X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;ds=sidebyside;f=src%2Fgallium%2Fdrivers%2Fradeonsi%2Fsi_blit.c;h=c3591a7e85a518706921d9df4fac40692b9ac8fd;hb=b9dad585e66b1031bdcbb148a19524ee2705baf7;hp=e02615fe6ca07b128f3f3c917c4495b20ee2e59d;hpb=d2261918202697febed0de18f66416e273001088;p=mesa.git diff --git a/src/gallium/drivers/radeonsi/si_blit.c b/src/gallium/drivers/radeonsi/si_blit.c index e02615fe6ca..c3591a7e85a 100644 --- a/src/gallium/drivers/radeonsi/si_blit.c +++ b/src/gallium/drivers/radeonsi/si_blit.c @@ -22,8 +22,8 @@ */ #include "si_pipe.h" -#include "util/u_blitter.h" #include "util/u_format.h" +#include "util/u_surface.h" enum si_blitter_op /* bitmask */ { @@ -57,10 +57,19 @@ static void si_blitter_begin(struct pipe_context *ctx, enum si_blitter_op op) util_blitter_save_rasterizer(sctx->blitter, sctx->queued.named.rasterizer); util_blitter_save_fragment_shader(sctx->blitter, sctx->ps_shader); util_blitter_save_geometry_shader(sctx->blitter, sctx->gs_shader); + util_blitter_save_tessctrl_shader(sctx->blitter, sctx->tcs_shader); + util_blitter_save_tesseval_shader(sctx->blitter, sctx->tes_shader); util_blitter_save_vertex_shader(sctx->blitter, sctx->vs_shader); util_blitter_save_vertex_elements(sctx->blitter, sctx->vertex_elements); - if (sctx->queued.named.viewport) { - util_blitter_save_viewport(sctx->blitter, &sctx->queued.named.viewport->viewport); + if (sctx->queued.named.sample_mask) { + util_blitter_save_sample_mask(sctx->blitter, + sctx->queued.named.sample_mask->sample_mask); + } + if (sctx->queued.named.viewport[0]) { + util_blitter_save_viewport(sctx->blitter, &sctx->queued.named.viewport[0]->viewport); + } + if (sctx->queued.named.scissor[0]) { + util_blitter_save_scissor(sctx->blitter, &sctx->queued.named.scissor[0]->scissor); } util_blitter_save_vertex_buffer_slot(sctx->blitter, sctx->vertex_buffer); util_blitter_save_so_targets(sctx->blitter, sctx->b.streamout.num_targets, @@ -71,12 +80,10 @@ static void si_blitter_begin(struct pipe_context *ctx, enum si_blitter_op op) if (op & SI_SAVE_TEXTURES) { util_blitter_save_fragment_sampler_states( - sctx->blitter, sctx->samplers[PIPE_SHADER_FRAGMENT].n_samplers, - (void**)sctx->samplers[PIPE_SHADER_FRAGMENT].samplers); + sctx->blitter, 2, + sctx->samplers[PIPE_SHADER_FRAGMENT].states.saved_states); - util_blitter_save_fragment_sampler_views(sctx->blitter, - util_last_bit(sctx->samplers[PIPE_SHADER_FRAGMENT].views.desc.enabled_mask & - ((1 << NUM_TEX_UNITS) - 1)), + util_blitter_save_fragment_sampler_views(sctx->blitter, 2, sctx->samplers[PIPE_SHADER_FRAGMENT].views.views); } @@ -110,7 +117,6 @@ static void si_blit_decompress_depth(struct pipe_context *ctx, unsigned layer, level, sample, checked_last_layer, max_layer, max_sample; float depth = 1.0f; const struct util_format_description *desc; - void **custom_dsa; struct r600_texture *flushed_depth_texture = staging ? staging : texture->flushed_depth_texture; @@ -120,20 +126,13 @@ static void si_blit_decompress_depth(struct pipe_context *ctx, max_sample = u_max_sample(&texture->resource.b.b); desc = util_format_description(flushed_depth_texture->resource.b.b.format); - switch (util_format_has_depth(desc) | util_format_has_stencil(desc) << 