *ctx) static int tgsi_eg_arl(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int r; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: @@ -2786,11 +2790,11 @@ static int tgsi_eg_arl(struct r600_shader_ctx *ctx) return -1; } - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.last = 1; alu.dst.sel = ctx->ar_reg; alu.dst.write = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; @@ -2800,12 +2804,12 @@ static int tgsi_eg_arl(struct r600_shader_ctx *ctx) * between ARL and AR usage. The easy way to do that is to remove * the MOVA here, and load it for the first AR access after ar_reg * has been modified in each clause. */ - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT; alu.src[0].sel = ctx->ar_reg; alu.src[0].chan = 0; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; return 0; @@ -2814,19 +2818,19 @@ static int tgsi_r600_arl(struct r600_shader_ctx *ctx) { /* TODO from r600c, ar values don't persist between clauses */ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int r; switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: memset(&alu, 0, sizeof(alu)); alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR; - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.dst.sel = ctx->ar_reg; alu.dst.write = 1; alu.last = 1; - if ((r = r600_bc_add_alu(ctx->bc, &alu))) + if ((r = r600_bytecode_add_alu(ctx->bc, &alu))) return r; memset(&alu, 0, sizeof(alu)); @@ -2836,18 +2840,18 @@ static int tgsi_r600_arl(struct r600_shader_ctx *ctx) alu.dst.write = 1; alu.last = 1; - if ((r = r600_bc_add_alu(ctx->bc, &alu))) + if ((r = r600_bytecode_add_alu(ctx->bc, &alu))) return r; break; case TGSI_OPCODE_ARR: memset(&alu, 0, sizeof(alu)); alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT; - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.dst.sel = ctx->ar_reg; alu.dst.write = 1; alu.last = 1; - if ((r = r600_bc_add_alu(ctx->bc, &alu))) + if ((r = r600_bytecode_add_alu(ctx->bc, &alu))) return r; break; default: @@ -2860,7 +2864,7 @@ static int tgsi_r600_arl(struct r600_shader_ctx *ctx) alu.src[0].sel = ctx->ar_reg; alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; ctx->bc->cf_last->r6xx_uses_waterfall = 1; @@ -2870,11 +2874,11 @@ static int tgsi_r600_arl(struct r600_shader_ctx *ctx) static int tgsi_opdst(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int i, r = 0; for (i = 0; i < 4; i++) { - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL); tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst); @@ -2882,17 +2886,17 @@ static int tgsi_opdst(struct r600_shader_ctx *ctx) if (i == 0 || i == 3) { alu.src[0].sel = V_SQ_ALU_SRC_1; } else { - r600_bc_src(&alu.src[0], &ctx->src[0], i); + r600_bytecode_src(&alu.src[0], &ctx->src[0], i); } if (i == 0 || i == 2) { alu.src[1].sel = V_SQ_ALU_SRC_1; } else { - r600_bc_src(&alu.src[1], &ctx->src[1], i); + r600_bytecode_src(&alu.src[1], &ctx->src[1], i); } if (i == 3) alu.last = 1; - r = r600_bc_add_alu(ctx->bc, &alu); + r = r600_bytecode_add_alu(ctx->bc, &alu); if (r) return r; } @@ -2901,10 +2905,10 @@ static int tgsi_opdst(struct r600_shader_ctx *ctx) static int emit_logic_pred(struct r600_shader_ctx *ctx, int opcode) { - struct r600_bc_alu alu; + struct r600_bytecode_alu alu; int r; - memset(&alu, 0, sizeof(struct r600_bc_alu)); + memset(&alu, 0, sizeof(struct r600_bytecode_alu)); alu.inst = opcode; alu.predicate = 1; @@ -2912,13 +2916,13 @@ static int emit_logic_pred(struct r600_shader_ctx *ctx, int opcode) alu.dst.write = 1; alu.dst.chan = 0; - r600_bc_src(&alu.src[0], &ctx->src[0], 0); + r600_bytecode_src(&alu.src[0], &ctx->src[0], 0); alu.src[1].sel = V_SQ_ALU_SRC_0; alu.src[1].chan = 0; alu.last = 1; - r = r600_bc_add_alu_type(ctx->bc, &alu, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE)); + r = r600_bytecode_add_alu_type(ctx->bc, &alu, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE)); if (r) return r; return 0; @@ -2926,25 +2930,34 @@ static int emit_logic_pred(struct r600_shader_ctx *ctx, int opcode) static int pops(struct r600_shader_ctx *ctx, int pops) { - int alu_pop = 3; - if (ctx->bc->cf_last) { - if (ctx->bc->cf_last->inst == CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU) << 3) - alu_pop = 0; - else if (ctx->bc->cf_last->inst == CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER) << 3) - alu_pop = 1; - } - alu_pop += pops; - if (alu_pop == 1) { - ctx->bc->cf_last->inst = CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER) << 3; - ctx->bc->force_add_cf = 1; - } else if (alu_pop == 2) { - ctx->bc->cf_last->inst = CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP2_AFTER) << 3; - ctx->bc->force_add_cf = 1; - } else { - r600_bc_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_POP)); + unsigned force_pop = ctx->bc->force_add_cf; + + if (!force_pop) { + int alu_pop = 3; + if (ctx->bc->cf_last) { + if (ctx->bc->cf_last->inst == CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU) << 3) + alu_pop = 0; + else if (ctx->bc->cf_last->inst == CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER) << 3) + alu_pop = 1; + } + alu_pop += pops; + if (alu_pop == 1) { + ctx->bc->cf_last->inst = CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER) << 3; + ctx->bc->force_add_cf = 1; + } else if (alu_pop == 2) { + ctx->bc->cf_last->inst = CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP2_AFTER) << 3; + ctx->bc->force_add_cf = 1; + } else { + force_pop = 1; + } + } + + if (force_pop) { + r600_bytecode_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_POP)); ctx->bc->cf_last->pop_count = pops; ctx->bc->cf_last->cf_addr = ctx->bc->cf_last->id + 2; } + return 0; } @@ -3011,8 +3024,8 @@ static void fc_set_mid(struct r600_shader_ctx *ctx, int fc_sp) { struct r600_cf_stack_entry *sp = &ctx->bc->fc_stack[fc_sp]; - sp->mid = (struct r600_bc_cf **)realloc((void *)sp->mid, - sizeof(struct r600_bc_cf *) * (sp->num_mid + 1)); + sp->mid = (struct r600_bytecode_cf **)realloc((void *)sp->mid, + sizeof(struct r600_bytecode_cf *) * (sp->num_mid + 1)); sp->mid[sp->num_mid] = ctx->bc->cf_last; sp->num_mid++; } @@ -3040,14 +3053,14 @@ static void fc_poplevel(struct r600_shader_ctx *ctx) #if 0 static int emit_return(struct r600_shader_ctx *ctx) { - r600_bc_add_cfinst(ctx->bc, V_SQ_CF_WORD1_SQ_CF_INST_RETURN); + r600_bytecode_add_cfinst(ctx->bc, V_SQ_CF_WORD1_SQ_CF_INST_RETURN); return 0; } static int emit_jump_to_offset(struct r600_shader_ctx *ctx, int pops, int offset) { - r600_bc_add_cfinst(ctx->bc, V_SQ_CF_WORD1_SQ_CF_INST_JUMP); + r600_bytecode_add_cfinst(ctx->bc, V_SQ_CF_WORD1_SQ_CF_INST_JUMP); ctx->bc->cf_last->pop_count = pops; /* TODO work out offset */ return 0; @@ -3076,7 +3089,7 @@ static void break_loop_on_flag(struct r600_shader_ctx *ctx, unsigned fc_sp) { emit_testflag(ctx); - r600_bc_add_cfinst(ctx->bc, ctx->inst_info->r600_opcode); + r600_bytecode_add_cfinst(ctx->bc, ctx->inst_info->r600_opcode); ctx->bc->cf_last->pop_count = 1; fc_set_mid(ctx, fc_sp); @@ -3089,7 +3102,7 @@ static int tgsi_if(struct r600_shader_ctx *ctx) { emit_logic_pred(ctx, CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE)); - r600_bc_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_JUMP)); + r600_bytecode_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_JUMP)); fc_pushlevel(ctx, FC_IF); @@ -3099,7 +3112,7 @@ static int tgsi_if(struct r600_shader_ctx *ctx) static int tgsi_else(struct r600_shader_ctx *ctx) { - r600_bc_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_ELSE)); + r600_bytecode_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_ELSE)); ctx->bc->cf_last->pop_count = 1; fc_set_mid(ctx, ctx->bc->fc_sp); @@ -3129,7 +3142,7 @@ static int tgsi_endif(struct r600_shader_ctx *ctx) static int tgsi_bgnloop(struct r600_shader_ctx *ctx) { - r600_bc_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL)); + r600_bytecode_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL)); fc_pushlevel(ctx, FC_LOOP); @@ -3142,7 +3155,7 @@ static int tgsi_endloop(struct r600_shader_ctx *ctx) { int i; - r600_bc_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END)); + r600_bytecode_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END)); if (ctx->bc->fc_stack[ctx->bc->fc_sp].type != FC_LOOP) { R600_ERR("loop/endloop in shader code are not paired.\n"); @@ -3182,7 +3195,7 @@ static int tgsi_loop_brk_cont(struct r600_shader_ctx *ctx) return -EINVAL; } - r600_bc_add_cfinst(ctx->bc, ctx->inst_info->r600_opcode); + r600_bytecode_add_cfinst(ctx->bc, ctx->inst_info->r600_opcode); ctx->bc->cf_last->pop_count = 1; fc_set_mid(ctx, fscp); @@ -3228,7 +3241,7 @@ static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[] = { {TGSI_OPCODE_FRC, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT, tgsi_op2}, {TGSI_OPCODE_CLAMP, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_FLR, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR, tgsi_op2}, - {TGSI_OPCODE_ROUND, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, + {TGSI_OPCODE_ROUND, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RNDNE, tgsi_op2}, {TGSI_OPCODE_EX2, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE, tgsi_trans_srcx_replicate}, {TGSI_OPCODE_LG2, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE, tgsi_trans_srcx_replicate}, {TGSI_OPCODE_POW, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_pow}, @@ -3353,6 +3366,18 @@ static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[] = { {TGSI_OPCODE_CASE, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_DEFAULT, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_ENDSWITCH, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, + {TGSI_OPCODE_LOAD, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_LOAD_MS, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_B, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_C, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_C_LZ, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_D, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_L, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_GATHER4, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_RESINFO, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_POS, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_INFO, 