From a3d9d7ec79d6f7205fab2324e47d8ea185431de0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Marek=20Ol=C5=A1=C3=A1k?= Date: Sun, 12 Aug 2012 20:06:33 +0200 Subject: [PATCH] r600g: implement compression for MSAA colorbuffers for evergreen This adds the FMASK and CMASK buffers. They share the same resource with color data. COMPRESSION and FAST_CLEAR are always enabled if both FMASK and CMASK are allocated. We initialize the CMASK to a "compressed" state (not "fast cleared"), so that we can keep FAST_CLEAR enabled all the time. Both FMASK and CMASK must be present at the moment. If either one is missing, the other one is not used. v2: add cayman regs in the list Reviewed-by: Jerome Glisse --- src/gallium/auxiliary/util/u_blitter.c | 45 ++++++ src/gallium/auxiliary/util/u_blitter.h | 5 + .../drivers/r600/evergreen_hw_context.c | 128 ++++++++++++++++++ src/gallium/drivers/r600/evergreen_state.c | 54 +++++++- src/gallium/drivers/r600/evergreend.h | 76 +++++++++-- src/gallium/drivers/r600/r600_blit.c | 79 +++++++++++ src/gallium/drivers/r600/r600_pipe.c | 4 + src/gallium/drivers/r600/r600_pipe.h | 6 + src/gallium/drivers/r600/r600_resource.h | 11 +- src/gallium/drivers/r600/r600_state.c | 4 +- src/gallium/drivers/r600/r600_state_common.c | 27 ++++ src/gallium/drivers/r600/r600_texture.c | 109 +++++++++++++++ 12 files changed, 529 insertions(+), 19 deletions(-) diff --git a/src/gallium/auxiliary/util/u_blitter.c b/src/gallium/auxiliary/util/u_blitter.c index 25c7119b9f2..ad4ccd9eb46 100644 --- a/src/gallium/auxiliary/util/u_blitter.c +++ b/src/gallium/auxiliary/util/u_blitter.c @@ -1553,3 +1553,48 @@ void util_blitter_custom_resolve_color(struct blitter_context *blitter, pipe_surface_reference(&srcsurf, NULL); pipe_surface_reference(&dstsurf, NULL); } + +void util_blitter_custom_color(struct blitter_context *blitter, + struct pipe_surface *dstsurf, + void *custom_blend) +{ + struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; + struct pipe_context *pipe = ctx->base.pipe; + struct pipe_framebuffer_state fb_state; + + assert(dstsurf->texture); + if (!dstsurf->texture) + return; + + /* check the saved state */ + blitter_set_running_flag(ctx); + blitter_check_saved_vertex_states(ctx); + blitter_check_saved_fragment_states(ctx); + blitter_check_saved_fb_state(ctx); + + /* bind states */ + pipe->bind_blend_state(pipe, custom_blend); + pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil); + pipe->bind_fs_state(pipe, blitter_get_fs_col(ctx, 1, FALSE)); + pipe->bind_vertex_elements_state(pipe, ctx->velem_state); + pipe->set_sample_mask(pipe, (1ull << MAX2(1, dstsurf->texture->nr_samples)) - 1); + + /* set a framebuffer state */ + fb_state.width = dstsurf->width; + fb_state.height = dstsurf->height; + fb_state.nr_cbufs = 1; + fb_state.cbufs[0] = dstsurf; + fb_state.zsbuf = 0; + pipe->set_framebuffer_state(pipe, &fb_state); + pipe->set_sample_mask(pipe, ~0); + + blitter_set_common_draw_rect_state(ctx); + blitter_set_dst_dimensions(ctx, dstsurf->width, dstsurf->height); + blitter->draw_rectangle(blitter, 0, 0, dstsurf->width, dstsurf->height, + 0, 0, NULL); + + blitter_restore_vertex_states(ctx); + blitter_restore_fragment_states(ctx); + blitter_restore_fb_state(ctx); + blitter_unset_running_flag(ctx); +} diff --git a/src/gallium/auxiliary/util/u_blitter.h b/src/gallium/auxiliary/util/u_blitter.h index f227902c163..e06e8b12d53 100644 --- a/src/gallium/auxiliary/util/u_blitter.h +++ b/src/gallium/auxiliary/util/u_blitter.h @@ -308,6 +308,11 @@ void util_blitter_custom_depth_stencil(struct blitter_context *blitter, unsigned sample_mask, void *dsa_stage, float depth); +/* Used by r600g for color decompression. */ +void util_blitter_custom_color(struct blitter_context *blitter, + struct pipe_surface *dstsurf, + void *custom_blend); + /* Used by r600g for MSAA color resolve. */ void util_blitter_custom_resolve_color(struct blitter_context *blitter, struct pipe_resource *dst, diff --git a/src/gallium/drivers/r600/evergreen_hw_context.c b/src/gallium/drivers/r600/evergreen_hw_context.c index f0c4ff7a482..d2f09498566 100644 --- a/src/gallium/drivers/r600/evergreen_hw_context.c +++ b/src/gallium/drivers/r600/evergreen_hw_context.c @@ -219,6 +219,14 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028C70_CB_COLOR0_INFO, REG_FLAG_NEED_BO, 0}, {R_028C74_CB_COLOR0_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028C78_CB_COLOR0_DIM, 0, 0}, + {R_028C7C_CB_COLOR0_CMASK, REG_FLAG_NEED_BO}, + {R_028C80_CB_COLOR0_CMASK_SLICE}, + {R_028C84_CB_COLOR0_FMASK, REG_FLAG_NEED_BO}, + {R_028C88_CB_COLOR0_FMASK_SLICE}, + {R_028C8C_CB_COLOR0_CLEAR_WORD0}, + {R_028C90_CB_COLOR0_CLEAR_WORD1}, + {R_028C94_CB_COLOR0_CLEAR_WORD2}, + {R_028C98_CB_COLOR0_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028C9C_CB_COLOR1_BASE, REG_FLAG_NEED_BO, 0}, {R_028CA0_CB_COLOR1_PITCH, 0, 0}, @@ -227,6 +235,14 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028CAC_CB_COLOR1_INFO, REG_FLAG_NEED_BO, 0}, {R_028CB0_CB_COLOR1_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028CB4_CB_COLOR1_DIM, 0, 0}, + {R_028CB8_CB_COLOR1_CMASK, REG_FLAG_NEED_BO, 0}, + {R_028CBC_CB_COLOR1_CMASK_SLICE, 0, 