diff options
-rw-r--r-- | src/gallium/drivers/radeon/r600_texture.c | 14 | ||||
-rw-r--r-- | src/gallium/drivers/radeon/radeon_winsys.h | 6 | ||||
-rw-r--r-- | src/gallium/drivers/radeonsi/si_blit.c | 4 | ||||
-rw-r--r-- | src/gallium/drivers/radeonsi/si_descriptors.c | 4 | ||||
-rw-r--r-- | src/gallium/drivers/radeonsi/si_state.c | 2 | ||||
-rw-r--r-- | src/gallium/winsys/amdgpu/drm/amdgpu_surface.c | 4 |
6 files changed, 19 insertions, 15 deletions
diff --git a/src/gallium/drivers/radeon/r600_texture.c b/src/gallium/drivers/radeon/r600_texture.c index bc981da4bed..46281cbd0e5 100644 --- a/src/gallium/drivers/radeon/r600_texture.c +++ b/src/gallium/drivers/radeon/r600_texture.c @@ -73,8 +73,8 @@ bool r600_prepare_for_dma_blit(struct r600_common_context *rctx, * src: Use the 3D path. DCC decompression is expensive. * dst: Use the 3D path to compress the pixels with DCC. */ - if ((rsrc->dcc_offset && rsrc->surface.level[src_level].dcc_enabled) || - (rdst->dcc_offset && rdst->surface.level[dst_level].dcc_enabled)) + if ((rsrc->dcc_offset && src_level < rsrc->surface.num_dcc_levels) || + (rdst->dcc_offset && dst_level < rdst->surface.num_dcc_levels)) return false; /* CMASK as: @@ -940,7 +940,7 @@ void r600_print_texture_info(struct r600_texture *rtex, FILE *f) for (i = 0; i <= rtex->resource.b.b.last_level; i++) fprintf(f, " DCCLevel[%i]: enabled=%u, offset=%"PRIu64", " "fast_clear_size=%"PRIu64"\n", - i, rtex->surface.level[i].dcc_enabled, + i, i < rtex->surface.num_dcc_levels, rtex->surface.level[i].dcc_offset, rtex->surface.level[i].dcc_fast_clear_size); } @@ -1744,7 +1744,7 @@ void vi_dcc_disable_if_incompatible_format(struct r600_common_context *rctx, struct r600_texture *rtex = (struct r600_texture *)tex; if (rtex->dcc_offset && - rtex->surface.level[level].dcc_enabled && + level < rtex->surface.num_dcc_levels && !vi_dcc_formats_compatible(tex->format, view_format)) if (!r600_texture_disable_dcc(rctx, (struct r600_texture*)tex)) rctx->decompress_dcc(&rctx->b, rtex); @@ -2096,7 +2096,7 @@ static void vi_separate_dcc_try_enable(struct r600_common_context *rctx, if (!vi_should_enable_separate_dcc(tex)) return; /* stats show that DCC decompression is too expensive */ - assert(tex->surface.level[0].dcc_enabled); + assert(tex->surface.num_dcc_levels); assert(!tex->dcc_separate_buffer); r600_texture_discard_cmask(rctx->screen, tex); @@ -2311,7 +2311,7 @@ void vi_dcc_clear_level(struct r600_common_context *rctx, struct pipe_resource *dcc_buffer; uint64_t dcc_offset; - assert(rtex->dcc_offset && rtex->surface.level[level].dcc_enabled); + assert(rtex->dcc_offset && level < rtex->surface.num_dcc_levels); if (rtex->dcc_separate_buffer) { dcc_buffer = &rtex->dcc_separate_buffer->b.b; @@ -2483,7 +2483,7 @@ void evergreen_do_fast_color_clear(struct r600_common_context *rctx, } /* Try to clear DCC first, otherwise try CMASK. */ - if (tex->dcc_offset && tex->surface.level[0].dcc_enabled) { + if (tex->dcc_offset && tex->surface.num_dcc_levels) { uint32_t reset_value; bool clear_words_needed; diff --git a/src/gallium/drivers/radeon/radeon_winsys.h b/src/gallium/drivers/radeon/radeon_winsys.h index 2330cddf631..f5b9f105836 100644 --- a/src/gallium/drivers/radeon/radeon_winsys.h +++ b/src/gallium/drivers/radeon/radeon_winsys.h @@ -283,7 +283,6 @@ struct radeon_surf_level { uint16_t nblk_y; uint32_t pitch_bytes; enum radeon_surf_mode mode; - bool dcc_enabled; }; struct radeon_surf { @@ -291,6 +290,11 @@ struct radeon_surf { unsigned blk_w:4; unsigned blk_h:4; unsigned bpe:5; + /* Number of mipmap levels where DCC is enabled starting from level 0. + * Non-zero levels may be disabled due to alignment constraints, but not + * the first level. + */ + unsigned num_dcc_levels:4; uint32_t