0, 0, tgsi_unsupported}, {TGSI_OPCODE_LAST, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, }; @@ -3386,7 +3411,7 @@ static struct r600_shader_tgsi_instruction eg_shader_tgsi_instruction[] = { {TGSI_OPCODE_FRC, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT, tgsi_op2}, {TGSI_OPCODE_CLAMP, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_FLR, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR, tgsi_op2}, - {TGSI_OPCODE_ROUND, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, + {TGSI_OPCODE_ROUND, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RNDNE, tgsi_op2}, {TGSI_OPCODE_EX2, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE, tgsi_trans_srcx_replicate}, {TGSI_OPCODE_LG2, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE, tgsi_trans_srcx_replicate}, {TGSI_OPCODE_POW, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_pow}, @@ -3511,6 +3536,18 @@ static struct r600_shader_tgsi_instruction eg_shader_tgsi_instruction[] = { {TGSI_OPCODE_CASE, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_DEFAULT, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_ENDSWITCH, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, + {TGSI_OPCODE_LOAD, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_LOAD_MS, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_B, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_C, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_C_LZ, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_D, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_L, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_GATHER4, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_RESINFO, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_POS, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_INFO, 0, 0, tgsi_unsupported}, {TGSI_OPCODE_LAST, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, }; @@ -3544,7 +3581,7 @@ static struct r600_shader_tgsi_instruction cm_shader_tgsi_instruction[] = { {TGSI_OPCODE_FRC, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT, tgsi_op2}, {TGSI_OPCODE_CLAMP, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_FLR, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR, tgsi_op2}, - {TGSI_OPCODE_ROUND, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, + {TGSI_OPCODE_ROUND, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RNDNE, tgsi_op2}, {TGSI_OPCODE_EX2, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE, cayman_emit_float_instr}, {TGSI_OPCODE_LG2, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE, cayman_emit_float_instr}, {TGSI_OPCODE_POW, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, cayman_pow}, @@ -3669,5 +3706,17 @@ static struct r600_shader_tgsi_instruction cm_shader_tgsi_instruction[] = { {TGSI_OPCODE_CASE, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_DEFAULT, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_ENDSWITCH, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, + {TGSI_OPCODE_LOAD, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_LOAD_MS, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_B, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_C, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_C_LZ, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_D, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_L, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_GATHER4, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_RESINFO, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_POS, 0, 0, tgsi_unsupported}, + {TGSI_OPCODE_SAMPLE_INFO, 0, 0, tgsi_unsupported}, {TGSI_OPCODE_LAST, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, }; diff --git a/src/gallium/drivers/r600/r600_shader.h b/src/gallium/drivers/r600/r600_shader.h index 3ba84bd8907..ada369ade68 100644 --- a/src/gallium/drivers/r600/r600_shader.h +++ b/src/gallium/drivers/r600/r600_shader.h @@ -37,9 +37,10 @@ struct r600_shader_io { struct r600_shader { unsigned processor_type; - struct r600_bc bc; + struct r600_bytecode bc; unsigned ninput; unsigned noutput; + unsigned npos; unsigned nlds; struct r600_shader_io input[32]; struct r600_shader_io output[32]; diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c index 01406f2bad6..fba2af8a6ac 100644 --- a/src/gallium/drivers/r600/r600_state.c +++ b/src/gallium/drivers/r600/r600_state.c @@ -662,19 +662,19 @@ void r600_polygon_offset_update(struct r600_pipe_context *rctx) offset_db_fmt_cntl |= S_028DF8_POLY_OFFSET_NEG_NUM_DB_BITS(depth); r600_pipe_state_add_reg(&state, R_028E00_PA_SU_POLY_OFFSET_FRONT_SCALE, - fui(rctx->rasterizer->offset_scale), 0xFFFFFFFF, NULL); + fui(rctx->rasterizer->offset_scale), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&state, R_028E04_PA_SU_POLY_OFFSET_FRONT_OFFSET, - fui(offset_units), 0xFFFFFFFF, NULL); + fui(offset_units), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&state, R_028E08_PA_SU_POLY_OFFSET_BACK_SCALE, - fui(rctx->rasterizer->offset_scale), 0xFFFFFFFF, NULL); + fui(rctx->rasterizer->offset_scale), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&state, R_028E0C_PA_SU_POLY_OFFSET_BACK_OFFSET, - fui(offset_units), 0xFFFFFFFF, NULL); + fui(offset_units), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&state, R_028DF8_PA_SU_POLY_OFFSET_DB_FMT_CNTL, - offset_db_fmt_cntl, 0xFFFFFFFF, NULL); + offset_db_fmt_cntl, 0xFFFFFFFF, NULL, 0); r600_context_pipe_state_set(&rctx->ctx, &state); } } @@ -689,10 +689,10 @@ static void r600_set_blend_color(struct pipe_context *ctx, return; rstate->id = R600_PIPE_STATE_BLEND_COLOR; - r600_pipe_state_add_reg(rstate, R_028414_CB_BLEND_RED, fui(state->color[0]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028418_CB_BLEND_GREEN, fui(state->color[1]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02841C_CB_BLEND_BLUE, fui(state->color[2]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028420_CB_BLEND_ALPHA, fui(state->color[3]), 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028414_CB_BLEND_RED, fui(state->color[0]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028418_CB_BLEND_GREEN, fui(state->color[1]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02841C_CB_BLEND_BLUE, fui(state->color[2]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028420_CB_BLEND_ALPHA, fui(state->color[3]), 0xFFFFFFFF, NULL, 0); free(rctx->states[R600_PIPE_STATE_BLEND_COLOR]); rctx->states[R600_PIPE_STATE_BLEND_COLOR] = rstate; r600_context_pipe_state_set(&rctx->ctx, rstate); @@ -742,7 +742,7 @@ static void *r600_create_blend_state(struct pipe_context *ctx, blend->cb_target_mask = target_mask; /* MULTIWRITE_ENABLE is controlled by r600_pipe_shader_ps(). */ r600_pipe_state_add_reg(rstate, R_028808_CB_COLOR_CONTROL, - color_control, 0xFFFFFFFD, NULL); + color_control, 0xFFFFFFFD, NULL, 0); for (int i = 0; i < 8; i++) { /* state->rt entries > 0 only written if independent blending */ @@ -773,9 +773,9 @@ static void *r600_create_blend_state(struct pipe_context *ctx, /* R600 does not support per-MRT blends */ if (rctx->family > CHIP_R600) - r600_pipe_state_add_reg(rstate, R_028780_CB_BLEND0_CONTROL + i * 4, bc, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028780_CB_BLEND0_CONTROL + i * 4, bc, 0xFFFFFFFF, NULL, 0); if (i == 0) - r600_pipe_state_add_reg(rstate, R_028804_CB_BLEND_CONTROL, bc, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028804_CB_BLEND_CONTROL, bc, 0xFFFFFFFF, NULL, 0); } return rstate; } @@ -842,28 +842,28 @@ static void *r600_create_dsa_state(struct pipe_context *ctx, S_028D10_FORCE_HIS_ENABLE0(V_028D10_FORCE_DISABLE) | S_028D10_FORCE_HIS_ENABLE1(V_028D10_FORCE_DISABLE); /* TODO db_render_override depends on query */ - r600_pipe_state_add_reg(rstate, R_028028_DB_STENCIL_CLEAR, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02802C_DB_DEPTH_CLEAR, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028410_SX_ALPHA_TEST_CONTROL, alpha_test_control, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028028_DB_STENCIL_CLEAR, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02802C_DB_DEPTH_CLEAR, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028410_SX_ALPHA_TEST_CONTROL, alpha_test_control, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028430_DB_STENCILREFMASK, stencil_ref_mask, - 0xFFFFFFFF & C_028430_STENCILREF, NULL); + 0xFFFFFFFF & C_028430_STENCILREF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028434_DB_STENCILREFMASK_BF, stencil_ref_mask_bf, - 0xFFFFFFFF & C_028434_STENCILREF_BF, NULL); - r600_pipe_state_add_reg(rstate, R_0286E0_SPI_FOG_FUNC_SCALE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0286E4_SPI_FOG_FUNC_BIAS, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0286DC_SPI_FOG_CNTL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028800_DB_DEPTH_CONTROL, db_depth_control, 0xFFFFFFFF, NULL); + 0xFFFFFFFF & C_028434_STENCILREF_BF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0286E0_SPI_FOG_FUNC_SCALE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0286E4_SPI_FOG_FUNC_BIAS, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0286DC_SPI_FOG_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028800_DB_DEPTH_CONTROL, db_depth_control, 0xFFFFFFFF, NULL, 0); /* The DB_SHADER_CONTROL mask is 0xFFFFFFBC since Z_EXPORT_ENABLE, * STENCIL_EXPORT_ENABLE and KILL_ENABLE are controlled by * r600_pipe_shader_ps().