0}, + {R_028CC0_CB_COLOR1_FMASK, REG_FLAG_NEED_BO, 0}, + {R_028CC4_CB_COLOR1_FMASK_SLICE, 0, 0}, + {R_028CC8_CB_COLOR1_CLEAR_WORD0}, + {R_028CCC_CB_COLOR1_CLEAR_WORD1}, + {R_028CD0_CB_COLOR1_CLEAR_WORD2}, + {R_028CD4_CB_COLOR1_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028CD8_CB_COLOR2_BASE, REG_FLAG_NEED_BO, 0}, {R_028CDC_CB_COLOR2_PITCH, 0, 0}, @@ -235,6 +251,14 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028CE8_CB_COLOR2_INFO, REG_FLAG_NEED_BO, 0}, {R_028CEC_CB_COLOR2_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028CF0_CB_COLOR2_DIM, 0, 0}, + {R_028CF4_CB_COLOR2_CMASK, REG_FLAG_NEED_BO, 0}, + {R_028CF8_CB_COLOR2_CMASK_SLICE, 0, 0}, + {R_028CFC_CB_COLOR2_FMASK, REG_FLAG_NEED_BO, 0}, + {R_028D00_CB_COLOR2_FMASK_SLICE, 0, 0}, + {R_028D04_CB_COLOR2_CLEAR_WORD0}, + {R_028D08_CB_COLOR2_CLEAR_WORD1}, + {R_028D0C_CB_COLOR2_CLEAR_WORD2}, + {R_028D10_CB_COLOR2_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028D14_CB_COLOR3_BASE, REG_FLAG_NEED_BO, 0}, {R_028D18_CB_COLOR3_PITCH, 0, 0}, @@ -243,6 +267,14 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028D24_CB_COLOR3_INFO, REG_FLAG_NEED_BO, 0}, {R_028D28_CB_COLOR3_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028D2C_CB_COLOR3_DIM, 0, 0}, + {R_028D30_CB_COLOR3_CMASK, REG_FLAG_NEED_BO}, + {R_028D34_CB_COLOR3_CMASK_SLICE}, + {R_028D38_CB_COLOR3_FMASK, REG_FLAG_NEED_BO}, + {R_028D3C_CB_COLOR3_FMASK_SLICE}, + {R_028D40_CB_COLOR3_CLEAR_WORD0}, + {R_028D44_CB_COLOR3_CLEAR_WORD1}, + {R_028D48_CB_COLOR3_CLEAR_WORD2}, + {R_028D4C_CB_COLOR3_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028D50_CB_COLOR4_BASE, REG_FLAG_NEED_BO, 0}, {R_028D54_CB_COLOR4_PITCH, 0, 0}, @@ -251,6 +283,14 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028D60_CB_COLOR4_INFO, REG_FLAG_NEED_BO, 0}, {R_028D64_CB_COLOR4_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028D68_CB_COLOR4_DIM, 0, 0}, + {R_028D6C_CB_COLOR4_CMASK, REG_FLAG_NEED_BO}, + {R_028D70_CB_COLOR4_CMASK_SLICE}, + {R_028D74_CB_COLOR4_FMASK, REG_FLAG_NEED_BO}, + {R_028D78_CB_COLOR4_FMASK_SLICE}, + {R_028D7C_CB_COLOR4_CLEAR_WORD0}, + {R_028D80_CB_COLOR4_CLEAR_WORD1}, + {R_028D84_CB_COLOR4_CLEAR_WORD2}, + {R_028D88_CB_COLOR4_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028D8C_CB_COLOR5_BASE, REG_FLAG_NEED_BO, 0}, {R_028D90_CB_COLOR5_PITCH, 0, 0}, @@ -259,6 +299,14 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028D9C_CB_COLOR5_INFO, REG_FLAG_NEED_BO, 0}, {R_028DA0_CB_COLOR5_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028DA4_CB_COLOR5_DIM, 0, 0}, + {R_028DA8_CB_COLOR5_CMASK, REG_FLAG_NEED_BO}, + {R_028DAC_CB_COLOR5_CMASK_SLICE}, + {R_028DB0_CB_COLOR5_FMASK, REG_FLAG_NEED_BO}, + {R_028DB4_CB_COLOR5_FMASK_SLICE}, + {R_028DB8_CB_COLOR5_CLEAR_WORD0}, + {R_028DBC_CB_COLOR5_CLEAR_WORD1}, + {R_028DC0_CB_COLOR5_CLEAR_WORD2}, + {R_028DC4_CB_COLOR5_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028DC8_CB_COLOR6_BASE, REG_FLAG_NEED_BO, 0}, {R_028DCC_CB_COLOR6_PITCH, 0, 0}, @@ -267,6 +315,14 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028DD8_CB_COLOR6_INFO, REG_FLAG_NEED_BO, 0}, {R_028DDC_CB_COLOR6_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028DE0_CB_COLOR6_DIM, 0, 0}, + {R_028DE4_CB_COLOR6_CMASK, REG_FLAG_NEED_BO}, + {R_028DE8_CB_COLOR6_CMASK_SLICE}, + {R_028DEC_CB_COLOR6_FMASK, REG_FLAG_NEED_BO}, + {R_028DF0_CB_COLOR6_FMASK_SLICE}, + {R_028DF4_CB_COLOR6_CLEAR_WORD0}, + {R_028DF8_CB_COLOR6_CLEAR_WORD1}, + {R_028DFC_CB_COLOR6_CLEAR_WORD2}, + {R_028E00_CB_COLOR6_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028E04_CB_COLOR7_BASE, REG_FLAG_NEED_BO, 0}, {R_028E08_CB_COLOR7_PITCH, 0, 0}, @@ -275,6 +331,14 @@ static const struct r600_reg evergreen_context_reg_list[] = { {R_028E14_CB_COLOR7_INFO, REG_FLAG_NEED_BO, 0}, {R_028E18_CB_COLOR7_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028E1C_CB_COLOR7_DIM, 0, 0}, + {R_028E20_CB_COLOR7_CMASK, REG_FLAG_NEED_BO}, + {R_028E24_CB_COLOR7_CMASK_SLICE}, + {R_028E28_CB_COLOR7_FMASK, REG_FLAG_NEED_BO}, + {R_028E2C_CB_COLOR7_FMASK_SLICE}, + {R_028E30_CB_COLOR7_CLEAR_WORD0}, + {R_028E34_CB_COLOR7_CLEAR_WORD1}, + {R_028E38_CB_COLOR7_CLEAR_WORD2}, + {R_028E3C_CB_COLOR7_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028E40_CB_COLOR8_BASE, REG_FLAG_NEED_BO, 0}, {R_028E44_CB_COLOR8_PITCH, 0, 0}, @@ -500,6 +564,14 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028C70_CB_COLOR0_INFO, REG_FLAG_NEED_BO, 0}, {R_028C74_CB_COLOR0_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028C78_CB_COLOR0_DIM, 0, 0}, + {R_028C7C_CB_COLOR0_CMASK, REG_FLAG_NEED_BO}, + {R_028C80_CB_COLOR0_CMASK_SLICE}, + {R_028C84_CB_COLOR0_FMASK, REG_FLAG_NEED_BO}, + {R_028C88_CB_COLOR0_FMASK_SLICE}, + {R_028C8C_CB_COLOR0_CLEAR_WORD0}, + {R_028C90_CB_COLOR0_CLEAR_WORD1}, + {R_028C94_CB_COLOR0_CLEAR_WORD2}, + {R_028C98_CB_COLOR0_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028C9C_CB_COLOR1_BASE, REG_FLAG_NEED_BO, 0}, {R_028CA0_CB_COLOR1_PITCH, 0, 0}, @@ -508,6 +580,14 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028CAC_CB_COLOR1_INFO, REG_FLAG_NEED_BO, 0}, {R_028CB0_CB_COLOR1_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028CB4_CB_COLOR1_DIM, 0, 