diff options
author | Christian König <[email protected]> | 2010-11-28 18:49:57 +0100 |
---|---|---|
committer | Christian König <[email protected]> | 2010-12-03 19:04:00 +0100 |
commit | 3bbbb3c54f633ab10e7e2302e8edf3cf1f801858 (patch) | |
tree | 5e007d57d59f1e3a026d68fb5e6d6e7322fbe663 /src | |
parent | 4abe7382882a451a7750ccc451b8568768d122cb (diff) |
move macroblock type handling into its own structure
Diffstat (limited to 'src')
-rw-r--r-- | src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c | 420 | ||||
-rw-r--r-- | src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h | 21 |
2 files changed, 151 insertions, 290 deletions
diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c index 7bb50596de6..b783612b174 100644 --- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c +++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c @@ -87,24 +87,21 @@ enum VS_OUTPUT VS_O_MV3 }; -enum MACROBLOCK_TYPE -{ - MACROBLOCK_TYPE_INTRA, - MACROBLOCK_TYPE_FWD_FRAME_PRED, - MACROBLOCK_TYPE_FWD_FIELD_PRED, - MACROBLOCK_TYPE_BKWD_FRAME_PRED, - MACROBLOCK_TYPE_BKWD_FIELD_PRED, - MACROBLOCK_TYPE_BI_FRAME_PRED, - MACROBLOCK_TYPE_BI_FIELD_PRED, - - NUM_MACROBLOCK_TYPES -}; - /* vertices for a quad covering a macroblock */ static const struct vertex2f const_quad[4] = { {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f} }; +static const unsigned const_mbtype_config[VL_NUM_MACROBLOCK_TYPES][2] = { + [VL_MACROBLOCK_TYPE_INTRA] = { 0, 0 }, + [VL_MACROBLOCK_TYPE_FWD_FRAME_PRED] = { 1, 1 }, + [VL_MACROBLOCK_TYPE_FWD_FIELD_PRED] = { 1, 2 }, + [VL_MACROBLOCK_TYPE_BKWD_FRAME_PRED] = { 1, 1 }, + [VL_MACROBLOCK_TYPE_BKWD_FIELD_PRED] = { 1, 2 }, + [VL_MACROBLOCK_TYPE_BI_FRAME_PRED] = { 2, 1 }, + [VL_MACROBLOCK_TYPE_BI_FIELD_PRED] = { 2, 2 } +}; + static void * create_vert_shader(struct vl_mpeg12_mc_renderer *r, unsigned ref_frames, unsigned mv_per_frame) { @@ -299,222 +296,111 @@ fetch_ycbcr(struct vl_mpeg12_mc_renderer *r, struct ureg_program *shader, struct } static void * -create_intra_frag_shader(struct vl_mpeg12_mc_renderer *r) +create_frag_shader(struct vl_mpeg12_mc_renderer *r, unsigned ref_frames, unsigned mv_per_frame) { struct ureg_program *shader; - struct ureg_dst field, texel; - struct ureg_dst fragment; - - shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); - if (!shader) - return NULL; - - fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0); - - /* - * texel = fetch_ycbcr() - * fragment = texel * scale + 0.5 - */ - field = calc_field(shader); - texel = fetch_ycbcr(r, shader, field); - ureg_ADD(shader, fragment, ureg_src(texel), ureg_scalar(ureg_imm1f(shader, 0.5f), TGSI_SWIZZLE_X)); - - ureg_release_temporary(shader, field); - ureg_release_temporary(shader, texel); - ureg_END(shader); - - return ureg_create_shader_and_destroy(shader, r->pipe); -} - -static void * -create_frame_pred_frag_shader(struct vl_mpeg12_mc_renderer *r) -{ - struct ureg_program *shader; - struct ureg_src tc; - struct ureg_src sampler; - struct ureg_dst field, texel, ref; + struct ureg_src tc[ref_frames * mv_per_frame], sampler[ref_frames], result; + struct ureg_dst field, texel, ref[ref_frames]; struct ureg_dst fragment; + unsigned