*/ - r600_pipe_state_add_reg(rstate, R_02880C_DB_SHADER_CONTROL, db_shader_control, 0xFFFFFFBC, NULL); - r600_pipe_state_add_reg(rstate, R_028D0C_DB_RENDER_CONTROL, db_render_control, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028D10_DB_RENDER_OVERRIDE, db_render_override, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028D2C_DB_SRESULTS_COMPARE_STATE1, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028D30_DB_PRELOAD_CONTROL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028D44_DB_ALPHA_TO_MASK, 0x0000AA00, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_02880C_DB_SHADER_CONTROL, db_shader_control, 0xFFFFFFBC, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028D0C_DB_RENDER_CONTROL, db_render_control, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028D10_DB_RENDER_OVERRIDE, db_render_override, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028D2C_DB_SRESULTS_COMPARE_STATE1, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028D30_DB_PRELOAD_CONTROL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028D44_DB_ALPHA_TO_MASK, 0x0000AA00, 0xFFFFFFFF, NULL, 0); return rstate; } @@ -907,7 +907,7 @@ static void *r600_create_rs_state(struct pipe_context *ctx, tmp |= S_0286D4_PNT_SPRITE_TOP_1(1); } } - r600_pipe_state_add_reg(rstate, R_0286D4_SPI_INTERP_CONTROL_0, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_0286D4_SPI_INTERP_CONTROL_0, tmp, 0xFFFFFFFF, NULL, 0); polygon_dual_mode = (state->fill_front != PIPE_POLYGON_MODE_FILL || state->fill_back != PIPE_POLYGON_MODE_FILL); @@ -921,33 +921,33 @@ static void *r600_create_rs_state(struct pipe_context *ctx, S_028814_POLY_OFFSET_PARA_ENABLE(state->offset_tri) | S_028814_POLY_MODE(polygon_dual_mode) | S_028814_POLYMODE_FRONT_PTYPE(r600_translate_fill(state->fill_front)) | - S_028814_POLYMODE_BACK_PTYPE(r600_translate_fill(state->fill_back)), 0xFFFFFFFF, NULL); + S_028814_POLYMODE_BACK_PTYPE(r600_translate_fill(state->fill_back)), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_02881C_PA_CL_VS_OUT_CNTL, S_02881C_USE_VTX_POINT_SIZE(state->point_size_per_vertex) | - S_02881C_VS_OUT_MISC_VEC_ENA(state->point_size_per_vertex), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028820_PA_CL_NANINF_CNTL, 0x00000000, 0xFFFFFFFF, NULL); + S_02881C_VS_OUT_MISC_VEC_ENA(state->point_size_per_vertex), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028820_PA_CL_NANINF_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); /* point size 12.4 fixed point */ tmp = (unsigned)(state->point_size * 8.0); - r600_pipe_state_add_reg(rstate, R_028A00_PA_SU_POINT_SIZE, S_028A00_HEIGHT(tmp) | S_028A00_WIDTH(tmp), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A04_PA_SU_POINT_MINMAX, 0x80000000, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028A00_PA_SU_POINT_SIZE, S_028A00_HEIGHT(tmp) | S_028A00_WIDTH(tmp), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A04_PA_SU_POINT_MINMAX, 0x80000000, 0xFFFFFFFF, NULL, 0); tmp = (unsigned)state->line_width * 8; - r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, S_028A08_WIDTH(tmp), 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, S_028A08_WIDTH(tmp), 0xFFFFFFFF, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028A0C_PA_SC_LINE_STIPPLE, 0x00000005, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A48_PA_SC_MPASS_PS_CNTL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028C00_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028A0C_PA_SC_LINE_STIPPLE, 0x00000005, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A48_PA_SC_MPASS_PS_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028C00_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C08_PA_SU_VTX_CNTL, S_028C08_PIX_CENTER_HALF(state->gl_rasterization_rules), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); - r600_pipe_state_add_reg(rstate, R_028C0C_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028C10_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028C14_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028C18_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028DFC_PA_SU_POLY_OFFSET_CLAMP, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02820C_PA_SC_CLIPRECT_RULE, clip_rule, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028C0C_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028C10_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028C14_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028C18_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028DFC_PA_SU_POLY_OFFSET_CLAMP, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02820C_PA_SC_CLIPRECT_RULE, clip_rule, 0xFFFFFFFF, NULL, 0); return rstate; } @@ -977,17 +977,17 @@ static void *r600_create_sampler_state(struct pipe_context *ctx, S_03C000_MIP_FILTER(r600_tex_mipfilter(state->min_mip_filter)) | S_03C000_MAX_ANISO(r600_tex_aniso_filter(state->max_anisotropy)) | S_03C000_DEPTH_COMPARE_FUNCTION(r600_tex_compare(state->compare_func)) | - S_03C000_BORDER_COLOR_TYPE(uc.ui ? V_03C000_SQ_TEX_BORDER_COLOR_REGISTER : 0), 0xFFFFFFFF, NULL); + S_03C000_BORDER_COLOR_TYPE(uc.ui ? V_03C000_SQ_TEX_BORDER_COLOR_REGISTER : 0), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg_noblock(rstate, R_03C004_SQ_TEX_SAMPLER_WORD1_0, S_03C004_MIN_LOD(S_FIXED(CLAMP(state->min_lod, 0, 15), 6)) | S_03C004_MAX_LOD(S_FIXED(CLAMP(state->max_lod, 0, 15), 6)) | - S_03C004_LOD_BIAS(S_FIXED(CLAMP(state->lod_bias, -16, 16), 6)), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg_noblock(rstate, R_03C008_SQ_TEX_SAMPLER_WORD2_0, S_03C008_TYPE(1), 0xFFFFFFFF, NULL); + S_03C004_LOD_BIAS(S_FIXED(CLAMP(state->lod_bias, -16, 16), 6)), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg_noblock(rstate, R_03C008_SQ_TEX_SAMPLER_WORD2_0, S_03C008_TYPE(1), 0xFFFFFFFF, NULL, 0); if (uc.ui) { - r600_pipe_state_add_reg_noblock(rstate, R_00A400_TD_PS_SAMPLER0_BORDER_RED, fui(state->border_color[0]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg_noblock(rstate, R_00A404_TD_PS_SAMPLER0_BORDER_GREEN, fui(state->border_color[1]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg_noblock(rstate, R_00A408_TD_PS_SAMPLER0_BORDER_BLUE, fui(state->border_color[2]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg_noblock(rstate, R_00A40C_TD_PS_SAMPLER0_BORDER_ALPHA, fui(state->border_color[3]), 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg_noblock(rstate, R_00A400_TD_PS_SAMPLER0_BORDER_RED, fui(state->border_color[0]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg_noblock(rstate, R_00A404_TD_PS_SAMPLER0_BORDER_GREEN, fui(state->border_color[1]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg_noblock(rstate, R_00A408_TD_PS_SAMPLER0_BORDER_BLUE, fui(state->border_color[2]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg_noblock(rstate, R_00A40C_TD_PS_SAMPLER0_BORDER_ALPHA, fui(state->border_color[3]), 0xFFFFFFFF, NULL, 0); } return rstate; } @@ -996,10 +996,9 @@ static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *c struct pipe_resource *texture, const struct pipe_sampler_view *state) { - struct r600_pipe_sampler_view *resource = CALLOC_STRUCT(r600_pipe_sampler_view); + struct r600_pipe_sampler_view *view = CALLOC_STRUCT(r600_pipe_sampler_view); struct r600_pipe_resource_state *rstate; - const struct util_format_description *desc; - struct r600_resource_texture *tmp; + struct r600_resource_texture *tmp = (struct r600_resource_texture*)texture; struct r600_resource *rbuffer; unsigned format, endian; uint32_t word4 = 0, yuv_format = 0, pitch = 0; @@ -1007,43 +1006,42 @@ static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *c struct r600_bo *bo[2]; unsigned width, height, depth, offset_level, last_level; - if (resource == NULL) + if (view == NULL) return NULL; - rstate = &resource->state; + rstate = &view->state; /* initialize base object */ - resource->base = *state; - resource->base.texture = NULL; + view->base = *state; + view->base.texture = NULL; pipe_reference(NULL, &texture->reference); - resource->base.texture = texture; - resource->base.reference.count = 1; - resource->base.context = ctx; + view->base.texture = texture; + view->base.reference.count = 1; + view->base.context = ctx; swizzle[0] = state->swizzle_r; swizzle[1] = state->swizzle_g; swizzle[2] = state->swizzle_b; swizzle[3] = state->swizzle_a; + format = r600_translate_texformat(ctx->screen, state->format, swizzle, &word4, &yuv_format); if (format == ~0) { format = 0; } - desc = util_format_description(state->format); - if (desc == NULL) { - R600_ERR("unknown format %d\n", state->format); - } - tmp = (struct r600_resource_texture *)texture; + if (tmp->depth && !tmp->is_flushing_texture) { r600_texture_depth_flush(ctx, texture, TRUE); tmp = tmp->flushed_depth_texture; } + endian = r600_colorformat_endian_swap(format); if (tmp->force_int_type) { word4 &= C_038010_NUM_FORMAT_ALL; word4 |= S_038010_NUM_FORMAT_ALL(V_038010_SQ_NUM_FORMAT_INT); } + rbuffer = &tmp->resource; bo[0] = rbuffer->bo; bo[1] = rbuffer->bo; @@ -1068,6 +1066,8 @@ static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *c rstate->bo[0] = bo[0]; rstate->bo[1] = bo[1]; + rstate->bo_usage[0] = RADEON_USAGE_READ; + rstate->bo_usage[1] = RADEON_USAGE_READ; rstate->val[0] = (S_038000_DIM(r600_tex_dim(texture->target)) | S_038000_TILE_MODE(array_mode) | @@ -1077,8 +1077,8 @@ static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *c rstate->val[1] = (S_038004_TEX_HEIGHT(height - 1) | S_038004_TEX_DEPTH(depth - 1) | S_038004_DATA_FORMAT(format)); - rstate->val[2] = (tmp->offset[offset_level] + r600_bo_offset(bo[0])) >> 8; - rstate->val[3] = (tmp->offset[offset_level+1] + r600_bo_offset(bo[1])) >> 8; + rstate->val[2] = tmp->offset[offset_level] >> 8; + rstate->val[3] = tmp->offset[offset_level+1] >> 8; rstate->val[4] = (word4 | S_038010_SRF_MODE_ALL(V_038010_SRF_MODE_ZERO_CLAMP_MINUS_ONE) | S_038010_REQUEST_SIZE(1) | @@ -1090,7 +1090,7 @@ static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *c rstate->val[6] = (S_038018_TYPE(V_038010_SQ_TEX_VTX_VALID_TEXTURE) | S_038018_MAX_ANISO(4 /* max 16 samples */)); - return &resource->base; + return &view->base; } static void r600_set_vs_sampler_view(struct pipe_context *ctx, unsigned count, @@ -1157,7 +1157,7 @@ static void r600_set_seamless_cubemap(struct r600_pipe_context *rctx, boolean en rstate->id = R600_PIPE_STATE_SEAMLESS_CUBEMAP; r600_pipe_state_add_reg(rstate, R_009508_TA_CNTL_AUX, (enable ? 