radeonsi: add DCC decompression (v2)

This is currently not needed but will be necessary when we have
features that do not work with DCC enabled, such as image stores
and sharing non-scanout surfaces.

v2: Marek: rebase, remove decompression from si_flush_resource (not needed)

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

diff --git a/src/gallium/drivers/radeonsi/si_blit.c b/src/gallium/drivers/radeonsi/si_blit.c
index 60b9f7b6a708827df9ff03b091a8e07a88053003..198f57d542fd6e0e8ca484f262c8579726f0d5dd 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_blit.c
+++ b/src/gallium/drivers/radeonsi/si_blit.c
@@ -271,18 +271,29 @@ void si_flush_depth_textures(struct si_context *sctx,
 static void si_blit_decompress_color(struct pipe_context *ctx,
                struct r600_texture *rtex,
                unsigned first_level, unsigned last_level,
-               unsigned first_layer, unsigned last_layer)
+               unsigned first_layer, unsigned last_layer,
+               bool need_dcc_decompress)
 {
        struct si_context *sctx = (struct si_context *)ctx;
        unsigned layer, level, checked_last_layer, max_layer;
 
-       if (!rtex->dirty_level_mask)
+       if (!rtex->dirty_level_mask && !need_dcc_decompress)
                return;
 
        for (level = first_level; level <= last_level; level++) {
-               if (!(rtex->dirty_level_mask & (1 << level)))
+               void* custom_blend;
+
+               if (!(rtex->dirty_level_mask & (1 << level)) && !need_dcc_decompress)
                        continue;
 
+               if (rtex->dcc_offset && need_dcc_decompress) {
+                       custom_blend = sctx->custom_blend_dcc_decompress;
+               } else if (rtex->fmask.size) {
+                       custom_blend = sctx->custom_blend_decompress;
+               } else {
+                       custom_blend = sctx->custom_blend_fastclear;
+               }
+
                /* The smaller the mipmap level, the less layers there are
                 * as far as 3D textures are concerned. */
                max_layer = util_max_layer(&rtex->resource.b.b, level);
@@ -298,9 +309,7 @@ static void si_blit_decompress_color(struct pipe_context *ctx,
                        cbsurf = ctx->create_surface(ctx, &rtex->resource.b.b, &surf_tmpl);
 
                        si_blitter_begin(ctx, SI_DECOMPRESS);
-                       util_blitter_custom_color(sctx->blitter, cbsurf,
-                               rtex->fmask.size ? sctx->custom_blend_decompress :
-                                                  sctx->custom_blend_fastclear);
+                       util_blitter_custom_color(sctx->blitter, cbsurf, custom_blend);
                        si_blitter_end(ctx);
 
                        pipe_surface_reference(&cbsurf, NULL);
@@ -334,7 +343,8 @@ void si_decompress_color_textures(struct si_context *sctx,
 
                si_blit_decompress_color(&sctx->b.b, tex,
                                         view->u.tex.first_level, view->u.tex.last_level,
-                                        0, util_max_layer(&tex->resource.b.b, view->u.tex.first_level));
+                                        0, util_max_layer(&tex->resource.b.b, view->u.tex.first_level),
+                                        false);
        }
 }
 
@@ -485,7 +495,7 @@ static void si_decompress_subresource(struct pipe_context *ctx,
                                                          first_layer, last_layer);
        } else if (rtex->fmask.size || rtex->cmask.size || rtex->dcc_offset) {
                si_blit_decompress_color(ctx, rtex, level, level,
-                                        first_layer, last_layer);
+                                        first_layer, last_layer, false);
        }
 }
 
@@ -763,7 +773,7 @@ static void si_flush_resource(struct pipe_context *ctx,
 
        if (!rtex->is_depth && (rtex->cmask.size || rtex->dcc_offset)) {
                si_blit_decompress_color(ctx, rtex, 0, res->last_level,
-                                        0, util_max_layer(res, 0));
+                                        0, util_max_layer(res, 0), false);
        }
 }
 

diff --git a/src/gallium/drivers/radeonsi/si_pipe.c b/src/gallium/drivers/radeonsi/si_pipe.c
index 37fd4a25d590f0c3ef0df4391212937e16b340b6..b445f6f010f76548a710c82670dbd635ce9ed2e9 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_pipe.c
+++ b/src/gallium/drivers/radeonsi/si_pipe.c
@@ -68,6 +68,8 @@ static void si_destroy_context(struct pipe_context *context)
                sctx->b.b.delete_blend_state(&sctx->b.b, sctx->custom_blend_decompress);
        if (sctx->custom_blend_fastclear)
                sctx->b.b.delete_blend_state(&sctx->b.b, sctx->custom_blend_fastclear);
+       if (sctx->custom_blend_dcc_decompress)
+               sctx->b.b.delete_blend_state(&sctx->b.b, sctx->custom_blend_dcc_decompress);
        util_unreference_framebuffer_state(&sctx->framebuffer.state);
 
        if (sctx->blitter)

diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index e6f92a31f9682095608ad02c58d770e8ec02b4bd..4d45e8ca3f1f05bc995a791b4ebbe14b48305802 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -196,6 +196,7 @@ struct si_context {
        void                            *custom_blend_resolve;
        void                            *custom_blend_decompress;
        void                            *custom_blend_fastclear;
+       void                            *custom_blend_dcc_decompress;
        void                            *pstipple_sampler_state;
        struct si_screen                *screen;
        struct pipe_fence_handle        *last_gfx_fence;

diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index 08a6505b0b3e699297db01a9ee6c02c312f05879..c7be30a2967e12b9d255102df1f59f55f71e2a95 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -3513,6 +3513,7 @@ void si_init_state_functions(struct si_context *sctx)
        sctx->custom_blend_resolve = si_create_blend_custom(sctx, V_028808_CB_RESOLVE);
        sctx->custom_blend_decompress = si_create_blend_custom(sctx, V_028808_CB_FMASK_DECOMPRESS);
        sctx->custom_blend_fastclear = si_create_blend_custom(sctx, V_028808_CB_ELIMINATE_FAST_CLEAR);
+       sctx->custom_blend_dcc_decompress = si_create_blend_custom(sctx, V_028808_CB_DCC_DECOMPRESS);
 
        sctx->b.b.set_clip_state = si_set_clip_state;
        sctx->b.b.set_scissor_states = si_set_scissor_states;