i, label; shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); if (!shader) return NULL; - tc = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_MV0, TGSI_INTERPOLATE_LINEAR); - sampler = ureg_DECL_sampler(shader, 3); - - ref = ureg_DECL_temporary(shader); fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0); - /* - * texel = fetch_ycbcr() - * ref = tex(tc, sampler) - * fragment = texel * scale + ref - */ - field = calc_field(shader); - texel = fetch_ycbcr(r, shader, field); - ureg_TEX(shader, ref, TGSI_TEXTURE_2D, tc, sampler); - ureg_ADD(shader, fragment, ureg_src(texel), ureg_src(ref)); - - ureg_release_temporary(shader, field); - ureg_release_temporary(shader, texel); - ureg_release_temporary(shader, ref); - ureg_END(shader); - - return ureg_create_shader_and_destroy(shader, r->pipe); -} - -static void * -create_field_pred_frag_shader(struct vl_mpeg12_mc_renderer *r) -{ - struct ureg_program *shader; - struct ureg_src tc[2]; - struct ureg_src sampler; - struct ureg_dst texel, ref, field; - struct ureg_dst fragment; - unsigned i, label; - - shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); - if (!shader) - return NULL; - - for (i = 0; i < 2; ++i) + for (i = 0; i < ref_frames * mv_per_frame; ++i) tc[i] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_MV0 + i, TGSI_INTERPOLATE_LINEAR); - sampler = ureg_DECL_sampler(shader, 3); - ref = ureg_DECL_temporary(shader); - fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0); + for (i = 0; i < ref_frames; ++i) { + sampler[i] = ureg_DECL_sampler(shader, i + 3); + ref[i] = ureg_DECL_temporary(shader); + } - /* - * texel = fetch_ycbcr() - * field = calc_field(); - * if(field == 1) - * ref = tex(tc[1], sampler) - * else - * ref = tex(tc[0], sampler) - * fragment = texel * scale + ref - */ field = calc_field(shader); texel = fetch_ycbcr(r, shader, field); - ureg_IF(shader, ureg_scalar(ureg_src(field), TGSI_SWIZZLE_Y), &label); - ureg_TEX(shader, ref, TGSI_TEXTURE_2D, tc[1], sampler); - ureg_ELSE(shader, &label); - ureg_TEX(shader, ref, TGSI_TEXTURE_2D, tc[0], sampler); - ureg_ENDIF(shader); - - ureg_ADD(shader, fragment, ureg_src(texel), ureg_src(ref)); - - ureg_release_temporary(shader, field); - ureg_release_temporary(shader, texel); - ureg_release_temporary(shader, ref); - ureg_END(shader); - - return ureg_create_shader_and_destroy(shader, r->pipe); -} + switch(ref_frames) { + case 0: + result = ureg_scalar(ureg_imm1f(shader, 0.5f), TGSI_SWIZZLE_X); + break; + + case 1: + if(mv_per_frame == 1) + ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, tc[0], sampler[0]); + else { + ureg_IF(shader, ureg_scalar(ureg_src(field), TGSI_SWIZZLE_Y), &label); + ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, tc[1], sampler[0]); + ureg_ELSE(shader, &label); + ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, tc[0], sampler[0]); + ureg_ENDIF(shader); + } + result = ureg_src(ref[0]); + break; -static void * -create_frame_bi_pred_frag_shader(struct vl_mpeg12_mc_renderer *r) -{ - struct ureg_program *shader; - struct ureg_src tc[2]; - struct ureg_src sampler[2]; - struct ureg_dst field, texel, ref[2]; - struct ureg_dst fragment; - unsigned i; + case 2: + if(mv_per_frame == 1) { + ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, tc[0], sampler[0]); + ureg_TEX(shader, ref[1], TGSI_TEXTURE_2D, tc[1], sampler[1]); + } else { + ureg_IF(shader, ureg_scalar(ureg_src(field), TGSI_SWIZZLE_Y), &label); + ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, tc[1], sampler[0]); + ureg_TEX(shader, ref[1], TGSI_TEXTURE_2D, tc[3], sampler[1]); + ureg_ELSE(shader, &label); + ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, tc[0], sampler[0]); + ureg_TEX(shader, ref[1], TGSI_TEXTURE_2D, tc[2], sampler[1]); + ureg_ENDIF(shader); + } - shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); - if (!shader) - return NULL; + ureg_LRP(shader, ref[0], ureg_scalar(ureg_imm1f(shader, 0.5f), TGSI_SWIZZLE_X), ureg_src(ref[0]), ureg_src(ref[1])); + result = ureg_src(ref[0]); + break; - for (i = 0; i < 2; ++i) { - tc[i] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_MV0 + i, TGSI_INTERPOLATE_LINEAR); - sampler[i] = ureg_DECL_sampler(shader, i + 3); + default: + assert(0); } + ureg_ADD(shader, fragment, ureg_src(texel), result); - ref[0] = ureg_DECL_temporary(shader); - ref[1] = ureg_DECL_temporary(shader); - fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0); - - /* - * texel = fetch_ycbcr() - * ref[0..1 = tex(tc[3..4], sampler[3..4]) - * ref[0] = lerp(ref[0], ref[1], 0.5) - * fragment = texel * scale + ref[0] - */ - field = calc_field(shader); - texel = fetch_ycbcr(r, shader, field); - ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, tc[0], sampler[0]); - ureg_TEX(shader, ref[1], TGSI_TEXTURE_2D, tc[1], sampler[1]); - ureg_LRP(shader, ref[0], ureg_scalar(ureg_imm1f(shader, 0.5f), TGSI_SWIZZLE_X), ureg_src(ref[0]), ureg_src(ref[1])); - - ureg_ADD(shader, fragment, ureg_src(texel), ureg_src(ref[0])); + for (i = 0; i < ref_frames; ++i) + ureg_release_temporary(shader, ref[i]); ureg_release_temporary(shader, field); ureg_release_temporary(shader, texel); - ureg_release_temporary(shader, ref[0]); - ureg_release_temporary(shader, ref[1]); ureg_END(shader); return ureg_create_shader_and_destroy(shader, r->pipe); } -static void * -create_field_bi_pred_frag_shader(struct vl_mpeg12_mc_renderer *r) +static bool +init_mbtype_handler(struct vl_mpeg12_mc_renderer *r, enum VL_MACROBLOCK_TYPE type) { - struct ureg_program *shader; - struct ureg_src tc[4]; - struct ureg_src sampler[2]; - struct ureg_dst texel, ref[2], field; - struct ureg_dst fragment; - unsigned i, label; - - shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); - if (!shader) - return NULL; - - for (i = 0; i < 4; ++i) - tc[i] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_MV0 + i, TGSI_INTERPOLATE_LINEAR); - for (i = 0; i < 2; ++i) - sampler[i] = ureg_DECL_sampler(shader, i + 3); + unsigned ref_frames, mv_per_frame; + struct vl_mc_mbtype_handler *handler; - texel = ureg_DECL_temporary(shader); - ref[0] = ureg_DECL_temporary(shader); - ref[1] = ureg_DECL_temporary(shader); - fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0); + assert(r); - /* - * texel = fetch_ycbcr() - * if(field == 1) - * ref[0..1] = tex(tc[1|3], sampler[0..1]) - * else - * ref[0..1] = tex(tc[0|2], sampler[0..1]) - * ref[0] = lerp(ref[0], ref[1], 