Merge branch 'gallium-0.2-radeon' into gallium-0.2

1 files changed, 328 insertions, 0 deletions

diff --git a/src/gallium/drivers/r300/r300_surface.c b/src/gallium/drivers/r300/r300_surface.c
new file mode 100644
index 00000000000..1ed4a4e3bcf
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_surface.c
@@ -0,0 +1,328 @@
+/*
+ * Copyright 2008 Corbin Simpson <[email protected]>
+ *
+ * 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
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, 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 (including the next
+ * paragraph) 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS 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. */
+
+#include "r300_surface.h"
+
+/* Provides pipe_context's "surface_fill". Commonly used for clearing
+ * buffers. */
+static void r300_surface_fill(struct pipe_context* pipe,
+                              struct pipe_surface* dest,
+                              unsigned x, unsigned y,
+                              unsigned w, unsigned h,
+                              unsigned color)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    CS_LOCALS(r300);
+    struct r300_capabilities* caps = ((struct r300_screen*)pipe->screen)->caps;
+    int i;
+    float r, g, b, a;
+    r = (float)((color >> 16) & 0xff) / 255.0f;
+    g = (float)((color >>  8) & 0xff) / 255.0f;
+    b = (float)((color >>  0) & 0xff) / 255.0f;
+    debug_printf("r300: Filling surface %p at (%d,%d),"
+        " dimensions %dx%d (stride %d), color 0x%x\n",
+        dest, x, y, w, h, dest->stride, color);
+
+    /* Fallback? */
+    if (0) {
+        debug_printf("r300: Falling back on surface clear...");
+        void* map = pipe->screen->surface_map(pipe->screen, dest,
+            PIPE_BUFFER_USAGE_CPU_WRITE);
+        pipe_fill_rect(map, &dest->block, &dest->stride, x, y, w, h, color);
+        pipe->screen->surface_unmap(pipe->screen, dest);
+        return;
+    }
+
+BEGIN_CS((caps->is_r500) ? 309 : 280);
+R300_PACIFY;
+OUT_CS_REG(R300_TX_INVALTAGS, 0x0);
+R300_PACIFY;
+/* Flush PVS. */
+OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
+
+OUT_CS_REG(R300_SE_VTE_CNTL, R300_VPORT_X_SCALE_ENA |
+    R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA |
+    R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA |
+    R300_VPORT_Z_OFFSET_ENA | R300_VTX_W0_FMT);
+/* Vertex size. */
+OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8);
+/* Max and min vertex index clamp. */
+OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, 0xFFFFFF);
+OUT_CS_REG(R300_VAP_VF_MIN_VTX_INDX, 0x0);
+/* XXX endian */
+OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP);
+OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0, 0x0);
+/* XXX magic number not in r300_reg */
+OUT_CS_REG(R300_VAP_PSC_SGN_NORM_CNTL, 0xAAAAAAAA);
+OUT_CS_REG(R300_VAP_CLIP_CNTL, 0x0);
+OUT_CS_REG_SEQ(R300_VAP_GB_VERT_CLIP_ADJ, 4);
+OUT_CS_32F(1.0);
+OUT_CS_32F(1.0);
+OUT_CS_32F(1.0);
+OUT_CS_32F(1.0);
+/* XXX is this too long? */
+OUT_CS_REG(VAP_PVS_VTX_TIMEOUT_REG, 0xFFFF);
+OUT_CS_REG(R300_GB_ENABLE, R300_GB_POINT_STUFF_ENABLE |
+    R300_GB_LINE_STUFF_ENABLE | R300_GB_TRIANGLE_STUFF_ENABLE);
+/* XXX more magic numbers */
+OUT_CS_REG(R300_GB_MSPOS0, 0x66666666);
+OUT_CS_REG(R300_GB_MSPOS1, 0x66666666);
+/* XXX why doesn't classic Mesa write the number of pipes, too? */