0}, + {R_028CB8_CB_COLOR1_CMASK, REG_FLAG_NEED_BO, 0}, + {R_028CBC_CB_COLOR1_CMASK_SLICE, 0, 0}, + {R_028CC0_CB_COLOR1_FMASK, REG_FLAG_NEED_BO, 0}, + {R_028CC4_CB_COLOR1_FMASK_SLICE, 0, 0}, + {R_028CC8_CB_COLOR1_CLEAR_WORD0}, + {R_028CCC_CB_COLOR1_CLEAR_WORD1}, + {R_028CD0_CB_COLOR1_CLEAR_WORD2}, + {R_028CD4_CB_COLOR1_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028CD8_CB_COLOR2_BASE, REG_FLAG_NEED_BO, 0}, {R_028CDC_CB_COLOR2_PITCH, 0, 0}, @@ -516,6 +596,14 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028CE8_CB_COLOR2_INFO, REG_FLAG_NEED_BO, 0}, {R_028CEC_CB_COLOR2_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028CF0_CB_COLOR2_DIM, 0, 0}, + {R_028CF4_CB_COLOR2_CMASK, REG_FLAG_NEED_BO, 0}, + {R_028CF8_CB_COLOR2_CMASK_SLICE, 0, 0}, + {R_028CFC_CB_COLOR2_FMASK, REG_FLAG_NEED_BO, 0}, + {R_028D00_CB_COLOR2_FMASK_SLICE, 0, 0}, + {R_028D04_CB_COLOR2_CLEAR_WORD0}, + {R_028D08_CB_COLOR2_CLEAR_WORD1}, + {R_028D0C_CB_COLOR2_CLEAR_WORD2}, + {R_028D10_CB_COLOR2_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028D14_CB_COLOR3_BASE, REG_FLAG_NEED_BO, 0}, {R_028D18_CB_COLOR3_PITCH, 0, 0}, @@ -524,6 +612,14 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028D24_CB_COLOR3_INFO, REG_FLAG_NEED_BO, 0}, {R_028D28_CB_COLOR3_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028D2C_CB_COLOR3_DIM, 0, 0}, + {R_028D30_CB_COLOR3_CMASK, REG_FLAG_NEED_BO}, + {R_028D34_CB_COLOR3_CMASK_SLICE}, + {R_028D38_CB_COLOR3_FMASK, REG_FLAG_NEED_BO}, + {R_028D3C_CB_COLOR3_FMASK_SLICE}, + {R_028D40_CB_COLOR3_CLEAR_WORD0}, + {R_028D44_CB_COLOR3_CLEAR_WORD1}, + {R_028D48_CB_COLOR3_CLEAR_WORD2}, + {R_028D4C_CB_COLOR3_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028D50_CB_COLOR4_BASE, REG_FLAG_NEED_BO, 0}, {R_028D54_CB_COLOR4_PITCH, 0, 0}, @@ -532,6 +628,14 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028D60_CB_COLOR4_INFO, REG_FLAG_NEED_BO, 0}, {R_028D64_CB_COLOR4_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028D68_CB_COLOR4_DIM, 0, 0}, + {R_028D6C_CB_COLOR4_CMASK, REG_FLAG_NEED_BO}, + {R_028D70_CB_COLOR4_CMASK_SLICE}, + {R_028D74_CB_COLOR4_FMASK, REG_FLAG_NEED_BO}, + {R_028D78_CB_COLOR4_FMASK_SLICE}, + {R_028D7C_CB_COLOR4_CLEAR_WORD0}, + {R_028D80_CB_COLOR4_CLEAR_WORD1}, + {R_028D84_CB_COLOR4_CLEAR_WORD2}, + {R_028D88_CB_COLOR4_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028D8C_CB_COLOR5_BASE, REG_FLAG_NEED_BO, 0}, {R_028D90_CB_COLOR5_PITCH, 0, 0}, @@ -540,6 +644,14 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028D9C_CB_COLOR5_INFO, REG_FLAG_NEED_BO, 0}, {R_028DA0_CB_COLOR5_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028DA4_CB_COLOR5_DIM, 0, 0}, + {R_028DA8_CB_COLOR5_CMASK, REG_FLAG_NEED_BO}, + {R_028DAC_CB_COLOR5_CMASK_SLICE}, + {R_028DB0_CB_COLOR5_FMASK, REG_FLAG_NEED_BO}, + {R_028DB4_CB_COLOR5_FMASK_SLICE}, + {R_028DB8_CB_COLOR5_CLEAR_WORD0}, + {R_028DBC_CB_COLOR5_CLEAR_WORD1}, + {R_028DC0_CB_COLOR5_CLEAR_WORD2}, + {R_028DC4_CB_COLOR5_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028DC8_CB_COLOR6_BASE, REG_FLAG_NEED_BO, 0}, {R_028DCC_CB_COLOR6_PITCH, 0, 0}, @@ -548,6 +660,14 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028DD8_CB_COLOR6_INFO, REG_FLAG_NEED_BO, 0}, {R_028DDC_CB_COLOR6_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028DE0_CB_COLOR6_DIM, 0, 0}, + {R_028DE4_CB_COLOR6_CMASK, REG_FLAG_NEED_BO}, + {R_028DE8_CB_COLOR6_CMASK_SLICE}, + {R_028DEC_CB_COLOR6_FMASK, REG_FLAG_NEED_BO}, + {R_028DF0_CB_COLOR6_FMASK_SLICE}, + {R_028DF4_CB_COLOR6_CLEAR_WORD0}, + {R_028DF8_CB_COLOR6_CLEAR_WORD1}, + {R_028DFC_CB_COLOR6_CLEAR_WORD2}, + {R_028E00_CB_COLOR6_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028E04_CB_COLOR7_BASE, REG_FLAG_NEED_BO, 0}, {R_028E08_CB_COLOR7_PITCH, 0, 0}, @@ -556,6 +676,14 @@ static const struct r600_reg cayman_context_reg_list[] = { {R_028E14_CB_COLOR7_INFO, REG_FLAG_NEED_BO, 0}, {R_028E18_CB_COLOR7_ATTRIB, REG_FLAG_NEED_BO, 0}, {R_028E1C_CB_COLOR7_DIM, 0, 0}, + {R_028E20_CB_COLOR7_CMASK, REG_FLAG_NEED_BO}, + {R_028E24_CB_COLOR7_CMASK_SLICE}, + {R_028E28_CB_COLOR7_FMASK, REG_FLAG_NEED_BO}, + {R_028E2C_CB_COLOR7_FMASK_SLICE}, + {R_028E30_CB_COLOR7_CLEAR_WORD0}, + {R_028E34_CB_COLOR7_CLEAR_WORD1}, + {R_028E38_CB_COLOR7_CLEAR_WORD2}, + {R_028E3C_CB_COLOR7_CLEAR_WORD3}, {GROUP_FORCE_NEW_BLOCK, 0, 0}, {R_028E40_CB_COLOR8_BASE, REG_FLAG_NEED_BO, 0}, {R_028E44_CB_COLOR8_PITCH, 0, 0}, diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c index e12706e856b..35bc391fe80 100644 --- a/src/gallium/drivers/r600/evergreen_state.c +++ b/src/gallium/drivers/r600/evergreen_state.c @@ -1250,8 +1250,8 @@ void evergreen_init_color_surface(struct r600_context *rctx, unsigned pitch, slice; unsigned color_info, color_attrib, color_dim = 0; unsigned format, swap, ntype, endian; - uint64_t offset; - unsigned tile_type, macro_aspect, tile_split, bankh, bankw, nbanks; + uint64_t offset, base_offset; + unsigned tile_type, macro_aspect, tile_split, bankh, bankw, fmask_bankh, nbanks; const struct util_format_description *desc; int i; bool blend_clamp = 0, blend_bypass = 0; @@ -1296,10 +1296,12 @@ void evergreen_init_color_surface(struct r600_context *rctx, macro_aspect = rtex->surface.mtilea; bankw = rtex->surface.bankw; bankh = rtex->surface.bankh; + fmask_bankh = rtex->fmask_bank_height; tile_split = eg_tile_split(tile_split); macro_aspect = eg_macro_tile_aspect(macro_aspect); bankw = eg_bank_wh(bankw); bankh = eg_bank_wh(bankh); + fmask_bankh = eg_bank_wh(fmask_bankh); /* 128 bit formats require tile type = 1 */ if (rscreen->chip_class == CAYMAN) { @@ -1319,7 +1321,8 @@ void evergreen_init_color_surface(struct r600_context *rctx, S_028C74_BANK_WIDTH(bankw) | S_028C74_BANK_HEIGHT(bankh) | S_028C74_MACRO_TILE_ASPECT(macro_aspect) | - S_028C74_NON_DISP_TILING_ORDER(tile_type); + S_028C74_NON_DISP_TILING_ORDER(tile_type) | + S_028C74_FMASK_BANK_HEIGHT(fmask_bankh); ntype = V_028C70_NUMBER_UNORM; if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB) @@ -1393,11 +1396,14 @@ void evergreen_init_color_surface(struct r600_context *rctx, surf->export_16bpc = true; } - offset += r600_resource_va(rctx->context.screen, pipe_tex); - offset >>= 8; + if (rtex->fmask_size && rtex->cmask_size) { + color_info |= S_028C70_COMPRESSION(1) | S_028C70_FAST_CLEAR(1); + } + + base_offset = r600_resource_va(rctx->context.screen, pipe_tex); /* XXX handle enabling of CB beyond BASE8 which has different offset */ - surf->cb_color_base = offset; + surf->cb_color_base = (base_offset + offset) >> 8; surf->cb_color_dim = color_dim; surf->cb_color_info = color_info; surf->cb_color_pitch = S_028C64_PITCH_TILE_MAX(pitch); @@ -1409,6 +1415,15 @@ void evergreen_init_color_surface(struct r600_context *rctx, S_028C6C_SLICE_MAX(surf->base.u.tex.last_layer); } surf->cb_color_attrib = color_attrib; + if (rtex->fmask_size && rtex->cmask_size) { + surf->cb_color_fmask = (base_offset + rtex->fmask_offset) >> 8; + surf->cb_color_cmask = (base_offset + rtex->cmask_offset) >> 8; + } else { + surf->cb_color_fmask = surf->cb_color_base; + surf->cb_color_cmask = surf->cb_color_base; + } + surf->cb_color_fmask_slice = S_028C88_TILE_MAX(slice); + surf->cb_color_cmask_slice = S_028C80_TILE_MAX(rtex->cmask_slice_tile_max); surf->color_initialized = true; } @@ -1631,6 +1646,7 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, struct r600_pipe_state *rstate = CALLOC_STRUCT(r600_pipe_state); struct r600_surface *surf; struct r600_resource *res; + struct r600_texture *rtex; uint32_t tl, br, i, nr_samples; if (rstate == NULL) @@ -1648,10 +1664,12 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, rctx->nr_cbufs = state->nr_cbufs; rctx->cb0_is_integer = state->nr_cbufs && util_format_is_pure_integer(state->cbufs[0]->format); + rctx->compressed_cb_mask = 0; for (i = 0; i < state->nr_cbufs; i++) { surf = (struct r600_surface*)state->cbufs[i]; res = (struct r600_resource*)surf->base.texture; + rtex = (struct r600_texture*)res; if (!surf->color_initialized) { evergreen_init_color_surface(rctx, surf); @@ -1675,6 +1693,18 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, surf->cb_color_view); r600_pipe_state_add_reg_bo(rstate, R_028C74_CB_COLOR0_ATTRIB + i * 0x3C, surf->cb_color_attrib, res, RADEON_USAGE_READWRITE); + r600_pipe_state_add_reg_bo(rstate, R_028C7C_CB_COLOR0_CMASK + i * 0x3c, + surf->cb_color_cmask, res, RADEON_USAGE_READWRITE); + r600_pipe_state_add_reg(rstate, R_028C80_CB_COLOR0_CMASK_SLICE + i * 0x3c, + surf->cb_color_cmask_slice); + r600_pipe_state_add_reg_bo(rstate, R_028C84_CB_COLOR0_FMASK + i * 0x3c, + surf->cb_color_fmask, res, RADEON_USAGE_READWRITE); + r600_pipe_state_add_reg(rstate, R_028C88_CB_COLOR0_FMASK_SLICE + i * 0x3c, + surf->cb_color_fmask_slice); + + if (rtex->fmask_size && rtex->cmask_size) { + rctx->compressed_cb_mask |= 1 << i; + } } /* set CB_COLOR1_INFO for possible dual-src blending */ if (i == 1 && !((struct r600_texture*)res)->is_rat) { @@ -3080,6 +3110,18 @@ void *evergreen_create_resolve_blend(struct r600_context *rctx) return rstate; } +void *evergreen_create_decompress_blend(struct r600_context *rctx) +{ + struct pipe_blend_state blend; + struct r600_pipe_state *rstate; + + memset(&blend, 0, sizeof(blend)); + blend.independent_blend_enable = true; + blend.rt[0].colormask = 0xf; + rstate = evergreen_create_blend_state_mode(&rctx->context, &blend, V_028808_CB_DECOMPRESS); + return rstate; +} + void *evergreen_create_db_flush_dsa(struct r600_context *rctx) { struct pipe_depth_stencil_alpha_state dsa = {{0}}; diff --git a/src/gallium/drivers/r600/evergreend.h b/src/gallium/drivers/r600/evergreend.h index d1b3cd33bb7..0c56aa6f693 100644 --- a/src/gallium/drivers/r600/evergreend.h +++ b/src/gallium/drivers/r600/evergreend.h @@ -377,6 +377,7 @@ #define S_028C74_BANK_WIDTH(x) (((x) & 0x3) << 13) #define S_028C74_BANK_HEIGHT(x) (((x) & 0x3) << 16) #define S_028C74_MACRO_TILE_ASPECT(x) (((x) & 0x3) << 19) +#define S_028C74_FMASK_BANK_HEIGHT(x) (((x) & 0x3) << 22) #define R_028C78_CB_COLOR0_DIM 0x028C78 #define S_028C78_WIDTH_MAX(x) (((x) & 0xFFFF) << 0) @@ -386,15 +387,6 @@ #define G_028C78_HEIGHT_MAX(x) (((x) >> 16) & 0xFFFF) #define C_028C78_HEIGHT_MAX 0x0000FFFF -#define R_028C7C_CB_COLOR0_CMASK 0x028C7C -#define R_028C80_CB_COLOR0_CMASK_SLICE 0x028C80 -#define R_028C84_CB_COLOR0_FMASK 0x028C84 -#define R_028C88_CB_COLOR0_FMASK_SLICE 0x028C88 - -#define R_028C8C_CB_COLOR0_CLEAR_WORD0 