0.5) - * fragment = texel * scale + ref[0] - */ - field = calc_field(shader); - texel = fetch_ycbcr(r, shader, field); + ref_frames = const_mbtype_config[type][0]; + mv_per_frame = const_mbtype_config[type][1]; - ureg_IF(shader, ureg_scalar(ureg_src(field), TGSI_SWIZZLE_Y), &label); - ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, tc[1], sampler[0]); - ureg_TEX(shader, ref[1], TGSI_TEXTURE_2D, tc[3], sampler[1]); - ureg_ELSE(shader, &label); - ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, tc[0], sampler[0]); - ureg_TEX(shader, ref[1], TGSI_TEXTURE_2D, tc[2], sampler[1]); - ureg_ENDIF(shader); + handler = &r->mbtype_handlers[type]; - ureg_LRP(shader, ref[0], ureg_scalar(ureg_imm1f(shader, 0.5f), TGSI_SWIZZLE_X), ureg_src(ref[0]), ureg_src(ref[1])); + handler->vs = create_vert_shader(r, ref_frames, mv_per_frame); + handler->fs = create_frag_shader(r, ref_frames, mv_per_frame); - ureg_ADD(shader, fragment, ureg_src(texel), ureg_src(ref[0])); + return handler->vs != NULL && handler->fs != NULL; +} - ureg_release_temporary(shader, field); - ureg_release_temporary(shader, texel); - ureg_release_temporary(shader, ref[0]); - ureg_release_temporary(shader, ref[1]); - ureg_END(shader); +static void +cleanup_mbtype_handler(struct vl_mpeg12_mc_renderer *r, enum VL_MACROBLOCK_TYPE type) +{ + assert(r); - return ureg_create_shader_and_destroy(shader, r->pipe); + r->pipe->delete_vs_state(r->pipe, r->mbtype_handlers[type].vs); + r->pipe->delete_fs_state(r->pipe, r->mbtype_handlers[type].fs); } + static bool init_pipe_state(struct vl_mpeg12_mc_renderer *r) { @@ -595,54 +481,6 @@ cleanup_pipe_state(struct vl_mpeg12_mc_renderer *r) } static bool -init_shaders(struct vl_mpeg12_mc_renderer *r) -{ - assert(r); - - r->i_vs = create_vert_shader(r, 0, 0); - r->i_fs = create_intra_frag_shader(r); - - r->p_vs[0] = create_vert_shader(r, 1, 1); - r->p_vs[1] = create_vert_shader(r, 1, 2); - r->p_fs[0] = create_frame_pred_frag_shader(r); - r->p_fs[1] = create_field_pred_frag_shader(r); - - r->b_vs[0] = create_vert_shader(r, 2, 1); - r->b_vs[1] = create_vert_shader(r, 2, 2); - r->b_fs[0] = create_frame_bi_pred_frag_shader(r); - r->b_fs[1] = create_field_bi_pred_frag_shader(r); - - return - r->i_vs != NULL && - r->i_fs != NULL && - r->p_vs[0] != NULL && - r->p_vs[1] != NULL && - r->p_fs[0] != NULL && - r->p_fs[1] != NULL && - r->b_vs[0] != NULL && - r->b_vs[1] != NULL && - r->b_fs[0] != NULL && - r->b_fs[1] != NULL; -} - -static void -cleanup_shaders(struct vl_mpeg12_mc_renderer *r) -{ - assert(r); - - r->pipe->delete_vs_state(r->pipe, r->i_vs); - r->pipe->delete_fs_state(r->pipe, r->i_fs); - r->pipe->delete_vs_state(r->pipe, r->p_vs[0]); - r->pipe->delete_vs_state(r->pipe, r->p_vs[1]); - r->pipe->delete_fs_state(r->pipe, r->p_fs[0]); - r->pipe->delete_fs_state(r->pipe, r->p_fs[1]); - r->pipe->delete_vs_state(r->pipe, r->b_vs[0]); - r->pipe->delete_vs_state(r->pipe, r->b_vs[1]); - r->pipe->delete_fs_state(r->pipe, r->b_fs[0]); - r->pipe->delete_fs_state(r->pipe, r->b_fs[1]); -} - -static bool init_buffers(struct vl_mpeg12_mc_renderer *r) { struct pipe_resource template; @@ -836,23 +674,23 @@ cleanup_buffers(struct vl_mpeg12_mc_renderer *r) FREE(r->macroblock_buf); } -static enum MACROBLOCK_TYPE +static enum VL_MACROBLOCK_TYPE get_macroblock_type(struct pipe_mpeg12_macroblock *mb) { assert(mb); switch (mb->mb_type) { case PIPE_MPEG12_MACROBLOCK_TYPE_INTRA: - return MACROBLOCK_TYPE_INTRA; + return VL_MACROBLOCK_TYPE_INTRA; case PIPE_MPEG12_MACROBLOCK_TYPE_FWD: return mb->mo_type == PIPE_MPEG12_MOTION_TYPE_FRAME ? - MACROBLOCK_TYPE_FWD_FRAME_PRED : MACROBLOCK_TYPE_FWD_FIELD_PRED; + VL_MACROBLOCK_TYPE_FWD_FRAME_PRED : VL_MACROBLOCK_TYPE_FWD_FIELD_PRED; case PIPE_MPEG12_MACROBLOCK_TYPE_BKWD: return mb->mo_type == PIPE_MPEG12_MOTION_TYPE_FRAME ? - MACROBLOCK_TYPE_BKWD_FRAME_PRED : MACROBLOCK_TYPE_BKWD_FIELD_PRED; + VL_MACROBLOCK_TYPE_BKWD_FRAME_PRED : VL_MACROBLOCK_TYPE_BKWD_FIELD_PRED; case PIPE_MPEG12_MACROBLOCK_TYPE_BI: return mb->mo_type == PIPE_MPEG12_MOTION_TYPE_FRAME ? - MACROBLOCK_TYPE_BI_FRAME_PRED : MACROBLOCK_TYPE_BI_FIELD_PRED; + VL_MACROBLOCK_TYPE_BI_FRAME_PRED : VL_MACROBLOCK_TYPE_BI_FIELD_PRED; default: assert(0); } @@ -994,7 +832,7 @@ static void gen_macroblock_stream(struct vl_mpeg12_mc_renderer *r, unsigned *num_macroblocks) { - unsigned offset[NUM_MACROBLOCK_TYPES]; + unsigned offset[VL_NUM_MACROBLOCK_TYPES]; struct vert_stream_0 *ycbcr_vb; struct vertex2f *ref_vb[2]; struct pipe_transfer *buf_transfer[3]; @@ -1004,13 +842,13 @@ gen_macroblock_stream(struct vl_mpeg12_mc_renderer *r, assert(num_macroblocks); for (i = 0; i < r->num_macroblocks; ++i) { - enum MACROBLOCK_TYPE mb_type = get_macroblock_type(&r->macroblock_buf[i]); + enum VL_MACROBLOCK_TYPE mb_type = get_macroblock_type(&r->macroblock_buf[i]); ++num_macroblocks[mb_type]; } offset[0] = 0; - for (i = 1; i < NUM_MACROBLOCK_TYPES; ++i) + for (i = 1; i < VL_NUM_MACROBLOCK_TYPES; ++i) offset[i] = offset[i - 1] + num_macroblocks[i - 1]; ycbcr_vb = (struct vert_stream_0 *)pipe_buffer_map @@ -1031,7 +869,7 @@ gen_macroblock_stream(struct vl_mpeg12_mc_renderer *r, ); for (i = 0; i < r->num_macroblocks; ++i) { - enum MACROBLOCK_TYPE mb_type = get_macroblock_type(&r->macroblock_buf[i]); + enum VL_MACROBLOCK_TYPE mb_type = get_macroblock_type(&r->macroblock_buf[i]); gen_macroblock_verts(r, &r->macroblock_buf[i], offset[mb_type], ycbcr_vb, ref_vb); @@ -1074,7 +912,7 @@ static struct pipe_sampler_view static void flush(struct vl_mpeg12_mc_renderer *r) { - unsigned num_macroblocks[NUM_MACROBLOCK_TYPES] = { 0 }; + unsigned num_macroblocks[VL_NUM_MACROBLOCK_TYPES] = { 0 }; unsigned vb_start = 0; assert(r); @@ -1089,80 +927,80 @@ flush(struct vl_mpeg12_mc_renderer *r) r->pipe->set_framebuffer_state(r->pipe, &r->fb_state); r->pipe->set_viewport_state(r->pipe, &r->viewport); - if (num_macroblocks[MACROBLOCK_TYPE_INTRA] > 0) { + if (num_macroblocks[VL_MACROBLOCK_TYPE_INTRA] > 0) { r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all); r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.i); r->pipe->set_fragment_sampler_views(r->pipe, 3, r->sampler_views.all); r->pipe->bind_fragment_sampler_states(r->pipe, 3, r->samplers.all); - r->pipe->bind_vs_state(r->pipe, r->i_vs); - r->pipe->bind_fs_state(r->pipe, r->i_fs); + r->pipe->bind_vs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_INTRA].vs); + r->pipe->bind_fs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_INTRA].fs); util_draw_arrays(r->pipe, PIPE_PRIM_QUADS, vb_start, - num_macroblocks[MACROBLOCK_TYPE_INTRA] * 