1) { - default: - assert(!"No depth or stencil to uncompress"); - return; - case 3: - custom_dsa = sctx->custom_dsa_flush_depth_stencil; - break; - case 2: - custom_dsa = sctx->custom_dsa_flush_stencil; - break; - case 1: - custom_dsa = sctx->custom_dsa_flush_depth; - break; - } + + if (util_format_has_depth(desc)) + sctx->dbcb_depth_copy_enabled = true; + if (util_format_has_stencil(desc)) + sctx->dbcb_stencil_copy_enabled = true; + + assert(sctx->dbcb_depth_copy_enabled || sctx->dbcb_stencil_copy_enabled); for (level = first_level; level <= last_level; level++) { if (!staging && !(texture->dirty_level_mask & (1 << level))) @@ -148,6 +147,9 @@ static void si_blit_decompress_depth(struct pipe_context *ctx, for (sample = first_sample; sample <= last_sample; sample++) { struct pipe_surface *zsurf, *cbsurf, surf_tmpl; + sctx->dbcb_copy_sample = sample; + sctx->db_render_state.dirty = true; + surf_tmpl.format = texture->resource.b.b.format; surf_tmpl.u.tex.level = level; surf_tmpl.u.tex.first_layer = layer; @@ -161,7 +163,7 @@ static void si_blit_decompress_depth(struct pipe_context *ctx, si_blitter_begin(ctx, SI_DECOMPRESS); util_blitter_custom_depth_stencil(sctx->blitter, zsurf, cbsurf, 1 << sample, - custom_dsa[sample], depth); + sctx->custom_dsa_flush, depth); si_blitter_end(ctx); pipe_surface_reference(&zsurf, NULL); @@ -177,6 +179,10 @@ static void si_blit_decompress_depth(struct pipe_context *ctx, texture->dirty_level_mask &= ~(1 << level); } } + + sctx->dbcb_depth_copy_enabled = false; + sctx->dbcb_stencil_copy_enabled = false; + sctx->db_render_state.dirty = true; } static void si_blit_decompress_depth_in_place(struct si_context *sctx, @@ -187,6 +193,9 @@ static void si_blit_decompress_depth_in_place(struct si_context *sctx, struct pipe_surface *zsurf, surf_tmpl = {{0}}; unsigned layer, max_layer, checked_last_layer, level; + sctx->db_inplace_flush_enabled = true; + sctx->db_render_state.dirty = true; + surf_tmpl.format = texture->resource.b.b.format; for (level = first_level; level <= last_level; level++) { @@ -208,7 +217,7 @@ static void si_blit_decompress_depth_in_place(struct si_context *sctx, si_blitter_begin(&sctx->b.b, SI_DECOMPRESS); util_blitter_custom_depth_stencil(sctx->blitter, zsurf, NULL, ~0, - sctx->custom_dsa_flush_inplace, + sctx->custom_dsa_flush, 1.0f); si_blitter_end(&sctx->b.b); @@ -221,23 +230,28 @@ static void si_blit_decompress_depth_in_place(struct si_context *sctx, texture->dirty_level_mask &= ~(1 << level); } } + + sctx->db_inplace_flush_enabled = false; + sctx->db_render_state.dirty = true; } void si_flush_depth_textures(struct si_context *sctx, struct si_textures_info *textures) { unsigned i; + unsigned mask = textures->depth_texture_mask; - for (i = 0; i < textures->n_views; ++i) { + while (mask) { struct pipe_sampler_view *view; struct r600_texture *tex; + i = u_bit_scan(&mask); + view = textures->views.views[i]; - if (!view) continue; + assert(view); tex = (struct r600_texture *)view->texture; - if (!tex->is_depth || tex->is_flushing_texture) - continue; + assert(tex->is_depth && !tex->is_flushing_texture); si_blit_decompress_depth_in_place(sctx, tex, view->u.tex.first_level, view->u.tex.last_level, @@ -321,6 +335,9 @@ static void si_clear(struct pipe_context *ctx, unsigned buffers, { struct si_context *sctx = (struct si_context *)ctx; struct pipe_framebuffer_state *fb = &sctx->framebuffer.state; + struct pipe_surface *zsbuf = fb->zsbuf; + struct r600_texture *zstex = + zsbuf ? (struct r600_texture*)zsbuf->texture : NULL; if (buffers & PIPE_CLEAR_COLOR) { evergreen_do_fast_color_clear(&sctx->b, fb, &sctx->framebuffer.atom, @@ -347,11 +364,35 @@ static void si_clear(struct pipe_context *ctx, unsigned buffers, } } + if (buffers & PIPE_CLEAR_DEPTH && + zstex && zstex->htile_buffer && + zsbuf->u.tex.level == 0 && + zsbuf->u.tex.first_layer == 0 && + zsbuf->u.tex.last_layer == util_max_layer(&zstex->resource.b.b, 0)) { + /* Need to disable EXPCLEAR temporarily if clearing + * to a new value. */ + if (zstex->depth_cleared && zstex->depth_clear_value != depth) { + sctx->db_depth_disable_expclear = true; + } + + zstex->depth_clear_value = depth; + sctx->framebuffer.atom.dirty = true; /* updates DB_DEPTH_CLEAR */ + sctx->db_depth_clear = true; + sctx->db_render_state.dirty = true; + } + si_blitter_begin(ctx, SI_CLEAR); util_blitter_clear(sctx->blitter, fb->width, fb->height, util_framebuffer_get_num_layers(fb), buffers, color, depth, stencil); si_blitter_end(ctx); + + if (sctx->db_depth_clear) { + sctx->db_depth_clear = false; + sctx->db_depth_disable_expclear = false; + zstex->depth_cleared = true; + sctx->db_render_state.dirty = true; + } } static void si_clear_render_target(struct pipe_context *ctx, @@ -416,189 +457,123 @@ struct texture_orig_info { unsigned npix0_y; }; -static void si_compressed_to_blittable(struct pipe_resource *tex, - unsigned level, - struct texture_orig_info *orig) -{ - struct r600_texture *rtex = (struct r600_texture*)tex; - unsigned pixsize = util_format_get_blocksize(rtex->resource.b.b.format); - int new_format; - int new_height, new_width; - - orig->format = tex->format; - orig->width0 = tex->width0; - orig->height0 = tex->height0; - orig->npix0_x = rtex->surface.level[0].npix_x; - orig->npix0_y = rtex->surface.level[0].npix_y; - orig->npix_x = rtex->surface.level[level].npix_x; - orig->npix_y = rtex->surface.level[level].npix_y; - - if (pixsize == 8) - new_format = PIPE_FORMAT_R16G16B16A16_UINT; /* 64-bit block */ - else - new_format = PIPE_FORMAT_R32G32B32A32_UINT; /* 128-bit block */ - - new_width = util_format_get_nblocksx(tex->format, orig->width0); - new_height = util_format_get_nblocksy(tex->format, orig->height0); - - tex->width0 = new_width; - tex->height0 = new_height; - tex->format = new_format; - rtex->surface.level[0].npix_x = util_format_get_nblocksx(orig->format, orig->npix0_x); - rtex->surface.level[0].npix_y = util_format_get_nblocksy(orig->format, orig->npix0_y); - rtex->surface.level[level].npix_x = util_format_get_nblocksx(orig->format, orig->npix_x); - rtex->surface.level[level].npix_y = util_format_get_nblocksy(orig->format, orig->npix_y); - - /* By dividing the dimensions by 4, we effectively decrement - * last_level by 2, therefore the last 2 mipmap levels disappear and - * aren't blittable. Note that the last 3 mipmap levels (4x4, 2x2, - * 1x1) have equal slice