0 : S_009508_DISABLE_CUBE_WRAP(1)), - 1, NULL); + 1, NULL, 0); free(rctx->states[R600_PIPE_STATE_SEAMLESS_CUBEMAP]); rctx->states[R600_PIPE_STATE_SEAMLESS_CUBEMAP] = rstate; @@ -1215,21 +1215,21 @@ static void r600_set_clip_state(struct pipe_context *ctx, for (int i = 0; i < state->nr; i++) { r600_pipe_state_add_reg(rstate, R_028E20_PA_CL_UCP0_X + i * 16, - fui(state->ucp[i][0]), 0xFFFFFFFF, NULL); + fui(state->ucp[i][0]), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028E24_PA_CL_UCP0_Y + i * 16, - fui(state->ucp[i][1]) , 0xFFFFFFFF, NULL); + fui(state->ucp[i][1]) , 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028E28_PA_CL_UCP0_Z + i * 16, - fui(state->ucp[i][2]), 0xFFFFFFFF, NULL); + fui(state->ucp[i][2]), 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028E2C_PA_CL_UCP0_W + i * 16, - fui(state->ucp[i][3]), 0xFFFFFFFF, NULL); + fui(state->ucp[i][3]), 0xFFFFFFFF, NULL, 0); } r600_pipe_state_add_reg(rstate, R_028810_PA_CL_CLIP_CNTL, S_028810_PS_UCP_MODE(3) | ((1 << state->nr) - 1) | S_028810_ZCLIP_NEAR_DISABLE(state->depth_clamp) | - S_028810_ZCLIP_FAR_DISABLE(state->depth_clamp), 0xFFFFFFFF, NULL); + S_028810_ZCLIP_FAR_DISABLE(state->depth_clamp), 0xFFFFFFFF, NULL, 0); free(rctx->states[R600_PIPE_STATE_CLIP]); rctx->states[R600_PIPE_STATE_CLIP] = rstate; @@ -1260,28 +1260,28 @@ static void r600_set_scissor_state(struct pipe_context *ctx, br = S_028244_BR_X(state->maxx) | S_028244_BR_Y(state->maxy); r600_pipe_state_add_reg(rstate, R_028210_PA_SC_CLIPRECT_0_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028214_PA_SC_CLIPRECT_0_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028218_PA_SC_CLIPRECT_1_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_02821C_PA_SC_CLIPRECT_1_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028220_PA_SC_CLIPRECT_2_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028224_PA_SC_CLIPRECT_2_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028228_PA_SC_CLIPRECT_3_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_02822C_PA_SC_CLIPRECT_3_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); free(rctx->states[R600_PIPE_STATE_SCISSOR]); rctx->states[R600_PIPE_STATE_SCISSOR] = rstate; @@ -1303,11 +1303,11 @@ static void r600_set_stencil_ref(struct pipe_context *ctx, tmp = S_028430_STENCILREF(state->ref_value[0]); r600_pipe_state_add_reg(rstate, R_028430_DB_STENCILREFMASK, tmp, - ~C_028430_STENCILREF, NULL); + ~C_028430_STENCILREF, NULL, 0); tmp = S_028434_STENCILREF_BF(state->ref_value[1]); r600_pipe_state_add_reg(rstate, R_028434_DB_STENCILREFMASK_BF, tmp, - ~C_028434_STENCILREF_BF, NULL); + ~C_028434_STENCILREF_BF, NULL, 0); free(rctx->states[R600_PIPE_STATE_STENCIL_REF]); rctx->states[R600_PIPE_STATE_STENCIL_REF] = rstate; @@ -1325,15 +1325,15 @@ static void r600_set_viewport_state(struct pipe_context *ctx, rctx->viewport = *state; rstate->id = R600_PIPE_STATE_VIEWPORT; - r600_pipe_state_add_reg(rstate, R_0282D0_PA_SC_VPORT_ZMIN_0, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0282D4_PA_SC_VPORT_ZMAX_0, 0x3F800000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02843C_PA_CL_VPORT_XSCALE_0, fui(state->scale[0]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028444_PA_CL_VPORT_YSCALE_0, fui(state->scale[1]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_02844C_PA_CL_VPORT_ZSCALE_0, fui(state->scale[2]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028440_PA_CL_VPORT_XOFFSET_0, fui(state->translate[0]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028448_PA_CL_VPORT_YOFFSET_0, fui(state->translate[1]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028450_PA_CL_VPORT_ZOFFSET_0, fui(state->translate[2]), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028818_PA_CL_VTE_CNTL, 0x0000043F, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_0282D0_PA_SC_VPORT_ZMIN_0, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0282D4_PA_SC_VPORT_ZMAX_0, 0x3F800000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02843C_PA_CL_VPORT_XSCALE_0, fui(state->scale[0]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028444_PA_CL_VPORT_YSCALE_0, fui(state->scale[1]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_02844C_PA_CL_VPORT_ZSCALE_0, fui(state->scale[2]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028440_PA_CL_VPORT_XOFFSET_0, fui(state->translate[0]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028448_PA_CL_VPORT_YOFFSET_0, fui(state->translate[1]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028450_PA_CL_VPORT_ZOFFSET_0, fui(state->translate[2]), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028818_PA_CL_VTE_CNTL, 0x0000043F, 0xFFFFFFFF, NULL, 0); free(rctx->states[R600_PIPE_STATE_VIEWPORT]); rctx->states[R600_PIPE_STATE_VIEWPORT] = rstate; @@ -1441,27 +1441,27 @@ static void r600_cb(struct r600_pipe_context *rctx, struct r600_pipe_state *rsta r600_pipe_state_add_reg(rstate, R_028040_CB_COLOR0_BASE + cb * 4, - (offset + r600_bo_offset(bo[0])) >> 8, 0xFFFFFFFF, bo[0]); + offset >> 8, 0xFFFFFFFF, bo[0], RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_0280A0_CB_COLOR0_INFO + cb * 4, - color_info, 0xFFFFFFFF, bo[0]); + color_info, 0xFFFFFFFF, bo[0], RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_028060_CB_COLOR0_SIZE + cb * 4, S_028060_PITCH_TILE_MAX(pitch) | S_028060_SLICE_TILE_MAX(slice), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028080_CB_COLOR0_VIEW + cb * 4, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0280E0_CB_COLOR0_FRAG + cb * 4, - r600_bo_offset(bo[1]) >> 8, 0xFFFFFFFF, bo[1]); + 0, 0xFFFFFFFF, bo[1], RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_0280C0_CB_COLOR0_TILE + cb * 4, - r600_bo_offset(bo[2]) >> 8, 0xFFFFFFFF, bo[2]); + 0, 0xFFFFFFFF, bo[2], RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_028100_CB_COLOR0_MASK + cb * 4, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); } static void r600_db(struct r600_pipe_context *rctx, struct r600_pipe_state *rstate, @@ -1492,16 +1492,16 @@ static void r600_db(struct r600_pipe_context *rctx, struct r600_pipe_state *rsta format = r600_translate_dbformat(state->zsbuf->texture->format); r600_pipe_state_add_reg(rstate, R_02800C_DB_DEPTH_BASE, - (offset + r600_bo_offset(rbuffer->bo)) >> 8, 0xFFFFFFFF, rbuffer->bo); + offset >> 8, 0xFFFFFFFF, rbuffer->bo, RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_028000_DB_DEPTH_SIZE, S_028000_PITCH_TILE_MAX(pitch) | S_028000_SLICE_TILE_MAX(slice), - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028004_DB_DEPTH_VIEW, 0x00000000, 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028004_DB_DEPTH_VIEW, 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028010_DB_DEPTH_INFO, S_028010_ARRAY_MODE(rtex->array_mode[level]) | S_028010_FORMAT(format), - 0xFFFFFFFF, rbuffer->bo); + 0xFFFFFFFF, rbuffer->bo, RADEON_USAGE_READWRITE); r600_pipe_state_add_reg(rstate, R_028D34_DB_PREFETCH_LIMIT, - (surf->aligned_height / 8) - 1, 0xFFFFFFFF, NULL); + (surf->aligned_height / 8) - 1, 0xFFFFFFFF, NULL, 0); } static void r600_set_framebuffer_state(struct pipe_context *ctx, @@ -1546,59 +1546,59 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_028030_PA_SC_SCREEN_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028034_PA_SC_SCREEN_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028204_PA_SC_WINDOW_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028208_PA_SC_WINDOW_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028240_PA_SC_GENERIC_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028244_PA_SC_GENERIC_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028250_PA_SC_VPORT_SCISSOR_0_TL, tl, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028254_PA_SC_VPORT_SCISSOR_0_BR, br, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028200_PA_SC_WINDOW_OFFSET, 0x00000000, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); if (rctx->chip_class >= R700) { r600_pipe_state_add_reg(rstate, R_028230_PA_SC_EDGERULE, 0xAAAAAAAA, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); } r600_pipe_state_add_reg(rstate, R_0287A0_CB_SHADER_CONTROL, - shader_control, 0xFFFFFFFF, NULL); + shader_control, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028238_CB_TARGET_MASK, - 0x00000000, target_mask, NULL); + 0x00000000, target_mask, NULL, 0); r600_pipe_state_add_reg(rstate, R_02823C_CB_SHADER_MASK, - shader_mask, 0xFFFFFFFF, NULL); + shader_mask, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C04_PA_SC_AA_CONFIG, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C1C_PA_SC_AA_SAMPLE_LOCS_MCTX, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C20_PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C30_CB_CLRCMP_CONTROL, - 0x01000000, 0xFFFFFFFF, NULL); + 0x01000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C34_CB_CLRCMP_SRC, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C38_CB_CLRCMP_DST, - 0x000000FF, 0xFFFFFFFF, NULL); + 0x000000FF, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C3C_CB_CLRCMP_MSK, - 0xFFFFFFFF, 0xFFFFFFFF, NULL); + 0xFFFFFFFF, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028C48_PA_SC_AA_MASK, - 0xFFFFFFFF, 0xFFFFFFFF, NULL); + 0xFFFFFFFF, 0xFFFFFFFF, NULL, 0); free(rctx->states[R600_PIPE_STATE_FRAMEBUFFER]); rctx->states[R600_PIPE_STATE_FRAMEBUFFER] = rstate; @@ -1674,7 +1674,7 @@ void r600_adjust_gprs(struct r600_pipe_context *rctx) if (rctx->chip_class >= EVERGREEN) return; - if (!rctx->ps_shader && !rctx->vs_shader) + if (!rctx->ps_shader || !rctx->vs_shader) return; if (rctx->ps_shader->shader.bc.ngpr > rctx->default_ps_gprs) @@ -1695,7 +1695,7 @@ void r600_adjust_gprs(struct r600_pipe_context *rctx) tmp |= S_008C04_NUM_PS_GPRS(num_ps_gprs); tmp |= S_008C04_NUM_VS_GPRS(num_vs_gprs); rstate.nregs = 0; - r600_pipe_state_add_reg(&rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, tmp, 0x0FFFFFFF, NULL); + r600_pipe_state_add_reg(&rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, tmp, 0x0FFFFFFF, NULL, 0); r600_context_pipe_state_set(&rctx->ctx, &rstate); } @@ -1866,20 +1866,20 @@ void r600_init_config(struct r600_pipe_context *rctx) tmp |= S_008C00_VS_PRIO(vs_prio); tmp |= S_008C00_GS_PRIO(gs_prio); tmp |= S_008C00_ES_PRIO(es_prio); - r600_pipe_state_add_reg(rstate, R_008C00_SQ_CONFIG, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C00_SQ_CONFIG, tmp, 0xFFFFFFFF, NULL, 0); /* SQ_GPR_RESOURCE_MGMT_1 */ tmp = 0; tmp |= S_008C04_NUM_PS_GPRS(num_ps_gprs); tmp |= S_008C04_NUM_VS_GPRS(num_vs_gprs); tmp |= S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs); - r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C04_SQ_GPR_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL, 0); /* SQ_GPR_RESOURCE_MGMT_2 */ tmp = 0; tmp |= S_008C08_NUM_GS_GPRS(num_gs_gprs); tmp |= S_008C08_NUM_ES_GPRS(num_es_gprs); - r600_pipe_state_add_reg(rstate, R_008C08_SQ_GPR_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C08_SQ_GPR_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL, 0); /* SQ_THREAD_RESOURCE_MGMT */ tmp = 0; @@ -1887,78 +1887,78 @@ void r600_init_config(struct r600_pipe_context *rctx) tmp |= S_008C0C_NUM_VS_THREADS(num_vs_threads); tmp |= S_008C0C_NUM_GS_THREADS(num_gs_threads); tmp |= S_008C0C_NUM_ES_THREADS(num_es_threads); - r600_pipe_state_add_reg(rstate, R_008C0C_SQ_THREAD_RESOURCE_MGMT, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C0C_SQ_THREAD_RESOURCE_MGMT, tmp, 0xFFFFFFFF, NULL, 0); /* SQ_STACK_RESOURCE_MGMT_1 */ tmp = 0; tmp |= S_008C10_NUM_PS_STACK_ENTRIES(num_ps_stack_entries); tmp |= S_008C10_NUM_VS_STACK_ENTRIES(num_vs_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C10_SQ_STACK_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C10_SQ_STACK_RESOURCE_MGMT_1, tmp, 0xFFFFFFFF, NULL, 0); /* SQ_STACK_RESOURCE_MGMT_2 */ tmp = 0; tmp |= S_008C14_NUM_GS_STACK_ENTRIES(num_gs_stack_entries); tmp |= S_008C14_NUM_ES_STACK_ENTRIES(num_es_stack_entries); - r600_pipe_state_add_reg(rstate, R_008C14_SQ_STACK_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008C14_SQ_STACK_RESOURCE_MGMT_2, tmp, 0xFFFFFFFF, NULL, 0); - r600_pipe_state_add_reg(rstate, R_009714_VC_ENHANCE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028350_SX_MISC, 0x00000000, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_009714_VC_ENHANCE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028350_SX_MISC, 0x00000000, 0xFFFFFFFF, NULL, 0); if (rctx->chip_class >= R700) { - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x00004000, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x00004000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_009508_TA_CNTL_AUX, S_009508_DISABLE_CUBE_ANISO(1) | S_009508_SYNC_GRADIENT(1) | S_009508_SYNC_WALKER(1) | - S_009508_SYNC_ALIGNER(1), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_009830_DB_DEBUG, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_009838_DB_WATERMARKS, 0x00420204, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0286C8_SPI_THREAD_GROUPING, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL, 0x00514002, 0xFFFFFFFF, NULL); + S_009508_SYNC_ALIGNER(1), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_009830_DB_DEBUG, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_009838_DB_WATERMARKS, 0x00420204, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0286C8_SPI_THREAD_GROUPING, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL, 0x00514002, 0xFFFFFFFF, NULL, 0); } else { - r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x00000000, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_009508_TA_CNTL_AUX, S_009508_DISABLE_CUBE_ANISO(1) | S_009508_SYNC_GRADIENT(1) | S_009508_SYNC_WALKER(1) | - S_009508_SYNC_ALIGNER(1), 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_009830_DB_DEBUG, 0x82000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_009838_DB_WATERMARKS, 0x01020204, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0286C8_SPI_THREAD_GROUPING, 0x00000001, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL, 0x00004012, 0xFFFFFFFF, NULL); - } - r600_pipe_state_add_reg(rstate, R_0288A8_SQ_ESGS_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0288AC_SQ_GSVS_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0288B0_SQ_ESTMP_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0288B4_SQ_GSTMP_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0288B8_SQ_VSTMP_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0288BC_SQ_PSTMP_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0288C0_SQ_FBUF_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0288C4_SQ_REDUC_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0288C8_SQ_GS_VERT_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A14_VGT_HOS_CNTL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A20_VGT_HOS_REUSE_DEPTH, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A24_VGT_GROUP_PRIM_TYPE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A28_VGT_GROUP_FIRST_DECR, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A2C_VGT_GROUP_DECR, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A30_VGT_GROUP_VECT_0_CNTL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A34_VGT_GROUP_VECT_1_CNTL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A40_VGT_GS_MODE, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AB0_VGT_STRMOUT_EN, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AB4_VGT_REUSE_OFF, 0x00000001, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AB8_VGT_VTX_CNT_EN, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028B20_VGT_STRMOUT_BUFFER_EN, 0x00000000, 0xFFFFFFFF, NULL); - - r600_pipe_state_add_reg(rstate, R_02840C_VGT_MULTI_PRIM_IB_RESET_INDX, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A84_VGT_PRIMITIVEID_EN, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AA0_VGT_INSTANCE_STEP_RATE_0, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028AA4_VGT_INSTANCE_STEP_RATE_1, 0x00000000, 0xFFFFFFFF, NULL); + S_009508_SYNC_ALIGNER(1), 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_009830_DB_DEBUG, 0x82000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_009838_DB_WATERMARKS, 0x01020204, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0286C8_SPI_THREAD_GROUPING, 0x00000001, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL, 0x00004012, 0xFFFFFFFF, NULL, 0); + } + r600_pipe_state_add_reg(rstate, R_0288A8_SQ_ESGS_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0288AC_SQ_GSVS_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0288B0_SQ_ESTMP_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0288B4_SQ_GSTMP_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0288B8_SQ_VSTMP_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0288BC_SQ_PSTMP_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0288C0_SQ_FBUF_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0288C4_SQ_REDUC_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0288C8_SQ_GS_VERT_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A14_VGT_HOS_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A18_VGT_HOS_MAX_TESS_LEVEL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A20_VGT_HOS_REUSE_DEPTH, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A24_VGT_GROUP_PRIM_TYPE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A28_VGT_GROUP_FIRST_DECR, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A2C_VGT_GROUP_DECR, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A30_VGT_GROUP_VECT_0_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A34_VGT_GROUP_VECT_1_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A38_VGT_GROUP_VECT_0_FMT_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A40_VGT_GS_MODE, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AB0_VGT_STRMOUT_EN, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AB4_VGT_REUSE_OFF, 0x00000001, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AB8_VGT_VTX_CNT_EN, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028B20_VGT_STRMOUT_BUFFER_EN, 0x00000000, 0xFFFFFFFF, NULL, 0); + + r600_pipe_state_add_reg(rstate, R_02840C_VGT_MULTI_PRIM_IB_RESET_INDX, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A84_VGT_PRIMITIVEID_EN, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AA0_VGT_INSTANCE_STEP_RATE_0, 0x00000000, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_028AA4_VGT_INSTANCE_STEP_RATE_1, 0x00000000, 0xFFFFFFFF, NULL, 0); r600_context_pipe_state_set(&rctx->ctx, rstate); } @@ -2022,38 +2022,38 @@ void r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader *shad S_0286D0_FRONT_FACE_ADDR(rshader->input[face_index].gpr); } - r600_pipe_state_add_reg(rstate, R_0286CC_SPI_PS_IN_CONTROL_0, spi_ps_in_control_0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0286D0_SPI_PS_IN_CONTROL_1, spi_ps_in_control_1, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0286D8_SPI_INPUT_Z, spi_input_z, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_0286CC_SPI_PS_IN_CONTROL_0, spi_ps_in_control_0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0286D0_SPI_PS_IN_CONTROL_1, spi_ps_in_control_1, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(rstate, R_0286D8_SPI_INPUT_Z, spi_input_z, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028840_SQ_PGM_START_PS, - r600_bo_offset(shader->bo) >> 8, 0xFFFFFFFF, shader->bo); + 0, 0xFFFFFFFF, shader->bo, RADEON_USAGE_READ); r600_pipe_state_add_reg(rstate, R_028850_SQ_PGM_RESOURCES_PS, S_028868_NUM_GPRS(rshader->bc.ngpr) | S_028868_STACK_SIZE(rshader->bc.nstack), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028854_SQ_PGM_EXPORTS_PS, - exports_ps, 0xFFFFFFFF, NULL); + exports_ps, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0288CC_SQ_PGM_CF_OFFSET_PS, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028808_CB_COLOR_CONTROL, S_028808_MULTIWRITE_ENABLE(!!rshader->fs_write_all), S_028808_MULTIWRITE_ENABLE(1), - NULL); + NULL, 0); /* only set some bits here, the other bits are set in the dsa state */ r600_pipe_state_add_reg(rstate, R_02880C_DB_SHADER_CONTROL, db_shader_control, S_02880C_Z_EXPORT_ENABLE(1) | S_02880C_STENCIL_REF_EXPORT_ENABLE(1) | S_02880C_KILL_ENABLE(1), - NULL); + NULL, 0); r600_pipe_state_add_reg(rstate, R_03E200_SQ_LOOP_CONST_0, 