4); - vb_start += num_macroblocks[MACROBLOCK_TYPE_INTRA] * 4; + num_macroblocks[VL_MACROBLOCK_TYPE_INTRA] * 4); + vb_start += num_macroblocks[VL_MACROBLOCK_TYPE_INTRA] * 4; } - if (num_macroblocks[MACROBLOCK_TYPE_FWD_FRAME_PRED] > 0) { + if (num_macroblocks[VL_MACROBLOCK_TYPE_FWD_FRAME_PRED] > 0) { r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all); r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.p); r->textures.individual.ref[0] = r->past->texture; r->sampler_views.individual.ref[0] = find_or_create_sampler_view(r, r->past); r->pipe->set_fragment_sampler_views(r->pipe, 4, r->sampler_views.all); r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all); - r->pipe->bind_vs_state(r->pipe, r->p_vs[0]); - r->pipe->bind_fs_state(r->pipe, r->p_fs[0]); + r->pipe->bind_vs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_FWD_FRAME_PRED].vs); + r->pipe->bind_fs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_FWD_FRAME_PRED].fs); util_draw_arrays(r->pipe, PIPE_PRIM_QUADS, vb_start, - num_macroblocks[MACROBLOCK_TYPE_FWD_FRAME_PRED] * 4); - vb_start += num_macroblocks[MACROBLOCK_TYPE_FWD_FRAME_PRED] * 4; + num_macroblocks[VL_MACROBLOCK_TYPE_FWD_FRAME_PRED] * 4); + vb_start += num_macroblocks[VL_MACROBLOCK_TYPE_FWD_FRAME_PRED] * 4; } - if (num_macroblocks[MACROBLOCK_TYPE_FWD_FIELD_PRED] > 0) { + if (num_macroblocks[VL_MACROBLOCK_TYPE_FWD_FIELD_PRED] > 0) { r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all); r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.p); r->textures.individual.ref[0] = r->past->texture; r->sampler_views.individual.ref[0] = find_or_create_sampler_view(r, r->past); r->pipe->set_fragment_sampler_views(r->pipe, 4, r->sampler_views.all); r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all); - r->pipe->bind_vs_state(r->pipe, r->p_vs[1]); - r->pipe->bind_fs_state(r->pipe, r->p_fs[1]); + r->pipe->bind_vs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_FWD_FIELD_PRED].vs); + r->pipe->bind_fs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_FWD_FIELD_PRED].fs); util_draw_arrays(r->pipe, PIPE_PRIM_QUADS, vb_start, - num_macroblocks[MACROBLOCK_TYPE_FWD_FIELD_PRED] * 4); - vb_start += num_macroblocks[MACROBLOCK_TYPE_FWD_FIELD_PRED] * 4; + num_macroblocks[VL_MACROBLOCK_TYPE_FWD_FIELD_PRED] * 4); + vb_start += num_macroblocks[VL_MACROBLOCK_TYPE_FWD_FIELD_PRED] * 4; } - if (num_macroblocks[MACROBLOCK_TYPE_BKWD_FRAME_PRED] > 0) { + if (num_macroblocks[VL_MACROBLOCK_TYPE_BKWD_FRAME_PRED] > 0) { r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all); r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.p); r->textures.individual.ref[0] = r->future->texture; r->sampler_views.individual.ref[0] = find_or_create_sampler_view(r, r->future); r->pipe->set_fragment_sampler_views(r->pipe, 