+OUT_CS_REG(R300_GB_TILE_CONFIG, R300_GB_TILE_ENABLE | R300_GB_TILE_SIZE_16);
+OUT_CS_REG(R300_GB_SELECT, R300_GB_FOG_SELECT_1_1_W);
+OUT_CS_REG(R300_GB_AA_CONFIG, 0x0);
+/* XXX point tex stuffing */
+OUT_CS_REG_SEQ(R300_GA_POINT_S0, 1);
+OUT_CS_32F(0.0);
+OUT_CS_REG_SEQ(R300_GA_POINT_S1, 1);
+OUT_CS_32F(1.0);
+OUT_CS_REG(R300_GA_TRIANGLE_STIPPLE, 0x5 |
+    (0x5 << R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_SHIFT));
+/* XXX should this be related to the actual point size? */
+OUT_CS_REG(R300_GA_POINT_MINMAX, 0x6 |
+    (0x1800 << R300_GA_POINT_MINMAX_MAX_SHIFT));
+/* XXX this big chunk should be refactored into rs_state */
+OUT_CS_REG(R300_GA_LINE_CNTL, 0x00030006);
+OUT_CS_REG(R300_GA_LINE_STIPPLE_CONFIG, 0x3BAAAAAB);
+OUT_CS_REG(R300_GA_LINE_STIPPLE_VALUE, 0x00000000);
+OUT_CS_REG(R300_GA_LINE_S0, 0x00000000);
+OUT_CS_REG(R300_GA_LINE_S1, 0x3F800000);
+OUT_CS_REG(R300_GA_ENHANCE, 0x00000002);
+OUT_CS_REG(R300_GA_COLOR_CONTROL, 0x0003AAAA);
+OUT_CS_REG(R300_GA_SOLID_RG, 0x00000000);
+OUT_CS_REG(R300_GA_SOLID_BA, 0x00000000);
+OUT_CS_REG(R300_GA_POLY_MODE, 0x00000000);
+OUT_CS_REG(R300_GA_ROUND_MODE, 0x00000001);
+OUT_CS_REG(R300_GA_OFFSET, 0x00000000);
+OUT_CS_REG(R300_GA_FOG_SCALE, 0x3DBF1412);
+OUT_CS_REG(R300_GA_FOG_OFFSET, 0x00000000);
+OUT_CS_REG(R300_SU_TEX_WRAP, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_FRONT_SCALE, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_FRONT_OFFSET, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_BACK_SCALE, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_BACK_OFFSET, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_ENABLE, 0x00000000);
+OUT_CS_REG(R300_SU_CULL_MODE, 0x00000000);
+OUT_CS_REG(R300_SU_DEPTH_SCALE, 0x4B7FFFFF);
+OUT_CS_REG(R300_SU_DEPTH_OFFSET, 0x00000000);
+OUT_CS_REG(R300_SC_HYPERZ, 0x0000001C);
+OUT_CS_REG(R300_SC_EDGERULE, 0x2DA49525);
+OUT_CS_REG(R300_SC_SCREENDOOR, 0x00FFFFFF);
+OUT_CS_REG(R300_FG_FOG_BLEND, 0x00000002);
+OUT_CS_REG(R300_FG_FOG_COLOR_R, 0x00000000);
+OUT_CS_REG(R300_FG_FOG_COLOR_G, 0x00000000);
+OUT_CS_REG(R300_FG_FOG_COLOR_B, 0x00000000);
+OUT_CS_REG(R300_FG_DEPTH_SRC, 0x00000000);
+OUT_CS_REG(R300_FG_DEPTH_SRC, 0x00000000);
+OUT_CS_REG(R300_RB3D_CCTL, 0x00000000);
+OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F);
+
+/* XXX: Oh the wonderful unknown */
+OUT_CS_REG_SEQ(0x4E54, 8);
+for (i = 0; i < 8; i++)
+    OUT_CS(0x00000000);
+OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000);
+OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000);
+OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD, 0xFFFFFFFF);
+OUT_CS_REG(R300_ZB_FORMAT, 0x00000002);
+OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, 0x00000003);
+OUT_CS_REG(R300_ZB_BW_CNTL, 0x00000000);
+OUT_CS_REG(R300_ZB_DEPTHCLEARVALUE, 0x00000000);
+OUT_CS_REG(0x4F30, 0x00000000);
+OUT_CS_REG(0x4F34, 0x00000000);
+OUT_CS_REG(R300_ZB_HIZ_OFFSET, 0x00000000);
+OUT_CS_REG(R300_ZB_HIZ_PITCH, 0x00000000);
+R300_PACIFY;
+OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0, 0x21030003);
+OUT_CS_REG(R300_FG_FOG_BLEND, 0x00000000);
+OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_EXT_0, 0xF688F688);
+OUT_CS_REG(R300_VAP_VTX_STATE_CNTL, 0x1);
+OUT_CS_REG(R300_VAP_VSM_VTX_ASSM, 0x405);
+OUT_CS_REG(R300_SE_VTE_CNTL, 0x0000043F);
+OUT_CS_REG(R300_VAP_VTX_SIZE, 0x00000008);
+OUT_CS_REG(R300_VAP_PSC_SGN_NORM_CNTL, 0xAAAAAAAA);
+OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_0, 0x00000003);
+OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_1, 0x00000000);
+OUT_CS_REG(R300_TX_ENABLE, 0x0);
+/* XXX viewport setup */
+OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
+OUT_CS_32F(1.0);
+OUT_CS_32F((float)x);
+OUT_CS_32F(1.0);
+OUT_CS_32F((float)y);
+OUT_CS_32F(1.0);
+OUT_CS_32F(0.0);
+
+OUT_CS_REG(R300_VAP_CLIP_CNTL, 0x0001C000);
+OUT_CS_REG(R300_GA_POINT_SIZE, ((h * 6) & R300_POINTSIZE_Y_MASK) |
+    ((w * 6) << R300_POINTSIZE_X_SHIFT));
+
+/* XXX RS block and fp setup */
+if (caps->is_r500) {
+    OUT_CS_REG_SEQ(R500_RS_IP_0, 8);
+    for (i = 0; i < 8; i++) {
+        /* I like the operator macros more than the shift macros... */
+        OUT_CS((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) |
+            (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
+            (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
+            (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT));
+    }
+    /* XXX */
+    OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
+    OUT_CS((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
+    OUT_CS(0x0);
+    OUT_CS_REG(R500_RS_INST_0, R500_RS_INST_COL_CN_WRITE);
+
+    OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
+    OUT_CS_REG(R500_US_PIXSIZE, 0x00000000);
+    OUT_CS_REG(R500_US_CODE_ADDR, R500_US_CODE_START_ADDR(0) |
+        R500_US_CODE_END_ADDR(1));
+    OUT_CS_REG(R500_US_CODE_RANGE, R500_US_CODE_RANGE_ADDR(0) |
+        R500_US_CODE_RANGE_SIZE(1));
+    OUT_CS_REG(R500_US_CODE_OFFSET, R500_US_CODE_OFFSET_ADDR(0));
+    R300_PACIFY;
+    OUT_CS_REG(R500_GA_US_VECTOR_INDEX,
+        0 | R500_GA_US_VECTOR_INDEX_TYPE_INSTR);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_INST_TYPE_OUT | R500_INST_TEX_SEM_WAIT | R500_INST_LAST |
+        R500_INST_RGB_OMASK_R | R500_INST_RGB_OMASK_G | R500_INST_RGB_OMASK_B |
+        R500_INST_ALPHA_OMASK | R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST |
+        R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST |
+        R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R |
+        R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B |
+        R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R |
+        R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 |
+        R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 |
+        R500_ALU_RGBA_A_SWIZ_0);
+} else {
+    OUT_CS_REG_SEQ(R300_RS_IP_0, 8);
+    for (i = 0; i < 8; i++) {
+        OUT_CS(R300_RS_SEL_T(R300_RS_SEL_K0) |
+            R300_RS_SEL_R(R300_RS_SEL_K0) | R300_RS_SEL_Q(R300_RS_SEL_K1));
+    }
+    /* XXX */
+    OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
+    OUT_CS((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
+    OUT_CS(1);
+    OUT_CS_REG(R300_RS_INST_0, R300_RS_INST_COL_CN_WRITE);
+
+    /* XXX magic numbers */
+    OUT_CS_REG(R300_US_CONFIG, 0);
+    OUT_CS_REG(R300_US_PIXSIZE, 2);
+    OUT_CS_REG(R300_US_CODE_OFFSET, 0x0);
+    OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0);
+    OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0);
+    OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0);
+    OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000);
+    OUT_CS_REG(R300_US_ALU_RGB_INST_0, 0x50A80);
+    OUT_CS_REG(R300_US_ALU_RGB_ADDR_0, 0x1C000000);