sizes, which is an important assumption - * for this to work. - * - * In order to make the last 2 mipmap levels blittable, we have to - * add the slice size of the last mipmap level to the texture - * address, so that even though the hw thinks it reads last_level-2, - * it will actually read last_level-1, and if we add the slice size*2, - * it will read last_level. That's how this workaround works. - */ - if (level > rtex->resource.b.b.last_level-2) - rtex->mipmap_shift = level - (rtex->resource.b.b.last_level-2); -} - -static void si_change_format(struct pipe_resource *tex, - unsigned level, - struct texture_orig_info *orig, - enum pipe_format format) -{ - struct r600_texture *rtex = (struct r600_texture*)tex; - - orig->format = tex->format; - orig->width0 = tex->width0; - orig->height0 = tex->height0; - orig->npix0_x = rtex->surface.level[0].npix_x; - orig->npix0_y = rtex->surface.level[0].npix_y; - orig->npix_x = rtex->surface.level[level].npix_x; - orig->npix_y = rtex->surface.level[level].npix_y; - - tex->format = format; -} - -static void si_reset_blittable_to_orig(struct pipe_resource *tex, - unsigned level, - struct texture_orig_info *orig) -{ - struct r600_texture *rtex = (struct r600_texture*)tex; - - tex->format = orig->format; - tex->width0 = orig->width0; - tex->height0 = orig->height0; - rtex->surface.level[0].npix_x = orig->npix0_x; - rtex->surface.level[0].npix_y = orig->npix0_y; - rtex->surface.level[level].npix_x = orig->npix_x; - rtex->surface.level[level].npix_y = orig->npix_y; - rtex->mipmap_shift = 0; -} - -static void si_resource_copy_region(struct pipe_context *ctx, - struct pipe_resource *dst, - unsigned dst_level, - unsigned dstx, unsigned dsty, unsigned dstz, - struct pipe_resource *src, - unsigned src_level, - const struct pipe_box *src_box) +void si_resource_copy_region(struct pipe_context *ctx, + struct pipe_resource *dst, + unsigned dst_level, + unsigned dstx, unsigned dsty, unsigned dstz, + struct pipe_resource *src, + unsigned src_level, + const struct pipe_box *src_box) { struct si_context *sctx = (struct si_context *)ctx; - struct r600_texture *rdst = (struct r600_texture*)dst; struct pipe_surface *dst_view, dst_templ; struct pipe_sampler_view src_templ, *src_view; - struct texture_orig_info orig_info[2]; + unsigned dst_width, dst_height, src_width0, src_height0; + unsigned src_force_level = 0; struct pipe_box sbox, dstbox; - boolean restore_orig[2]; - /* Fallback for buffers. */ + /* Handle buffers first. */ if (dst->target == PIPE_BUFFER && src->target == PIPE_BUFFER) { - si_copy_buffer(sctx, dst, src, dstx, src_box->x, src_box->width); + si_copy_buffer(sctx, dst, src, dstx, src_box->x, src_box->width, false); return; } - memset(orig_info, 0, sizeof(orig_info)); + assert(u_max_sample(dst) == u_max_sample(src)); /* The driver doesn't decompress resources automatically while * u_blitter is rendering. */ si_decompress_subresource(ctx, src, src_level, src_box->z, src_box->z + src_box->depth - 1); - restore_orig[0] = restore_orig[1] = FALSE; + dst_width = u_minify(dst->width0, dst_level); + dst_height = u_minify(dst->height0, dst_level); + src_width0 = src->width0; + src_height0 = src->height0; + + util_blitter_default_dst_texture(&dst_templ, dst, dst_level, dstz); + util_blitter_default_src_texture(&src_templ, src, src_level); if (util_format_is_compressed(src->format) && util_format_is_compressed(dst->format)) { - si_compressed_to_blittable(src, src_level, &orig_info[0]); - restore_orig[0] = TRUE; - sbox.x = util_format_get_nblocksx(orig_info[0].format, src_box->x); - sbox.y = util_format_get_nblocksy(orig_info[0].format, src_box->y); + unsigned blocksize = util_format_get_blocksize(src->format); + + if (blocksize == 8) + src_templ.format = PIPE_FORMAT_R16G16B16A16_UINT; /* 64-bit block */ + else + src_templ.format = PIPE_FORMAT_R32G32B32A32_UINT; /* 128-bit block */ + dst_templ.format = src_templ.format; + + dst_width = util_format_get_nblocksx(dst->format, dst_width); + dst_height = util_format_get_nblocksy(dst->format, dst_height); + src_width0 = util_format_get_nblocksx(src->format, src_width0); + src_height0 = util_format_get_nblocksy(src->format, src_height0); + + dstx = util_format_get_nblocksx(dst->format, dstx); + dsty = util_format_get_nblocksy(dst->format, dsty); + + sbox.x = util_format_get_nblocksx(src->format, src_box->x); + sbox.y = util_format_get_nblocksy(src->format, src_box->y); sbox.z = src_box->z; - sbox.width = util_format_get_nblocksx(orig_info[0].format, src_box->width); - sbox.height = util_format_get_nblocksy(orig_info[0].format, src_box->height); + sbox.width = util_format_get_nblocksx(src->format, src_box->width); + sbox.height = util_format_get_nblocksy(src->format, src_box->height); sbox.depth = src_box->depth; src_box = &sbox; - si_compressed_to_blittable(dst, dst_level, &orig_info[1]); - restore_orig[1] = TRUE; - /* translate the dst box as well */ - dstx = util_format_get_nblocksx(orig_info[1].format, dstx); - dsty = util_format_get_nblocksy(orig_info[1].format, dsty); + src_force_level = src_level; } else if (!util_blitter_is_copy_supported(sctx->blitter, dst, src)) { - unsigned blocksize = util_format_get_blocksize(src->format); - - switch (blocksize) { - case 1: - si_change_format(src, src_level, &orig_info[0], - PIPE_FORMAT_R8_UNORM); - si_change_format(dst, dst_level, &orig_info[1], - PIPE_FORMAT_R8_UNORM); - break; - case 2: - si_change_format(src, src_level, &orig_info[0], - PIPE_FORMAT_R8G8_UNORM); - si_change_format(dst, dst_level, &orig_info[1], - PIPE_FORMAT_R8G8_UNORM); - break; - case 4: - si_change_format(src, src_level, &orig_info[0], - PIPE_FORMAT_R8G8B8A8_UNORM); - si_change_format(dst, dst_level, &orig_info[1], - PIPE_FORMAT_R8G8B8A8_UNORM); - break; - case 8: - si_change_format(src, src_level, &orig_info[0], - PIPE_FORMAT_R16G16B16A16_UINT); - si_change_format(dst, dst_level, &orig_info[1], - PIPE_FORMAT_R16G16B16A16_UINT); - break; - case 16: - si_change_format(src, src_level, &orig_info[0], - PIPE_FORMAT_R32G32B32A32_UINT); - si_change_format(dst, dst_level, &orig_info[1], - PIPE_FORMAT_R32G32B32A32_UINT); - break; - default: - fprintf(stderr, "Unhandled format %s with blocksize %u\n", - util_format_short_name(src->format), blocksize); - assert(0); + if (util_format_is_subsampled_422(src->format)) { + src_templ.format = PIPE_FORMAT_R8G8B8A8_UINT; + dst_templ.format = PIPE_FORMAT_R8G8B8A8_UINT; + + dst_width = util_format_get_nblocksx(dst->format, dst_width); + src_width0 = util_format_get_nblocksx(src->format, src_width0); + + dstx = util_format_get_nblocksx(dst->format, dstx); + + sbox = *src_box; + sbox.x = util_format_get_nblocksx(src->format, src_box->x); + sbox.width = util_format_get_nblocksx(src->format, src_box->width); + src_box = &sbox; + } else { + unsigned blocksize = util_format_get_blocksize(src->format); + + switch (blocksize) { + case 1: + dst_templ.format = PIPE_FORMAT_R8_UNORM; + src_templ.format = PIPE_FORMAT_R8_UNORM; + break; + case 2: + dst_templ.format = PIPE_FORMAT_R8G8_UNORM; + src_templ.format = PIPE_FORMAT_R8G8_UNORM; + break; + case 4: + dst_templ.format = PIPE_FORMAT_R8G8B8A8_UNORM; + src_templ.format = PIPE_FORMAT_R8G8B8A8_UNORM; + break; + case 8: + dst_templ.format = PIPE_FORMAT_R16G16B16A16_UINT; + src_templ.format = PIPE_FORMAT_R16G16B16A16_UINT; + break; + case 16: + dst_templ.format = PIPE_FORMAT_R32G32B32A32_UINT; + src_templ.format = PIPE_FORMAT_R32G32B32A32_UINT; + break; + default: + fprintf(stderr, "Unhandled format %s with blocksize %u\n", + util_format_short_name(src->format), blocksize); + assert(0); + } } - restore_orig[0] = TRUE; - restore_orig[1] = TRUE; } /* Initialize the surface. */ - util_blitter_default_dst_texture(&dst_templ, dst, dst_level, dstz); dst_view = r600_create_surface_custom(ctx, dst, &dst_templ, - rdst->surface.level[dst_level].npix_x, - rdst->surface.level[dst_level].npix_y); + dst_width, dst_height); /* Initialize the sampler view. */ - util_blitter_default_src_texture(&src_templ, src, src_level); - src_view = ctx->create_sampler_view(ctx, src, &src_templ); + src_view = si_create_sampler_view_custom(ctx, src, &src_templ, + src_width0, src_height0, + src_force_level); u_box_3d(dstx, dsty, dstz, abs(src_box->width), abs(src_box->height), abs(src_box->depth), &dstbox); @@ -606,18 +581,12 @@ static void si_resource_copy_region(struct pipe_context *ctx, /* Copy. */ si_blitter_begin(ctx, SI_COPY); util_blitter_blit_generic(sctx->blitter, dst_view, &dstbox, - src_view, src_box, src->width0, src->height0, + src_view, src_box, src_width0, src_height0, PIPE_MASK_RGBAZS, PIPE_TEX_FILTER_NEAREST, NULL); si_blitter_end(ctx); pipe_surface_reference(&dst_view, NULL); pipe_sampler_view_reference(&src_view, NULL); - - if (restore_orig[0]) - si_reset_blittable_to_orig(src, src_level, &orig_info[0]); - - if (restore_orig[1]) - si_reset_blittable_to_orig(dst, dst_level, &orig_info[1]); } /* For MSAA integer resolving to work, we change the format to NORM using this function. */ @@ -721,6 +690,10 @@ static void si_blit(struct pipe_context *ctx, info->src.box.z, info->src.box.z + info->src.box.depth - 1); + if (sctx->screen->b.debug_flags & DBG_FORCE_DMA && + util_try_blit_via_copy_region(ctx, info)) + return; + si_blitter_begin(ctx, SI_BLIT | (info->render_condition_enable ? 0 : SI_DISABLE_RENDER_COND)); util_blitter_blit(sctx->blitter, info); @@ -736,7 +709,7 @@ static void si_flush_resource(struct pipe_context *ctx, if (!rtex->is_depth && rtex->cmask.size) { si_blit_decompress_color(ctx, rtex, 0, res->last_level, - 0, res->array_size - 1); + 0, util_max_layer(res, 0)); } }