4, r->sampler_views.all); r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all); - r->pipe->bind_vs_state(r->pipe, r->p_vs[0]); - r->pipe->bind_fs_state(r->pipe, r->p_fs[0]); + r->pipe->bind_vs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_BKWD_FRAME_PRED].vs); + r->pipe->bind_fs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_BKWD_FRAME_PRED].fs); util_draw_arrays(r->pipe, PIPE_PRIM_QUADS, vb_start, - num_macroblocks[MACROBLOCK_TYPE_BKWD_FRAME_PRED] * 4); - vb_start += num_macroblocks[MACROBLOCK_TYPE_BKWD_FRAME_PRED] * 4; + num_macroblocks[VL_MACROBLOCK_TYPE_BKWD_FRAME_PRED] * 4); + vb_start += num_macroblocks[VL_MACROBLOCK_TYPE_BKWD_FRAME_PRED] * 4; } - if (num_macroblocks[MACROBLOCK_TYPE_BKWD_FIELD_PRED] > 0) { + if (num_macroblocks[VL_MACROBLOCK_TYPE_BKWD_FIELD_PRED] > 0) { r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all); r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.p); r->textures.individual.ref[0] = r->future->texture; r->sampler_views.individual.ref[0] = find_or_create_sampler_view(r, r->future); r->pipe->set_fragment_sampler_views(r->pipe, 4, r->sampler_views.all); r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all); - r->pipe->bind_vs_state(r->pipe, r->p_vs[1]); - r->pipe->bind_fs_state(r->pipe, r->p_fs[1]); + r->pipe->bind_vs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_BKWD_FIELD_PRED].vs); + r->pipe->bind_fs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_BKWD_FIELD_PRED].fs); util_draw_arrays(r->pipe, PIPE_PRIM_QUADS, vb_start, - num_macroblocks[MACROBLOCK_TYPE_BKWD_FIELD_PRED] * 4); - vb_start += num_macroblocks[MACROBLOCK_TYPE_BKWD_FIELD_PRED] * 4; + num_macroblocks[VL_MACROBLOCK_TYPE_BKWD_FIELD_PRED] * 4); + vb_start += num_macroblocks[VL_MACROBLOCK_TYPE_BKWD_FIELD_PRED] * 4; } - if (num_macroblocks[MACROBLOCK_TYPE_BI_FRAME_PRED] > 0) { + if (num_macroblocks[VL_MACROBLOCK_TYPE_BI_FRAME_PRED] > 0) { r->pipe->set_vertex_buffers(r->pipe, 4, r->vertex_bufs.all); r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.b); r->textures.individual.ref[0] = r->past->texture; @@ -1171,15 +1009,15 @@ flush(struct vl_mpeg12_mc_renderer *r) r->sampler_views.individual.ref[1] = find_or_create_sampler_view(r, r->future); r->pipe->set_fragment_sampler_views(r->pipe, 5, r->sampler_views.all); r->pipe->bind_fragment_sampler_states(r->pipe, 5, r->samplers.all); - r->pipe->bind_vs_state(r->pipe, r->b_vs[0]); - r->pipe->bind_fs_state(r->pipe, r->b_fs[0]); + r->pipe->bind_vs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_BI_FRAME_PRED].vs); + r->pipe->bind_fs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_BI_FRAME_PRED].fs); util_draw_arrays(r->pipe, PIPE_PRIM_QUADS, vb_start, - num_macroblocks[MACROBLOCK_TYPE_BI_FRAME_PRED] * 4); - vb_start += num_macroblocks[MACROBLOCK_TYPE_BI_FRAME_PRED] * 4; + num_macroblocks[VL_MACROBLOCK_TYPE_BI_FRAME_PRED] * 4); + vb_start += num_macroblocks[VL_MACROBLOCK_TYPE_BI_FRAME_PRED] * 4; } - if (num_macroblocks[MACROBLOCK_TYPE_BI_FIELD_PRED] > 0) { + if (num_macroblocks[VL_MACROBLOCK_TYPE_BI_FIELD_PRED] > 0) { r->pipe->set_vertex_buffers(r->pipe, 4, r->vertex_bufs.all); r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.b); r->textures.individual.ref[0] = r->past->texture; @@ -1188,12 +1026,12 @@ flush(struct vl_mpeg12_mc_renderer *r) r->sampler_views.individual.ref[1] = find_or_create_sampler_view(r, r->future); r->pipe->set_fragment_sampler_views(r->pipe, 