+    OUT_CS_REG(R300_US_ALU_ALPHA_INST_0, 0x40889);
+    OUT_CS_REG(R300_US_ALU_ALPHA_ADDR_0, 0x1000000);
+    OUT_CS_REG_SEQ(R300_US_OUT_FMT_0, 4);
+    OUT_CS(R300_C0_SEL_B | R300_C1_SEL_G | R300_C2_SEL_R | R300_C3_SEL_A);
+    OUT_CS(R300_US_OUT_FMT_UNUSED);
+    OUT_CS(R300_US_OUT_FMT_UNUSED);
+    OUT_CS(R300_US_OUT_FMT_UNUSED);
+    OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W0);
+}
+/* XXX these magic numbers should be explained when
+ * this becomes a cached state object */
+OUT_CS_REG(R300_VAP_CNTL, 0xA | (0x5 << R300_PVS_NUM_CNTLRS_SHIFT) |
+    (caps->num_vert_fpus << R300_PVS_NUM_FPUS_SHIFT));
+OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, 0x00100000);
+OUT_CS_REG(R300_VAP_PVS_CONST_CNTL, 0x00000000);
+OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, 0x00000001);
+R300_PACIFY;
+/* XXX translate these back into normal instructions */
+OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x1);
+OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0x0);
+OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xF00203);
+OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xD10001);
+OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x1248001);
+OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x0);
+OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xF02203);
+OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xD10021);
+OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x1248021);
+OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x0);
+R300_PACIFY;
+END_CS;
+
+r300_emit_blend_state(r300, &blend_clear_state);
+r300_emit_blend_color_state(r300, &blend_color_clear_state);
+r300_emit_dsa_state(r300, &dsa_clear_state);
+
+BEGIN_CS(36);
+R300_PACIFY;
+/* Flush colorbuffer and blend caches. */
+OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT,
+    R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D |
+    R300_RB3D_DSTCACHE_CTLSTAT_DC_FINISH_SIGNAL);
+OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT,
+    R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
+    R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
+
+OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0, 1);
+OUT_CS_RELOC(dest->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+/* XXX this should not be so rigid and it still doesn't work right */
+OUT_CS_REG(R300_RB3D_COLORPITCH0, (dest->stride >> 2) | R300_COLOR_FORMAT_ARGB8888);
+OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F);
+/* XXX Packet3 */
+OUT_CS(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8));
+OUT_CS(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING |
+(1 << R300_PRIM_NUM_VERTICES_SHIFT));
+OUT_CS_32F(w / 2.0);
+OUT_CS_32F(h / 2.0);
+/* XXX this should be the depth value to clear to */
+OUT_CS_32F(1.0);
+OUT_CS_32F(1.0);
+OUT_CS_32F(r);
+OUT_CS_32F(g);
+OUT_CS_32F(b);
+OUT_CS_32F(1.0);
+
+/* XXX figure out why this is 0xA and not 0x2 */
+OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA);
+/* XXX OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT,
+    R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
+    R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); */
+R300_PACIFY;
+
+END_CS;
+FLUSH_CS;
+
+    r300->dirty_state = R300_NEW_KITCHEN_SINK;
+}
+
+void r300_init_surface_functions(struct r300_context* r300)
+{
+    r300->context.surface_fill = r300_surface_fill;
+}