0x01000FFF, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); } void r600_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shader) @@ -2062,7 +2062,7 @@ void r600_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shad struct r600_pipe_state *rstate = &shader->rstate; struct r600_shader *rshader = &shader->shader; unsigned spi_vs_out_id[10]; - unsigned i, tmp; + unsigned i, tmp, nparams; /* clear previous register */ rstate->nregs = 0; @@ -2081,28 +2081,36 @@ void r600_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *shad for (i = 0; i < 10; i++) { r600_pipe_state_add_reg(rstate, R_028614_SPI_VS_OUT_ID_0 + i * 4, - spi_vs_out_id[i], 0xFFFFFFFF, NULL); + spi_vs_out_id[i], 0xFFFFFFFF, NULL, 0); } + /* Certain attributes (position, psize, etc.) don't count as params. + * VS is required to export at least one param and r600_shader_from_tgsi() + * takes care of adding a dummy export. + */ + nparams = rshader->noutput - rshader->npos; + if (nparams < 1) + nparams = 1; + r600_pipe_state_add_reg(rstate, R_0286C4_SPI_VS_OUT_CONFIG, - S_0286C4_VS_EXPORT_COUNT(rshader->noutput - 2), - 0xFFFFFFFF, NULL); + S_0286C4_VS_EXPORT_COUNT(nparams - 1), + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028868_SQ_PGM_RESOURCES_VS, S_028868_NUM_GPRS(rshader->bc.ngpr) | S_028868_STACK_SIZE(rshader->bc.nstack), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0288D0_SQ_PGM_CF_OFFSET_VS, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028858_SQ_PGM_START_VS, - r600_bo_offset(shader->bo) >> 8, 0xFFFFFFFF, shader->bo); + 0, 0xFFFFFFFF, shader->bo, RADEON_USAGE_READ); r600_pipe_state_add_reg(rstate, R_03E200_SQ_LOOP_CONST_0 + (32 * 4), 0x01000FFF, - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); } void r600_fetch_shader(struct pipe_context *ctx, @@ -2115,12 +2123,12 @@ void r600_fetch_shader(struct pipe_context *ctx, rstate->id = R600_PIPE_STATE_FETCH_SHADER; rstate->nregs = 0; r600_pipe_state_add_reg(rstate, R_0288A4_SQ_PGM_RESOURCES_FS, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_0288DC_SQ_PGM_CF_OFFSET_FS, - 0x00000000, 0xFFFFFFFF, NULL); + 0x00000000, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(rstate, R_028894_SQ_PGM_START_FS, - r600_bo_offset(ve->fetch_shader) >> 8, - 0xFFFFFFFF, ve->fetch_shader); + 0, + 0xFFFFFFFF, ve->fetch_shader, RADEON_USAGE_READ); } void *r600_create_db_flush_dsa(struct r600_pipe_context *rctx) @@ -2149,7 +2157,7 @@ void *r600_create_db_flush_dsa(struct r600_pipe_context *rctx) r600_pipe_state_add_reg(rstate, R_02880C_DB_SHADER_CONTROL, 0x0, - S_02880C_DUAL_EXPORT_ENABLE(1), NULL); + S_02880C_DUAL_EXPORT_ENABLE(1), NULL, 0); r600_pipe_state_add_reg(rstate, R_028D0C_DB_RENDER_CONTROL, S_028D0C_DEPTH_COPY_ENABLE(1) | @@ -2157,7 +2165,7 @@ void *r600_create_db_flush_dsa(struct r600_pipe_context *rctx) S_028D0C_COPY_CENTROID(1), S_028D0C_DEPTH_COPY_ENABLE(1) | S_028D0C_STENCIL_COPY_ENABLE(1) | - S_028D0C_COPY_CENTROID(1), NULL); + S_028D0C_COPY_CENTROID(1), NULL, 0); return rstate; } @@ -2178,10 +2186,12 @@ void r600_pipe_init_buffer_resource(struct r600_pipe_context *rctx, void r600_pipe_mod_buffer_resource(struct r600_pipe_resource_state *rstate, struct r600_resource *rbuffer, - unsigned offset, unsigned stride) + unsigned offset, unsigned stride, + enum radeon_bo_usage usage) { rstate->val[0] = offset; rstate->bo[0] = rbuffer->bo; + rstate->bo_usage[0] = usage; rstate->val[1] = rbuffer->bo_size - offset - 1; rstate->val[2] = S_038008_ENDIAN_SWAP(r600_endian_swap(32)) | S_038008_STRIDE(stride); diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c index 408eaed491b..853458f0156 100644 --- a/src/gallium/drivers/r600/r600_state_common.c +++ b/src/gallium/drivers/r600/r600_state_common.c @@ -150,7 +150,7 @@ void r600_delete_state(struct pipe_context *ctx, void *state) rctx->states[rstate->id] = NULL; } for (int i = 0; i < rstate->nregs; i++) { - r600_bo_reference(rctx->radeon, &rstate->regs[i].bo, NULL); + r600_bo_reference(&rstate->regs[i].bo, NULL); } free(rstate); } @@ -181,7 +181,7 @@ void r600_delete_vertex_element(struct pipe_context *ctx, void *state) if (rctx->vertex_elements == state) rctx->vertex_elements = NULL; - r600_bo_reference(rctx->radeon, &v->fetch_shader, NULL); + r600_bo_reference(&v->fetch_shader, NULL); u_vbuf_mgr_destroy_vertex_elements(rctx->vbuf_mgr, v->vmgr_elements); FREE(state); } @@ -336,7 +336,7 @@ static void r600_update_alpha_ref(struct r600_pipe_context *rctx) rstate.nregs = 0; if (rctx->export_16bpc) alpha_ref &= ~0x1FFF; - r600_pipe_state_add_reg(&rstate, R_028438_SX_ALPHA_REF, alpha_ref, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(&rstate, R_028438_SX_ALPHA_REF, alpha_ref, 0xFFFFFFFF, NULL, 0); r600_context_pipe_state_set(&rctx->ctx, &rstate); rctx->alpha_ref_dirty = false; @@ -349,7 +349,7 @@ static void r600_spi_block_init(struct r600_pipe_context *rctx, struct r600_pipe rstate->nregs = 0; rstate->id = R600_PIPE_STATE_SPI; for (i = 0; i < 32; i++) { - r600_pipe_state_add_reg(rstate, R_028644_SPI_PS_INPUT_CNTL_0 + i * 4, 0, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028644_SPI_PS_INPUT_CNTL_0 + i * 4, 0, 0xFFFFFFFF, NULL, 0); } } @@ -418,7 +418,6 @@ void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index, } r600_upload_const_buffer(rctx, &rbuffer, &offset); - offset += r600_bo_offset(rbuffer->r.bo); switch (shader) { case PIPE_SHADER_VERTEX: @@ -426,10 +425,10 @@ void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index, r600_pipe_state_add_reg(&rctx->vs_const_buffer, R_028180_ALU_CONST_BUFFER_SIZE_VS_0, ALIGN_DIVUP(buffer->width0 >> 4, 16), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&rctx->vs_const_buffer, R_028980_ALU_CONST_CACHE_VS_0, - offset >> 8, 0xFFFFFFFF, rbuffer->r.bo); + offset >> 8, 0xFFFFFFFF, rbuffer->r.bo, RADEON_USAGE_READ); r600_context_pipe_state_set(&rctx->ctx, &rctx->vs_const_buffer); rstate = &rctx->vs_const_buffer_resource[index]; @@ -442,10 +441,10 @@ void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index, } if (rctx->chip_class >= EVERGREEN) { - evergreen_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16); + evergreen_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16, RADEON_USAGE_READ); evergreen_context_pipe_state_set_vs_resource(&rctx->ctx, rstate, index); } else { - r600_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16); + r600_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16, RADEON_USAGE_READ); r600_context_pipe_state_set_vs_resource(&rctx->ctx, rstate, index); } break; @@ -454,10 +453,10 @@ void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index, r600_pipe_state_add_reg(&rctx->ps_const_buffer, R_028140_ALU_CONST_BUFFER_SIZE_PS_0, ALIGN_DIVUP(buffer->width0 >> 4, 16), - 0xFFFFFFFF, NULL); + 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&rctx->ps_const_buffer, R_028940_ALU_CONST_CACHE_PS_0, - offset >> 8, 0xFFFFFFFF, rbuffer->r.bo); + offset >> 8, 0xFFFFFFFF, rbuffer->r.bo, RADEON_USAGE_READ); r600_context_pipe_state_set(&rctx->ctx, &rctx->ps_const_buffer); rstate = &rctx->ps_const_buffer_resource[index]; @@ -469,10 +468,10 @@ void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index, } } if (rctx->chip_class >= EVERGREEN) { - evergreen_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16); + evergreen_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16, RADEON_USAGE_READ); evergreen_context_pipe_state_set_ps_resource(&rctx->ctx, rstate, index); } else { - r600_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16); + r600_pipe_mod_buffer_resource(rstate, &rbuffer->r, offset, 16, RADEON_USAGE_READ); r600_context_pipe_state_set_ps_resource(&rctx->ctx, rstate, index); } break; @@ -518,7 +517,7 @@ static void r600_vertex_buffer_update(struct r600_pipe_context *rctx) } if (vertex_buffer == NULL || rbuffer == NULL) continue; - offset += vertex_buffer->buffer_offset + r600_bo_offset(rbuffer->bo); + offset += vertex_buffer->buffer_offset; if (!rstate->id) { if (rctx->chip_class >= EVERGREEN) { @@ -529,10 +528,10 @@ static void r600_vertex_buffer_update(struct r600_pipe_context *rctx) } if (rctx->chip_class >= EVERGREEN) { - evergreen_pipe_mod_buffer_resource(rstate, rbuffer, offset, vertex_buffer->stride); + evergreen_pipe_mod_buffer_resource(rstate, rbuffer, offset, vertex_buffer->stride, RADEON_USAGE_READ); evergreen_context_pipe_state_set_fs_resource(&rctx->ctx, rstate, i); } else { - r600_pipe_mod_buffer_resource(rstate, rbuffer, offset, vertex_buffer->stride); + r600_pipe_mod_buffer_resource(rstate, rbuffer, offset, vertex_buffer->stride, RADEON_USAGE_READ); r600_context_pipe_state_set_fs_resource(&rctx->ctx, rstate, i); } } @@ -615,16 +614,18 @@ void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info) if (rctx->vgt.id != R600_PIPE_STATE_VGT) { rctx->vgt.id = R600_PIPE_STATE_VGT; rctx->vgt.nregs = 0; - r600_pipe_state_add_reg(&rctx->vgt, R_008958_VGT_PRIMITIVE_TYPE, prim, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(&rctx->vgt, R_028238_CB_TARGET_MASK, rctx->cb_target_mask & mask, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(&rctx->vgt, R_028400_VGT_MAX_VTX_INDX, draw.info.max_index, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(&rctx->vgt, R_028404_VGT_MIN_VTX_INDX, draw.info.min_index, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(&rctx->vgt, R_028408_VGT_INDX_OFFSET, draw.info.index_bias, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(&rctx->vgt, R_03CFF0_SQ_VTX_BASE_VTX_LOC, 0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(&rctx->vgt, R_03CFF4_SQ_VTX_START_INST_LOC, draw.info.start_instance, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(&rctx->vgt, R_008958_VGT_PRIMITIVE_TYPE, prim, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(&rctx->vgt, R_028238_CB_TARGET_MASK, rctx->cb_target_mask & mask, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(&rctx->vgt, R_028400_VGT_MAX_VTX_INDX, draw.info.max_index, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(&rctx->vgt, R_028404_VGT_MIN_VTX_INDX, draw.info.min_index, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(&rctx->vgt, R_028408_VGT_INDX_OFFSET, draw.info.index_bias, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(&rctx->vgt, R_02840C_VGT_MULTI_PRIM_IB_RESET_INDX, draw.info.restart_index, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(&rctx->vgt, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN, draw.info.primitive_restart, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(&rctx->vgt, R_03CFF0_SQ_VTX_BASE_VTX_LOC, 0, 0xFFFFFFFF, NULL, 0); + r600_pipe_state_add_reg(&rctx->vgt, R_03CFF4_SQ_VTX_START_INST_LOC, draw.info.start_instance, 0xFFFFFFFF, NULL, 0); r600_pipe_state_add_reg(&rctx->vgt, R_028814_PA_SU_SC_MODE_CNTL, 0, - S_028814_PROVOKING_VTX_LAST(1), NULL); + S_028814_PROVOKING_VTX_LAST(1), NULL, 0); } @@ -634,6 +635,8 @@ void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info) r600_pipe_state_mod_reg(&rctx->vgt, draw.info.max_index); r600_pipe_state_mod_reg(&rctx->vgt, draw.info.min_index); r600_pipe_state_mod_reg(&rctx->vgt, draw.info.index_bias); + r600_pipe_state_mod_reg(&rctx->vgt, draw.info.restart_index); + r600_pipe_state_mod_reg(&rctx->vgt, draw.info.primitive_restart); r600_pipe_state_mod_reg(&rctx->vgt, 0); r600_pipe_state_mod_reg(&rctx->vgt, draw.info.start_instance); if (draw.info.mode == PIPE_PRIM_QUADS || draw.info.mode == PIPE_PRIM_QUAD_STRIP || draw.info.mode == PIPE_PRIM_POLYGON) { @@ -676,11 +679,14 @@ void _r600_pipe_state_add_reg(struct r600_context *ctx, struct r600_pipe_state *state, u32 offset, u32 value, u32 mask, u32 range_id, u32 block_id, - struct r600_bo *bo) + struct r600_bo *bo, + enum radeon_bo_usage usage) { struct r600_range *range; struct r600_block *block; + if (bo) assert(usage); + range = &ctx->range[range_id]; block = range->blocks[block_id]; state->regs[state->nregs].block = block; @@ -689,6 +695,7 @@ void _r600_pipe_state_add_reg(struct r600_context *ctx, state->regs[state->nregs].value = value; state->regs[state->nregs].mask = mask; state->regs[state->nregs].bo = bo; + state->regs[state->nregs].bo_usage = usage; state->nregs++; assert(state->nregs < R600_BLOCK_MAX_REG); @@ -696,13 +703,17 @@ void _r600_pipe_state_add_reg(struct r600_context *ctx, void r600_pipe_state_add_reg_noblock(struct r600_pipe_state *state, u32 offset, u32 value, u32 mask, - struct r600_bo *bo) + struct r600_bo *bo, + enum radeon_bo_usage usage) { + if (bo) assert(usage); + state->regs[state->nregs].id = offset; state->regs[state->nregs].block = NULL; state->regs[state->nregs].value = value; state->regs[state->nregs].mask = mask; state->regs[state->nregs].bo = bo; + state->regs[state->nregs].bo_usage = usage; state->nregs++; assert(state->nregs < R600_BLOCK_MAX_REG); diff --git a/src/gallium/drivers/r600/r600_texture.c b/src/gallium/drivers/r600/r600_texture.c index e9e8b277243..7c1bd9d8ec6 100644 --- a/src/gallium/drivers/r600/r600_texture.c +++ b/src/gallium/drivers/r600/r600_texture.c @@ -31,7 +31,6 @@ #include <util/u_math.h> #include <util/u_inlines.h> #include <util/u_memory.h> -#include "state_tracker/drm_driver.h" #include "pipebuffer/pb_buffer.h" #include "r600_pipe.h" #include "r600_resource.h" @@ -67,7 +66,7 @@ static void r600_copy_from_staging_texture(struct pipe_context *ctx, struct r600 rtransfer->staging_texture, 0, &sbox); - ctx->flush(ctx, NULL); + r600_flush(ctx, NULL, RADEON_FLUSH_ASYNC); } unsigned r600_texture_get_offset(struct r600_resource_texture *rtex, @@ -174,15 +173,15 @@ static unsigned r600_texture_get_nblocksx(struct pipe_screen *screen, { struct pipe_resource *ptex = &rtex->resource.b.b.b; unsigned nblocksx, block_align, width; - unsigned blocksize = util_format_get_blocksize(ptex->format); + unsigned blocksize = util_format_get_blocksize(rtex->real_format); if (rtex->pitch_override) return rtex->pitch_override / blocksize; width = mip_minify(ptex->width0, level); - nblocksx = util_format_get_nblocksx(ptex->format, width); + nblocksx = util_format_get_nblocksx(rtex->real_format, width); - block_align = r600_get_block_alignment(screen, ptex->format, + block_align = r600_get_block_alignment(screen, rtex->real_format, rtex->array_mode[level]); nblocksx = align(nblocksx, block_align); return nblocksx; @@ -196,9 +195,19 @@ static unsigned r600_texture_get_nblocksy(struct pipe_screen *screen, unsigned height, tile_height; height = mip_minify(ptex->height0, level); - height = util_format_get_nblocksy(ptex->format, height); + height = util_format_get_nblocksy(rtex->real_format, height); tile_height = r600_get_height_alignment(screen, rtex->array_mode[level]); + + /* XXX Hack around an alignment issue. Less tests fail with this. + * + * The thing is depth-stencil buffers should be tiled, i.e. + * the alignment should be >=8. If I make them tiled, stencil starts + * working because it no longer overlaps with the depth buffer + * in memory, but texturing like drawpix-stencil breaks. */ + if (util_format_is_depth_or_stencil(rtex->real_format) && tile_height < 8) + tile_height = 8; + height = align(height, tile_height); return height; } @@ -221,7 +230,7 @@ static void r600_texture_set_array_mode(struct pipe_screen *screen, unsigned w, h, tile_height, tile_width; tile_height = r600_get_height_alignment(screen, array_mode); - tile_width = r600_get_block_alignment(screen, ptex->format, array_mode); + tile_width = r600_get_block_alignment(screen, rtex->real_format, array_mode); w = mip_minify(ptex->width0, level); h = mip_minify(ptex->height0, level); @@ -239,14 +248,14 @@ static void r600_setup_miptree(struct pipe_screen *screen, unsigned array_mode) { struct pipe_resource *ptex = &rtex->resource.b.b.b; - struct radeon *radeon = (struct radeon *)screen->winsys; + struct radeon *radeon = ((struct r600_screen*)screen)->radeon; enum chip_class chipc = r600_get_family_class(radeon); unsigned size, layer_size, i, offset; - unsigned nblocksx, nblocksy, extra_size = 0; + unsigned nblocksx, nblocksy; for (i = 0, offset = 0; i <= ptex->last_level; i++) { - unsigned blocksize = util_format_get_blocksize(ptex->format); - unsigned base_align = r600_get_base_alignment(screen, ptex->format, array_mode); + unsigned blocksize = util_format_get_blocksize(rtex->real_format); + unsigned base_align = r600_get_base_alignment(screen, rtex->real_format, array_mode); r600_texture_set_array_mode(screen, rtex, i, array_mode); @@ -265,10 +274,6 @@ static void r600_setup_miptree(struct pipe_screen *screen, else size = layer_size * ptex->array_size; - /* evergreen stores depth and stencil separately */ - if ((chipc >= EVERGREEN) && util_format_is_depth_or_stencil(ptex->format)) - extra_size = align(extra_size + (nblocksx * nblocksy * 1), base_align); - /* align base image and start of miptree */ if ((i == 0) || (i == 1)) offset = align(offset, base_align); @@ -279,7 +284,7 @@ static void r600_setup_miptree(struct pipe_screen *screen, offset += size; } - rtex->size = offset + extra_size; + rtex->size = offset; } /* Figure out whether u_blitter will fallback to a transfer operation. @@ -329,7 +334,7 @@ static boolean r600_texture_get_handle(struct pipe_screen* screen, { struct r600_resource_texture *rtex = (struct r600_resource_texture*)ptex; struct r600_resource *resource = &rtex->resource; - struct radeon *radeon = (struct radeon *)screen->winsys; + struct radeon *radeon = ((struct r600_screen*)screen)->radeon; return r600_bo_get_winsys_handle(radeon, resource->bo, rtex->pitch_in_bytes[0], whandle); @@ -340,13 +345,12 @@ static void r600_texture_destroy(struct pipe_screen *screen, { struct r600_resource_texture *rtex = (struct r600_resource_texture*)ptex; struct r600_resource *resource = &rtex->resource; - struct radeon *radeon = (struct radeon *)screen->winsys; if (rtex->flushed_depth_texture) pipe_resource_reference((struct pipe_resource **)&rtex->flushed_depth_texture, NULL); if (resource->bo) { - r600_bo_reference(radeon, &resource->bo, NULL); + r600_bo_reference(&resource->bo, NULL); } FREE(rtex); } @@ -369,11 +373,12 @@ r600_texture_create_object(struct pipe_screen *screen, unsigned array_mode, unsigned pitch_in_bytes_override, unsigned max_buffer_size, - struct r600_bo *bo) + struct r600_bo *bo, + boolean alloc_bo) { struct r600_resource_texture *rtex; struct r600_resource *resource; - struct radeon *radeon = (struct radeon *)screen->winsys; + struct radeon *radeon = ((struct r600_screen*)screen)->radeon; rtex = CALLOC_STRUCT(r600_resource_texture); if (rtex == NULL) @@ -386,59 +391,112 @@ r600_texture_create_object(struct pipe_screen *screen, resource->b.b.b.screen = screen; resource->bo = bo; rtex->pitch_override = pitch_in_bytes_override; + rtex->real_format = base->format; + + /* We must split depth and stencil into two separate buffers on Evergreen. */ + if (!