5, r->sampler_views.all); r->pipe->bind_fragment_sampler_states(r->pipe, 5, r->samplers.all); - r->pipe->bind_vs_state(r->pipe, r->b_vs[1]); - r->pipe->bind_fs_state(r->pipe, r->b_fs[1]); + r->pipe->bind_vs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_BI_FIELD_PRED].vs); + r->pipe->bind_fs_state(r->pipe, r->mbtype_handlers[VL_MACROBLOCK_TYPE_BI_FIELD_PRED].fs); util_draw_arrays(r->pipe, PIPE_PRIM_QUADS, vb_start, - num_macroblocks[MACROBLOCK_TYPE_BI_FIELD_PRED] * 4); - vb_start += num_macroblocks[MACROBLOCK_TYPE_BI_FIELD_PRED] * 4; + num_macroblocks[VL_MACROBLOCK_TYPE_BI_FIELD_PRED] * 4); + vb_start += num_macroblocks[VL_MACROBLOCK_TYPE_BI_FIELD_PRED] * 4; } r->pipe->flush(r->pipe, PIPE_FLUSH_RENDER_CACHE, r->fence); @@ -1297,6 +1135,7 @@ vl_mpeg12_mc_renderer_init(struct vl_mpeg12_mc_renderer *renderer, bool pot_buffers) { struct pipe_resource *idct_matrix; + unsigned i; assert(renderer); assert(pipe); @@ -1323,8 +1162,9 @@ vl_mpeg12_mc_renderer_init(struct vl_mpeg12_mc_renderer *renderer, if (!init_pipe_state(renderer)) goto error_pipe_state; - if (!init_shaders(renderer)) - goto error_shaders; + for(i = 0; i<VL_NUM_MACROBLOCK_TYPES; ++i) + if (!init_mbtype_handler(renderer, i)) + goto error_mbtypes; if (!init_buffers(renderer)) goto error_buffers; @@ -1361,9 +1201,10 @@ error_idct_matrix: cleanup_buffers(renderer); error_buffers: - cleanup_shaders(renderer); + for(i = 0; i<VL_NUM_MACROBLOCK_TYPES; ++i) + cleanup_mbtype_handler(renderer, i); -error_shaders: +error_mbtypes: cleanup_pipe_state(renderer); error_pipe_state: @@ -1374,6 +1215,8 @@ error_pipe_state: void vl_mpeg12_mc_renderer_cleanup(struct vl_mpeg12_mc_renderer *renderer) { + unsigned i; + assert(renderer); vl_idct_cleanup(&renderer->idct_y); @@ -1382,7 +1225,8 @@ vl_mpeg12_mc_renderer_cleanup(struct vl_mpeg12_mc_renderer *renderer) util_delete_keymap(renderer->texview_map, renderer->pipe); cleanup_pipe_state(renderer); - cleanup_shaders(renderer); + for(i = 0; i<VL_NUM_MACROBLOCK_TYPES; ++i) + cleanup_mbtype_handler(renderer, i); cleanup_buffers(renderer); pipe_surface_reference(&renderer->surface, NULL); diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h index ed48b5b6b45..69fa4bb7f23 100644 --- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h +++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h @@ -45,6 +45,24 @@ enum VL_MPEG12_MC_RENDERER_BUFFER_MODE VL_MPEG12_MC_RENDERER_BUFFER_PICTURE /* Larger batches, more memory */ }; +enum VL_MACROBLOCK_TYPE +{ + VL_MACROBLOCK_TYPE_INTRA, + VL_MACROBLOCK_TYPE_FWD_FRAME_PRED, + VL_MACROBLOCK_TYPE_FWD_FIELD_PRED, + VL_MACROBLOCK_TYPE_BKWD_FRAME_PRED, + VL_MACROBLOCK_TYPE_BKWD_FIELD_PRED, + VL_MACROBLOCK_TYPE_BI_FRAME_PRED, + VL_MACROBLOCK_TYPE_BI_FIELD_PRED, + + VL_NUM_MACROBLOCK_TYPES +}; + +struct vl_mc_mbtype_handler +{ + void *vs, *fs; +}; + struct vl_mpeg12_mc_renderer { struct pipe_context *pipe; @@ -79,8 +97,7 @@ struct vl_mpeg12_mc_renderer struct { struct pipe_sampler_view *y, *cb, *cr, *ref[2]; } individual; } sampler_views; - void *i_vs, *p_vs[2], *b_vs[2]; - void *i_fs, *p_fs[2], *b_fs[2]; + struct vl_mc_mbtype_handler mbtype_handlers[VL_NUM_MACROBLOCK_TYPES]; union { |