0x028C8C -#define R_028C90_CB_COLOR0_CLEAR_WORD1 0x028C90 -#define R_028C94_CB_COLOR0_CLEAR_WORD2 0x028C94 -#define R_028C98_CB_COLOR0_CLEAR_WORD3 0x028C98 /* alpha same */ #define R_028410_SX_ALPHA_TEST_CONTROL 0x028410 @@ -1978,6 +1970,16 @@ #define S_028C6C_SLICE_MAX(x) (((x) & 0x7FF) << 13) #define G_028C6C_SLICE_MAX(x) (((x) >> 13) & 0x7FF) #define C_028C6C_SLICE_MAX 0xFF001FFF +#define R_028C7C_CB_COLOR0_CMASK 0x028C7C +#define R_028C80_CB_COLOR0_CMASK_SLICE 0x028C80 +#define S_028C80_TILE_MAX(x) (((x) & 0x3FFF) << 0) +#define R_028C84_CB_COLOR0_FMASK 0x028C84 +#define R_028C88_CB_COLOR0_FMASK_SLICE 0x028C88 +#define S_028C88_TILE_MAX(x) (((x) & 0x3FFFFF) << 0) +#define R_028C8C_CB_COLOR0_CLEAR_WORD0 0x028C8C +#define R_028C90_CB_COLOR0_CLEAR_WORD1 0x028C90 +#define R_028C94_CB_COLOR0_CLEAR_WORD2 0x028C94 +#define R_028C98_CB_COLOR0_CLEAR_WORD3 0x028C98 #define R_028C9C_CB_COLOR1_BASE 0x00028C9C #define R_028CA0_CB_COLOR1_PITCH 0x00028CA0 #define R_028CA4_CB_COLOR1_SLICE 0x00028CA4 @@ -1985,6 +1987,14 @@ #define R_028CAC_CB_COLOR1_INFO 0x00028CAC #define R_028CB0_CB_COLOR1_ATTRIB 0x00028CB0 #define R_028CB4_CB_COLOR1_DIM 0x00028CB4 +#define R_028CB8_CB_COLOR1_CMASK 0x028CB8 +#define R_028CBC_CB_COLOR1_CMASK_SLICE 0x028CBC +#define R_028CC0_CB_COLOR1_FMASK 0x028CC0 +#define R_028CC4_CB_COLOR1_FMASK_SLICE 0x028CC4 +#define R_028CC8_CB_COLOR1_CLEAR_WORD0 0x028CC8 +#define R_028CCC_CB_COLOR1_CLEAR_WORD1 0x028CCC +#define R_028CD0_CB_COLOR1_CLEAR_WORD2 0x028CD0 +#define R_028CD4_CB_COLOR1_CLEAR_WORD3 0x028CD4 #define R_028CD8_CB_COLOR2_BASE 0x00028CD8 #define R_028CDC_CB_COLOR2_PITCH 0x00028CDC #define R_028CE0_CB_COLOR2_SLICE 0x00028CE0 @@ -1992,6 +2002,14 @@ #define R_028CE8_CB_COLOR2_INFO 0x00028CE8 #define R_028CEC_CB_COLOR2_ATTRIB 0x00028CEC #define R_028CF0_CB_COLOR2_DIM 0x00028CF0 +#define R_028CF4_CB_COLOR2_CMASK 0x028CF4 +#define R_028CF8_CB_COLOR2_CMASK_SLICE 0x028CF8 +#define R_028CFC_CB_COLOR2_FMASK 0x028CFC +#define R_028D00_CB_COLOR2_FMASK_SLICE 0x028D00 +#define R_028D04_CB_COLOR2_CLEAR_WORD0 0x028D04 +#define R_028D08_CB_COLOR2_CLEAR_WORD1 0x028D08 +#define R_028D0C_CB_COLOR2_CLEAR_WORD2 0x028D0C +#define R_028D10_CB_COLOR2_CLEAR_WORD3 0x028D10 #define R_028D14_CB_COLOR3_BASE 0x00028D14 #define R_028D18_CB_COLOR3_PITCH 0x00028D18 #define R_028D1C_CB_COLOR3_SLICE 0x00028D1C @@ -1999,6 +2017,14 @@ #define R_028D24_CB_COLOR3_INFO 0x00028D24 #define R_028D28_CB_COLOR3_ATTRIB 0x00028D28 #define R_028D2C_CB_COLOR3_DIM 0x00028D2C +#define R_028D30_CB_COLOR3_CMASK 0x028D30 +#define R_028D34_CB_COLOR3_CMASK_SLICE 0x028D34 +#define R_028D38_CB_COLOR3_FMASK 0x028D38 +#define R_028D3C_CB_COLOR3_FMASK_SLICE 0x028D3C +#define R_028D40_CB_COLOR3_CLEAR_WORD0 0x028D40 +#define R_028D44_CB_COLOR3_CLEAR_WORD1 0x028D44 +#define R_028D48_CB_COLOR3_CLEAR_WORD2 0x028D48 +#define R_028D4C_CB_COLOR3_CLEAR_WORD3 0x028D4C #define R_028D50_CB_COLOR4_BASE 0x00028D50 #define R_028D54_CB_COLOR4_PITCH 0x00028D54 #define R_028D58_CB_COLOR4_SLICE 0x00028D58 @@ -2006,6 +2032,14 @@ #define R_028D60_CB_COLOR4_INFO 0x00028D60 #define R_028D64_CB_COLOR4_ATTRIB 0x00028D64 #define R_028D68_CB_COLOR4_DIM 0x00028D68 +#define R_028D6C_CB_COLOR4_CMASK 0x028D6C +#define R_028D70_CB_COLOR4_CMASK_SLICE 0x028D70 +#define R_028D74_CB_COLOR4_FMASK 0x028D74 +#define R_028D78_CB_COLOR4_FMASK_SLICE 0x028D78 +#define R_028D7C_CB_COLOR4_CLEAR_WORD0 0x028D7C +#define R_028D80_CB_COLOR4_CLEAR_WORD1 0x028D80 +#define R_028D84_CB_COLOR4_CLEAR_WORD2 0x028D84 +#define R_028D88_CB_COLOR4_CLEAR_WORD3 0x028D88 #define R_028D8C_CB_COLOR5_BASE 0x00028D8C #define R_028D90_CB_COLOR5_PITCH 0x00028D90 #define R_028D94_CB_COLOR5_SLICE 0x00028D94 @@ -2013,6 +2047,14 @@ #define R_028D9C_CB_COLOR5_INFO 0x00028D9C #define R_028DA0_CB_COLOR5_ATTRIB 0x00028DA0 #define R_028DA4_CB_COLOR5_DIM 0x00028DA4 +#define R_028DA8_CB_COLOR5_CMASK 0x028DA8 +#define R_028DAC_CB_COLOR5_CMASK_SLICE 0x028DAC +#define R_028DB0_CB_COLOR5_FMASK 0x028DB0 +#define R_028DB4_CB_COLOR5_FMASK_SLICE 0x028DB4 +#define R_028DB8_CB_COLOR5_CLEAR_WORD0 0x028DB8 +#define R_028DBC_CB_COLOR5_CLEAR_WORD1 0x028DBC +#define R_028DC0_CB_COLOR5_CLEAR_WORD2 0x028DC0 +#define R_028DC4_CB_COLOR5_CLEAR_WORD3 0x028DC4 #define R_028DC8_CB_COLOR6_BASE 0x00028DC8 #define R_028DCC_CB_COLOR6_PITCH 0x00028DCC #define R_028DD0_CB_COLOR6_SLICE 0x00028DD0 @@ -2020,6 +2062,14 @@ #define R_028DD8_CB_COLOR6_INFO 0x00028DD8 #define R_028DDC_CB_COLOR6_ATTRIB 0x00028DDC #define R_028DE0_CB_COLOR6_DIM 0x00028DE0 +#define R_028DE4_CB_COLOR6_CMASK 0x028DE4 +#define R_028DE8_CB_COLOR6_CMASK_SLICE 0x028DE8 +#define R_028DEC_CB_COLOR6_FMASK 0x028DEC +#define R_028DF0_CB_COLOR6_FMASK_SLICE 0x028DF0 +#define R_028DF4_CB_COLOR6_CLEAR_WORD0 0x028DF4 +#define R_028DF8_CB_COLOR6_CLEAR_WORD1 0x028DF8 +#define R_028DFC_CB_COLOR6_CLEAR_WORD2 0x028DFC +#define R_028E00_CB_COLOR6_CLEAR_WORD3 0x028E00 #define R_028E04_CB_COLOR7_BASE 0x00028E04 #define R_028E08_CB_COLOR7_PITCH 0x00028E08 #define R_028E0C_CB_COLOR7_SLICE 0x00028E0C @@ -2027,6 +2077,14 @@ #define R_028E14_CB_COLOR7_INFO 0x00028E14 #define R_028E18_CB_COLOR7_ATTRIB 0x00028E18 #define R_028E1C_CB_COLOR7_DIM 0x00028E1C +#define R_028E20_CB_COLOR7_CMASK 0x028E20 +#define R_028E24_CB_COLOR7_CMASK_SLICE 0x028E24 +#define R_028E28_CB_COLOR7_FMASK 0x028E28 +#define R_028E2C_CB_COLOR7_FMASK_SLICE 0x028E2C +#define R_028E30_CB_COLOR7_CLEAR_WORD0 0x028E30 +#define R_028E34_CB_COLOR7_CLEAR_WORD1 0x028E34 +#define R_028E38_CB_COLOR7_CLEAR_WORD2 0x028E38 +#define R_028E3C_CB_COLOR7_CLEAR_WORD3 0x028E3C #define R_028E40_CB_COLOR8_BASE 0x00028E40 #define R_028E44_CB_COLOR8_PITCH 0x00028E44 #define R_028E48_CB_COLOR8_SLICE 0x00028E48 diff --git a/src/gallium/drivers/r600/r600_blit.c b/src/gallium/drivers/r600/r600_blit.c index c4ad0f20ed4..e9ed0740e65 100644 --- a/src/gallium/drivers/r600/r600_blit.c +++ b/src/gallium/drivers/r600/r600_blit.c @@ -243,6 +243,76 @@ void r600_decompress_depth_textures(struct r600_context *rctx, } } +static void r600_blit_decompress_color(struct pipe_context *ctx, + struct r600_texture *rtex, + unsigned first_level, unsigned last_level, + unsigned first_layer, unsigned last_layer) +{ + struct r600_context *rctx = (struct r600_context *)ctx; + unsigned layer, level, checked_last_layer, max_layer; + + if (!rtex->dirty_level_mask) + return; + + for (level = first_level; level <= last_level; level++) { + if (!(rtex->dirty_level_mask & (1 << level))) + continue; + + /* The smaller the mipmap level, the less layers there are + * as far as 3D textures are concerned. */ + max_layer = u_max_layer(&rtex->resource.b.b, level); + checked_last_layer = last_layer < max_layer ? last_layer : max_layer; + + for (layer = first_layer; layer <= checked_last_layer; layer++) { + struct pipe_surface *cbsurf, surf_tmpl; + + surf_tmpl.format = rtex->resource.b.b.format; + surf_tmpl.u.tex.level = level; + surf_tmpl.u.tex.first_layer = layer; + surf_tmpl.u.tex.last_layer = layer; + surf_tmpl.usage = PIPE_BIND_RENDER_TARGET; + cbsurf = ctx->create_surface(ctx, &rtex->resource.b.b, &surf_tmpl); + + r600_blitter_begin(ctx, R600_DECOMPRESS); + util_blitter_custom_color(rctx->blitter, cbsurf, + rctx->custom_blend_decompress); + r600_blitter_end(ctx); + + pipe_surface_reference(&cbsurf, NULL); + } + + /* The texture will always be dirty if some layers or samples aren't flushed. + * I don't think this case occurs often though. */ + if (first_layer == 0 && last_layer == max_layer) { + rtex->dirty_level_mask &= ~(1 << level); + } + } +} + +void r600_decompress_color_textures(struct r600_context *rctx, + struct r600_samplerview_state *textures) +{ + unsigned i; + unsigned mask = textures->compressed_colortex_mask; + + while (mask) { + struct pipe_sampler_view *view; + struct r600_texture *tex; + + i = u_bit_scan(&mask); + + view = &textures->views[i]->base; + assert(view); + + tex = (struct r600_texture *)view->texture; + assert(tex->cmask_size && tex->fmask_size); + + r600_blit_decompress_color(&rctx->context, tex, + view->u.tex.first_level, view->u.tex.last_level, + 0, u_max_layer(&tex->resource.b.b, view->u.tex.first_level)); + } +} + static void r600_copy_first_sample(struct pipe_context *ctx, const struct pipe_resolve_info *info) { @@ -262,6 +332,11 @@ static void r600_copy_first_sample(struct pipe_context *ctx, info->src.layer, info->src.layer, 0, 0); } + if (rsrc->fmask_size && rsrc->cmask_size) { + r600_blit_decompress_color(ctx, rsrc, + 0, 0, + info->src.layer, info->src.layer); + } /* this is correct for upside-down blits too */ u_box_2d(info->src.x0, @@ -601,6 +676,10 @@ static void r600_resource_copy_region(struct pipe_context *ctx, src_box->z, src_box->z + src_box->depth - 1, 0, u_max_sample(src)); } + if (rsrc->fmask_size && rsrc->cmask_size) { + r600_blit_decompress_color(ctx, rsrc, src_level, src_level, + src_box->z, src_box->z + src_box->depth - 1); + } restore_orig[0] = restore_orig[1] = FALSE; diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c index 7936c475bb6..0d489100dfb 100644 --- a/src/gallium/drivers/r600/r600_pipe.c +++ b/src/gallium/drivers/r600/r600_pipe.c @@ -168,6 +168,9 @@ static void r600_destroy_context(struct pipe_context *context) if (rctx->custom_blend_resolve) { rctx->context.delete_blend_state(&rctx->context, rctx->custom_blend_resolve); } + if (rctx->custom_blend_decompress) { + rctx->context.delete_blend_state(&rctx->context, rctx->custom_blend_decompress); + } util_unreference_framebuffer_state(&rctx->framebuffer); r600_context_fini(rctx); @@ -262,6 +265,7 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void goto fail; rctx->custom_dsa_flush = evergreen_create_db_flush_dsa(rctx); rctx->custom_blend_resolve = evergreen_create_resolve_blend(rctx); + rctx->custom_blend_decompress = evergreen_create_decompress_blend(rctx); rctx->has_vertex_cache = !