(base->flags & R600_RESOURCE_FLAG_TRANSFER) && + r600_get_family_class(((struct r600_screen*)screen)->radeon) >= EVERGREEN && + util_format_is_depth_and_stencil(base->format)) { + struct pipe_resource stencil; + unsigned stencil_pitch_override = 0; + + switch (base->format) { + case PIPE_FORMAT_Z24_UNORM_S8_USCALED: + rtex->real_format = PIPE_FORMAT_Z24X8_UNORM; + break; + case PIPE_FORMAT_S8_USCALED_Z24_UNORM: + rtex->real_format = PIPE_FORMAT_X8Z24_UNORM; + break; + case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED: + rtex->real_format = PIPE_FORMAT_Z32_FLOAT; + break; + default: + assert(0); + FREE(rtex); + return NULL; + } + + /* Divide the pitch in bytes by 4 for stencil, because it has a smaller pixel size. */ + if (pitch_in_bytes_override) { + assert(base->format == PIPE_FORMAT_Z24_UNORM_S8_USCALED || + base->format == PIPE_FORMAT_S8_USCALED_Z24_UNORM); + stencil_pitch_override = pitch_in_bytes_override / 4; + } + + /* Allocate the stencil buffer. */ + stencil = *base; + stencil.format = PIPE_FORMAT_S8_USCALED; + rtex->stencil = r600_texture_create_object(screen, &stencil, array_mode, + stencil_pitch_override, + max_buffer_size, NULL, FALSE); + if (!rtex->stencil) { + FREE(rtex); + return NULL; + } + /* Proceed in creating the depth buffer. */ + } + /* only mark depth textures the HW can hit as depth textures */ - if (util_format_is_depth_or_stencil(base->format) && permit_hardware_blit(screen, base)) + if (util_format_is_depth_or_stencil(rtex->real_format) && permit_hardware_blit(screen, base)) rtex->depth = 1; r600_setup_miptree(screen, rtex, array_mode); + /* If we initialized separate stencil for Evergreen. place it after depth. */ + if (rtex->stencil) { + unsigned stencil_align, stencil_offset; + + stencil_align = r600_get_base_alignment(screen, rtex->stencil->real_format, array_mode); + stencil_offset = align(rtex->size, stencil_align); + + for (unsigned i = 0; i <= rtex->stencil->resource.b.b.b.last_level; i++) + rtex->stencil->offset[i] += stencil_offset; + + rtex->size = stencil_offset + rtex->stencil->size; + } + resource->size = rtex->size; - if (!resource->bo) { + /* Now create the backing buffer. */ + if (!resource->bo && alloc_bo) { struct pipe_resource *ptex = &rtex->resource.b.b.b; - int base_align = r600_get_base_alignment(screen, ptex->format, array_mode); + unsigned base_align = r600_get_base_alignment(screen, ptex->format, array_mode); resource->bo = r600_bo(radeon, rtex->size, base_align, base->bind, base->usage); if (!resource->bo) { + pipe_resource_reference((struct pipe_resource**)&rtex->stencil, NULL); FREE(rtex); return NULL; } } + + if (rtex->stencil) + rtex->stencil->resource.bo = rtex->resource.bo; return rtex; } +DEBUG_GET_ONCE_BOOL_OPTION(tiling_enabled, "R600_TILING", FALSE); + struct pipe_resource *r600_texture_create(struct pipe_screen *screen, const struct pipe_resource *templ) { + struct radeon *radeon = ((struct r600_screen*)screen)->radeon; unsigned array_mode = 0; - static int force_tiling = -1; - /* Would like some magic "get_bool_option_once" routine. - */ - if (force_tiling == -1) { -#if 0 - /* reenable when 2D tiling is fixed better */ - struct r600_screen *rscreen = (struct r600_screen *)screen; - if (r600_get_minor_version(rscreen->radeon) >= 9) - force_tiling = debug_get_bool_option("R600_TILING", TRUE); -#endif - force_tiling = debug_get_bool_option("R600_TILING", FALSE); - } - - if (force_tiling && permit_hardware_blit(screen, templ)) { - if (!(templ->flags & R600_RESOURCE_FLAG_TRANSFER) && - !(templ->bind & PIPE_BIND_SCANOUT)) { + if (!(templ->flags & R600_RESOURCE_FLAG_TRANSFER) && + !(templ->bind & PIPE_BIND_SCANOUT)) { + if (util_format_is_compressed(templ->format)) { + array_mode = V_038000_ARRAY_1D_TILED_THIN1; + } + else if (debug_get_option_tiling_enabled() && + r600_get_minor_version(radeon) >= 9 && + permit_hardware_blit(screen, templ)) { array_mode = V_038000_ARRAY_2D_TILED_THIN1; } } - if (!(templ->flags & R600_RESOURCE_FLAG_TRANSFER) && - util_format_is_compressed(templ->format)) - array_mode = V_038000_ARRAY_1D_TILED_THIN1; - return (struct pipe_resource *)r600_texture_create_object(screen, templ, array_mode, - 0, 0, NULL); - + 0, 0, NULL, TRUE); } static struct pipe_surface *r600_create_surface(struct pipe_context *pipe, @@ -483,8 +541,9 @@ struct pipe_resource *r600_texture_from_handle(struct pipe_screen *screen, const struct pipe_resource *templ, struct winsys_handle *whandle) { - struct radeon *rw = (struct radeon*)screen->winsys; + struct radeon *rw = ((struct r600_screen*)screen)->radeon; struct r600_bo *bo = NULL; + unsigned stride = 0; unsigned array_mode = 0; /* Support only 2D textures without mipmaps */ @@ -492,15 +551,13 @@ struct pipe_resource *r600_texture_from_handle(struct pipe_screen *screen, templ->depth0 != 1 || templ->last_level != 0) return NULL; - bo = r600_bo_handle(rw, whandle->handle, &array_mode); + bo = r600_bo_handle(rw, whandle, &stride, &array_mode); if (bo == NULL) { return NULL; } return (struct pipe_resource *)r600_texture_create_object(screen, templ, array_mode, - whandle->stride, - 0, - bo); + stride, 0, bo, FALSE); } int r600_texture_depth_flush(struct pipe_context *ctx, @@ -590,6 +647,9 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx, (texture->flags & R600_RESOURCE_FLAG_TRANSFER)) use_staging_texture = FALSE; + if (use_staging_texture && (usage & PIPE_TRANSFER_MAP_DIRECTLY)) + return NULL; + trans = CALLOC_STRUCT(r600_transfer); if (trans == NULL) return NULL; @@ -648,7 +708,7 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx, if (usage & PIPE_TRANSFER_READ) { r600_copy_to_staging_texture(ctx, trans); /* Always referenced in the blit. */ - ctx->flush(ctx, NULL); + r600_flush(ctx, NULL, 0); } return &trans->transfer; } @@ -684,10 +744,11 @@ void r600_texture_transfer_destroy(struct pipe_context *ctx, void* r600_texture_transfer_map(struct pipe_context *ctx, struct pipe_transfer* transfer) { + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_transfer *rtransfer = (struct r600_transfer*)transfer; struct r600_bo *bo; enum pipe_format format = transfer->resource->format; - struct radeon *radeon = (struct radeon *)ctx->screen->winsys; + struct radeon *radeon = rctx->screen->radeon; unsigned offset = 0; char *map; @@ -706,7 +767,7 @@ void* r600_texture_transfer_map(struct pipe_context *ctx, transfer->box.x / util_format_get_blockwidth(format) * util_format_get_blocksize(format); } - if (!(map = r600_bo_map(radeon, bo, transfer->usage, ctx))) { + if (!(map = r600_bo_map(radeon, bo, rctx->ctx.cs, transfer->usage))) { return NULL; } @@ -717,7 +778,7 @@ void r600_texture_transfer_unmap(struct pipe_context *ctx, struct pipe_transfer* transfer) { struct r600_transfer *rtransfer = (struct r600_transfer*)transfer; - struct radeon *radeon = (struct radeon *)ctx->screen->winsys; + struct radeon *radeon = ((struct r600_screen*)ctx->screen)->radeon; struct r600_bo *bo; if (rtransfer->staging_texture) { @@ -754,11 +815,7 @@ static unsigned r600_get_swizzle_combined(const unsigned char *swizzle_format, }; if (swizzle_view) { - /* Combine two sets of swizzles. */ - for (i = 0; i < 4; i++) { - swizzle[i] = swizzle_view[i] <= UTIL_FORMAT_SWIZZLE_W ? - swizzle_format[swizzle_view[i]] : swizzle_view[i]; - } + util_format_compose_swizzles(swizzle_format, swizzle_view, swizzle); } else { memcpy(swizzle, swizzle_format, 4); } diff --git a/src/gallium/drivers/r600/r600d.h b/src/gallium/drivers/r600/r600d.h index f6eec24cc05..de458cf398a 100644 --- a/src/gallium/drivers/r600/r600d.h +++ b/src/gallium/drivers/r600/r600d.h @@ -28,6 +28,32 @@ #define R600_TEXEL_PITCH_ALIGNMENT_MASK 0x7 +/* evergreen values */ +#define EG_RESOURCE_OFFSET 0x00030000 +#define EG_RESOURCE_END 0x00034000 +#define EG_LOOP_CONST_OFFSET 0x0003A200 +#define EG_LOOP_CONST_END 0x0003A26C +#define EG_BOOL_CONST_OFFSET 0x0003A500 +#define EG_BOOL_CONST_END 0x0003A506 + +#define R600_CONFIG_REG_OFFSET 0X00008000 +#define R600_CONFIG_REG_END 0X0000AC00 +#define R600_CONTEXT_REG_OFFSET 0X00028000 +#define R600_CONTEXT_REG_END 0X00029000 +#define R600_ALU_CONST_OFFSET 0X00030000 +#define R600_ALU_CONST_END 0X00032000 +#define R600_RESOURCE_OFFSET 0X00038000 +#define R600_RESOURCE_END 0X0003C000 +#define R600_SAMPLER_OFFSET 0X0003C000 +#define R600_SAMPLER_END 0X0003CFF0 +#define R600_CTL_CONST_OFFSET 0X0003CFF0 +#define R600_CTL_CONST_END 0X0003E200 +#define R600_LOOP_CONST_OFFSET 0X0003E200 +#define R600_LOOP_CONST_END 0X0003E380 +#define R600_BOOL_CONST_OFFSET 0X0003E380 +#define R600_BOOL_CONST_END 0X00040000 + + #define PKT3_NOP 0x10 #define PKT3_INDIRECT_BUFFER_END 0x17 #define PKT3_SET_PREDICATION 0x20 @@ -66,11 +92,38 @@ #define PKT3_SET_SAMPLER 0x6E #define PKT3_SET_CTL_CONST 0x6F #define PKT3_SURFACE_BASE_UPDATE 0x73 +#define SURFACE_BASE_UPDATE_DEPTH (1 << 0) +#define SURFACE_BASE_UPDATE_COLOR(x) (2 << (x)) +#define SURFACE_BASE_UPDATE_STRMOUT(x) (0x200 << (x)) + +#define EVENT_TYPE_PS_PARTIAL_FLUSH 0x10 +#define EVENT_TYPE_CACHE_FLUSH_AND_INV_TS_EVENT 0x14 +#define EVENT_TYPE_ZPASS_DONE 0x15 +#define EVENT_TYPE_CACHE_FLUSH_AND_INV_EVENT 0x16 +#define EVENT_TYPE(x) ((x) << 0) +#define EVENT_INDEX(x) ((x) << 8) + /* 0 - any non-TS event + * 1 - ZPASS_DONE + * 2 - SAMPLE_PIPELINESTAT + * 3 - SAMPLE_STREAMOUTSTAT* + * 4 - *S_PARTIAL_FLUSH + * 5 - TS events + */ #define PREDICATION_OP_CLEAR 0x0 #define PREDICATION_OP_ZPASS 0x1 #define PREDICATION_OP_PRIMCOUNT 0x2 +#define PRED_OP(x) ((x) << 16) + +#define PREDICATION_CONTINUE (1 << 31) + +#define PREDICATION_HINT_WAIT (0 << 12) +#define PREDICATION_HINT_NOWAIT_DRAW (1 << 12) + +#define PREDICATION_DRAW_NOT_VISIBLE (0 << 8) +#define PREDICATION_DRAW_VISIBLE (1 << 8) + #define PKT_TYPE_S(x) (((x) & 0x3) << 30) #define PKT_TYPE_G(x) (((x) >> 30) & 0x3) #define PKT_TYPE_C 0x3FFFFFFF @@ -83,8 +136,9 @@ #define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8) #define PKT3_IT_OPCODE_G(x) (((x) >> 8) & 0xFF) #define PKT3_IT_OPCODE_C 0xFFFF00FF +#define PKT3_PRED_S(x) (((x) >> 0) & 0x1) #define PKT0(index, count) (PKT_TYPE_S(0) | PKT0_BASE_INDEX_S(index) | PKT_COUNT_S(count)) -#define PKT3(op, count) (PKT_TYPE_S(3) | PKT3_IT_OPCODE_S(op) | PKT_COUNT_S(count)) +#define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT3_IT_OPCODE_S(op) | PKT_COUNT_S(count) | PKT3_PRED_S(predicate)) /* Registers */ #define R_008C00_SQ_CONFIG 0x00008C00 diff --git a/src/gallium/drivers/r600/r700_asm.c b/src/gallium/drivers/r600/r700_asm.c index b3c7d1494fc..74efe226530 100644 --- a/src/gallium/drivers/r600/r700_asm.c +++ b/src/gallium/drivers/r600/r700_asm.c @@ -26,7 +26,7 @@ #include "r600_asm.h" #include "r700_sq.h" -void r700_bc_cf_vtx_build(uint32_t *bytecode, const struct r600_bc_cf *cf) +void r700_bytecode_cf_vtx_build(uint32_t *bytecode, const struct r600_bytecode_cf *cf) { unsigned count = (cf->ndw / 4) - 1; *bytecode++ = S_SQ_CF_WORD0_ADDR(cf->addr >> 1); @@ -36,7 +36,7 @@ void r700_bc_cf_vtx_build(uint32_t *bytecode, const struct r600_bc_cf *cf) S_SQ_CF_WORD1_COUNT_3(count >> 3); } -int r700_bc_alu_build(struct r600_bc *bc, struct r600_bc_alu *alu, unsigned id) +int r700_bytecode_alu_build(struct r600_bytecode *bc, struct r600_bytecode_alu *alu, unsigned id) { bc->bytecode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) | S_SQ_ALU_WORD0_SRC0_REL(alu->src[0].rel) | |