flags; /* These are return values. Some of them can be set by the caller, but diff --git a/src/gallium/drivers/radeonsi/si_blit.c b/src/gallium/drivers/radeonsi/si_blit.c index db41f565a94..0f46d71ddaf 100644 --- a/src/gallium/drivers/radeonsi/si_blit.c +++ b/src/gallium/drivers/radeonsi/si_blit.c @@ -429,7 +429,7 @@ static void si_blit_decompress_color(struct pipe_context *ctx, /* disable levels without DCC */ for (int i = first_level; i <= last_level; i++) { if (!rtex->dcc_offset || - !rtex->surface.level[i].dcc_enabled) + i >= rtex->surface.num_dcc_levels) level_mask &= ~(1 << i); } } else if (rtex->fmask.size) { @@ -1029,7 +1029,7 @@ static bool do_hardware_msaa_resolve(struct pipe_context *ctx, * This is still the fastest codepath even with this clear. */ if (dst->dcc_offset && - dst->surface.level[info->dst.level].dcc_enabled) { + info->dst.level < dst->surface.num_dcc_levels) { vi_dcc_clear_level(&sctx->b, dst, info->dst.level, 0xFFFFFFFF); dst->dirty_level_mask &= ~(1 << info->dst.level); diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c b/src/gallium/drivers/radeonsi/si_descriptors.c index 19cae65e75e..9358542ac78 100644 --- a/src/gallium/drivers/radeonsi/si_descriptors.c +++ b/src/gallium/drivers/radeonsi/si_descriptors.c @@ -394,7 +394,7 @@ void si_set_mutable_tex_desc_fields(struct r600_texture *tex, is_stencil)); state[4] |= S_008F20_PITCH(pitch - 1); - if (tex->dcc_offset && tex->surface.level[first_level].dcc_enabled) { + if (tex->dcc_offset && first_level < tex->surface.num_dcc_levels) { state[6] |= S_008F28_COMPRESSION_EN(1); state[7] = ((!tex->dcc_separate_buffer ? tex->resource.gpu_address : 0) + tex->dcc_offset + @@ -669,7 +669,7 @@ static void si_set_shader_image(struct si_context *ctx, unsigned level = view->u.tex.level; unsigned width, height, depth; bool uses_dcc = tex->dcc_offset && - tex->surface.level[level].dcc_enabled; + level < tex->surface.num_dcc_levels; assert(!tex->is_depth); assert(tex->fmask.size == 0); diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c index 0633b64de33..bf89d8bf1f2 100644 --- a/src/gallium/drivers/radeonsi/si_state.c +++ b/src/gallium/drivers/radeonsi/si_state.c @@ -2500,7 +2500,7 @@ static void si_emit_framebuffer_state(struct si_context *sctx, struct r600_atom cb_color_info = cb->cb_color_info | tex->cb_color_info; - if (tex->dcc_offset && cb->level_info->dcc_enabled) { + if (tex->dcc_offset && cb->base.u.tex.level < tex->surface.num_dcc_levels) { bool is_msaa_resolve_dst = state->cbufs[0] && state->cbufs[0]->texture->nr_samples > 1 && state->cbufs[1] == &cb->base && diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_surface.c b/src/gallium/winsys/amdgpu/drm/amdgpu_surface.c index 45edcc28734..8c57287f073 100644 --- a/src/gallium/winsys/amdgpu/drm/amdgpu_surface.c +++ b/src/gallium/winsys/amdgpu/drm/amdgpu_surface.c @@ -218,7 +218,6 @@ static int compute_level(struct amdgpu_winsys *ws, /* Clear DCC fields at the beginning. */ surf_level->dcc_offset = 0; - surf_level->dcc_enabled = false; /* The previous level's flag tells us if we can use DCC for this level. */ if (AddrSurfInfoIn->flags.dccCompatible && @@ -236,7 +235,7 @@ static int compute_level(struct amdgpu_winsys *ws, if (ret == ADDR_OK) { surf_level->dcc_offset = surf->dcc_size; surf_level->dcc_fast_clear_size = AddrDccOut->dccFastClearSize; - surf_level->dcc_enabled = true; + surf->num_dcc_levels = level + 1; surf->dcc_size = surf_level->dcc_offset + AddrDccOut->dccRamSize; surf->dcc_alignment = MAX2(surf->dcc_alignment, AddrDccOut->dccRamBaseAlign); } @@ -488,6 +487,7 @@ static int amdgpu_surface_init(struct radeon_winsys *rws, } } + surf->num_dcc_levels = 0; surf->surf_size = 0; surf->dcc_size = 0; surf->dcc_alignment = 1; |