(rctx->family == CHIP_CEDAR || rctx->family == CHIP_PALM || rctx->family == CHIP_SUMO || diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h index 0aff0f680d8..65725b6da62 100644 --- a/src/gallium/drivers/r600/r600_pipe.h +++ b/src/gallium/drivers/r600/r600_pipe.h @@ -268,6 +268,7 @@ struct r600_samplerview_state { uint32_t enabled_mask; uint32_t dirty_mask; uint32_t compressed_depthtex_mask; /* which textures are depth */ + uint32_t compressed_colortex_mask; }; struct r600_textures_info { @@ -327,12 +328,14 @@ struct r600_context { unsigned r6xx_num_clause_temp_gprs; void *custom_dsa_flush; void *custom_blend_resolve; + void *custom_blend_decompress; struct r600_screen *screen; struct radeon_winsys *ws; struct r600_pipe_state *states[R600_PIPE_NSTATES]; struct r600_vertex_element *vertex_elements; struct pipe_framebuffer_state framebuffer; + unsigned compressed_cb_mask; unsigned compute_cb_target_mask; unsigned db_shader_control; unsigned pa_sc_line_stipple; @@ -481,6 +484,7 @@ void evergreen_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader void evergreen_fetch_shader(struct pipe_context *ctx, struct r600_vertex_element *ve); void *evergreen_create_db_flush_dsa(struct r600_context *rctx); void *evergreen_create_resolve_blend(struct r600_context *rctx); +void *evergreen_create_decompress_blend(struct r600_context *rctx); void evergreen_polygon_offset_update(struct r600_context *rctx); boolean evergreen_is_format_supported(struct pipe_screen *screen, enum pipe_format format, @@ -504,6 +508,8 @@ void r600_blit_decompress_depth(struct pipe_context *ctx, unsigned first_sample, unsigned last_sample); void r600_decompress_depth_textures(struct r600_context *rctx, struct r600_samplerview_state *textures); +void r600_decompress_color_textures(struct r600_context *rctx, + struct r600_samplerview_state *textures); /* r600_buffer.c */ bool r600_init_resource(struct r600_screen *rscreen, diff --git a/src/gallium/drivers/r600/r600_resource.h b/src/gallium/drivers/r600/r600_resource.h index 7beab0fa110..dbb958e6925 100644 --- a/src/gallium/drivers/r600/r600_resource.h +++ b/src/gallium/drivers/r600/r600_resource.h @@ -55,6 +55,11 @@ struct r600_texture { struct r600_texture *flushed_depth_texture; boolean is_flushing_texture; struct radeon_surface surface; + + /* FMASK and CMASK can only be used with MSAA textures for now. + * MSAA textures cannot have mipmaps. */ + unsigned fmask_offset, fmask_size, fmask_bank_height; + unsigned cmask_offset, cmask_size, cmask_slice_tile_max; }; #define R600_TEX_IS_TILED(tex, level) ((tex)->array_mode[level] != V_038000_ARRAY_LINEAR_GENERAL && (tex)->array_mode[level] != V_038000_ARRAY_LINEAR_ALIGNED) @@ -74,12 +79,14 @@ struct r600_surface { unsigned cb_color_base; unsigned cb_color_view; unsigned cb_color_size; /* R600 only */ - unsigned cb_color_frag; /* R600 only */ - unsigned cb_color_tile; /* R600 only */ unsigned cb_color_dim; /* EG only */ unsigned cb_color_pitch; /* EG only */ unsigned cb_color_slice; /* EG only */ unsigned cb_color_attrib; /* EG only */ + unsigned cb_color_fmask; /* CB_COLORn_FMASK (EG) or CB_COLORn_FRAG (r600) */ + unsigned cb_color_fmask_slice; /* EG only */ + unsigned cb_color_cmask; /* CB_COLORn_CMASK (EG) or CB_COLORn_TILE (r600) */ + unsigned cb_color_cmask_slice; /* EG only */ /* DB registers. */ unsigned db_depth_info; /* DB_Z_INFO (EG) or DB_DEPTH_INFO (r600) */ diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c index d06c25000f1..8f6b0cd988b 100644 --- a/src/gallium/drivers/r600/r600_state.c +++ b/src/gallium/drivers/r600/r600_state.c @@ -1426,9 +1426,9 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_028080_CB_COLOR0_VIEW + i * 4, surf->cb_color_view); r600_pipe_state_add_reg_bo(rstate, R_0280E0_CB_COLOR0_FRAG + i * 4, - surf->cb_color_frag, res, RADEON_USAGE_READWRITE); + surf->cb_color_fmask, res, RADEON_USAGE_READWRITE); r600_pipe_state_add_reg_bo(rstate, R_0280C0_CB_COLOR0_TILE + i * 4, - surf->cb_color_tile, res, RADEON_USAGE_READWRITE); + surf->cb_color_cmask, res, RADEON_USAGE_READWRITE); } /* set CB_COLOR1_INFO for possible dual-src blending */ if (i == 1) { diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c index 5895fefc14d..77c3e20599e 100644 --- a/src/gallium/drivers/r600/r600_state_common.c +++ b/src/gallium/drivers/r600/r600_state_common.c @@ -628,6 +628,12 @@ void r600_set_sampler_views(struct pipe_context *pipe, dst->views.compressed_depthtex_mask &= ~(1 << i); } + if (rtex->cmask_size && rtex->fmask_size) { + dst->views.compressed_colortex_mask |= 1 << i; + } else { + dst->views.compressed_colortex_mask &= ~(1 << i); + } + /* Changing from array to non-arrays textures and vice * versa requires updating TEX_ARRAY_OVERRIDE on R6xx-R7xx. */ if (rctx->chip_class <= R700 && @@ -649,6 +655,7 @@ void r600_set_sampler_views(struct pipe_context *pipe, dst->views.enabled_mask |= new_mask; dst->views.dirty_mask |= new_mask; dst->views.compressed_depthtex_mask &= dst->views.enabled_mask; + dst->views.compressed_colortex_mask &= dst->views.enabled_mask; r600_sampler_views_dirty(rctx, &dst->views); } @@ -1046,6 +1053,12 @@ static void r600_update_derived_state(struct r600_context *rctx) if (rctx->ps_samplers.views.compressed_depthtex_mask) { r600_decompress_depth_textures(rctx, &rctx->ps_samplers.views); } + if (rctx->vs_samplers.views.compressed_colortex_mask) { + r600_decompress_color_textures(rctx, &rctx->vs_samplers.views); + } + if (rctx->ps_samplers.views.compressed_colortex_mask) { + r600_decompress_color_textures(rctx, &rctx->ps_samplers.views); + } } r600_shader_select(ctx, rctx->ps_shader, &ps_dirty); @@ -1262,6 +1275,20 @@ void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *dinfo) rtex->dirty_level_mask |= 1 << surf->u.tex.level; } + if (rctx->compressed_cb_mask) { + struct pipe_surface *surf; + struct r600_texture *rtex; + unsigned mask = rctx->compressed_cb_mask; + + do { + unsigned i = u_bit_scan(&mask); + surf = rctx->framebuffer.cbufs[i]; + rtex = (struct r600_texture*)surf->texture; + + rtex->dirty_level_mask |= 1 << surf->u.tex.level; + + } while (mask); + } pipe_resource_reference(&ib.buffer, NULL); } diff --git a/src/gallium/drivers/r600/r600_texture.c b/src/gallium/drivers/r600/r600_texture.c index 90a834ec634..acb77d5c4dd 100644 --- a/src/gallium/drivers/r600/r600_texture.c +++ b/src/gallium/drivers/r600/r600_texture.c @@ -252,6 +252,97 @@ static const struct u_resource_vtbl r600_texture_vtbl = NULL /* transfer_inline_write */ }; +static void r600_texture_allocate_fmask(struct r600_screen *rscreen, + struct r600_texture *rtex) +{ + /* FMASK is allocated pretty much like an ordinary texture. + * Here we use bpe in the units of bits, not bytes. */ + struct radeon_surface fmask = rtex->surface; + unsigned nr_samples = rtex->resource.b.b.nr_samples; + + switch (nr_samples) { + case 2: + /* This should be 8,1, but we should set nsamples > 1 + * for the allocator to treat it as a multisample surface. + * Let's set 4,2 then. */ + case 4: + fmask.bpe = 4; + fmask.nsamples = 2; + break; + case 8: + fmask.bpe = 8; + fmask.nsamples = 4; + break; + case 16: + fmask.bpe = 16; + fmask.nsamples = 4; + break; + default: + R600_ERR("Invalid sample count for FMASK allocation.\n"); + return; + } + + if (rscreen->chip_class >= EVERGREEN) { + fmask.bankh = nr_samples <= 4 ? 4 : 1; + } + + if (rscreen->ws->surface_init(rscreen->ws, &fmask)) { + R600_ERR("Got error in surface_init while allocating FMASK.\n"); + return; + } + assert(fmask.level[0].mode == RADEON_SURF_MODE_2D); + + /* Reserve space for FMASK while converting bits back to bytes. */ + rtex->fmask_bank_height = fmask.bankh; + rtex->fmask_offset = align(rtex->size, MAX2(256, fmask.bo_alignment)); + rtex->fmask_size = (fmask.bo_size + 7) / 8; + rtex->size = rtex->fmask_offset + rtex->fmask_size; +#if 0 + printf("FMASK width=%u, height=%i, bits=%u, size=%u\n", + fmask.npix_x, fmask.npix_y, fmask.bpe * fmask.nsamples, rtex->fmask_size); +#endif +} + +static void r600_texture_allocate_cmask(struct r600_screen *rscreen, + struct r600_texture *rtex) +{ + unsigned cmask_tile_width = 8; + unsigned cmask_tile_height = 8; + unsigned cmask_tile_elements = cmask_tile_width * cmask_tile_height; + unsigned element_bits = 4; + unsigned cmask_cache_bits = 1024; + unsigned num_pipes = rscreen->tiling_info.num_channels; + unsigned pipe_interleave_bytes = rscreen->tiling_info.group_bytes; + + unsigned elements_per_macro_tile = (cmask_cache_bits / element_bits) * num_pipes; + unsigned pixels_per_macro_tile = elements_per_macro_tile * cmask_tile_elements; + unsigned sqrt_pixels_per_macro_tile = sqrt(pixels_per_macro_tile); + unsigned macro_tile_width = util_next_power_of_two(sqrt_pixels_per_macro_tile); + unsigned macro_tile_height = pixels_per_macro_tile / macro_tile_width; + + unsigned pitch_elements = align(rtex->surface.npix_x, macro_tile_width); + unsigned height = align(rtex->surface.npix_y, macro_tile_height); + + unsigned base_align = num_pipes * pipe_interleave_bytes; + unsigned slice_bytes = + ((pitch_elements * height * element_bits + 7) / 8) / cmask_tile_elements; + unsigned size = rtex->surface.array_size * align(slice_bytes, base_align); + + assert(macro_tile_width % 128 == 0); + assert(macro_tile_height % 128 == 0); + + rtex->cmask_slice_tile_max = ((pitch_elements * height) / (128*128)) - 1; + rtex->cmask_offset = align(rtex->size, MAX2(256, base_align)); + rtex->cmask_size = size; + rtex->size = rtex->cmask_offset + rtex->cmask_size; +#if 0 + printf("CMASK: macro tile width = %u, macro tile height = %u, " + "pitch elements = %u, height = %u, slice tile max = %u\n", + macro_tile_width, macro_tile_height, pitch_elements, height, + rtex->cmask_slice_tile_max); +#endif +} + static struct r600_texture * r600_texture_create_object(struct pipe_screen *screen, const struct pipe_resource *base, @@ -287,6 +378,17 @@ r600_texture_create_object(struct pipe_screen *screen, return NULL; } + if (base->nr_samples > 1 && !rtex->is_depth && alloc_bo) { + r600_texture_allocate_fmask(rscreen, rtex); + r600_texture_allocate_cmask(rscreen, rtex); + } + + if (!rtex->is_depth && base->nr_samples > 1 && + (!rtex->fmask_size || !rtex->cmask_size)) { + FREE(rtex); + return NULL; + } + /* Now create the backing buffer. */ if (!buf && alloc_bo) { unsigned base_align = rtex->surface.bo_alignment; @@ -301,6 +403,13 @@ r600_texture_create_object(struct pipe_screen *screen, resource->cs_buf = rscreen->ws->buffer_get_cs_handle(buf); resource->domains = RADEON_DOMAIN_GTT | RADEON_DOMAIN_VRAM; } + + if (rtex->cmask_size) { + /* Initialize the cmask to 0xCC (= compressed state). */ + char *ptr = rscreen->ws->buffer_map(resource->cs_buf, NULL, PIPE_TRANSFER_WRITE); + memset(ptr + rtex->cmask_offset, 0xCC, rtex->cmask_size); + rscreen->ws->buffer_unmap(resource->cs_buf); + } return rtex; }