tu/a750: Disable HW binning when there is GS

Blob doesn't use hw binning with GS on all a6xx and a7xx, however
in Turnip it worked without issues until a750. On a750 there are CTS
failures when e.g. dEQP-VK.subgroups.arithmetic.framebuffer.* in
parallel with "forcebin". It is exacerbated by using "syncdraw".

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29074>

src/freedreno/common/freedreno_dev_info.h

@@ -247,6 +247,13 @@ struct fd_dev_info {
        * fully compatible.
        */
       bool ubwc_unorm_snorm_int_compatible;
+
+      /* Blob doesn't use hw binning with GS on all a6xx and a7xx, however
+       * in Turnip it worked without issues until a750. On a750 there are CTS
+       * failures when e.g. dEQP-VK.subgroups.arithmetic.framebuffer.* in
+       * parallel with "forcebin". It is exacerbated by using "syncdraw".
+       */
+      bool no_gs_hw_binning_quirk;
    } a7xx;
 };
 

src/freedreno/common/freedreno_devices.py

@@ -825,6 +825,7 @@ a7xx_750 = A7XXProps(
         # example dEQP-VK.image.load_store.with_format.2d.*. Disable this for
         # now.
         #supports_ibo_ubwc = True,
+        no_gs_hw_binning_quirk = True,
     )
 
 a730_magic_regs = dict(

src/freedreno/vulkan/tu_cmd_buffer.cc

@@ -858,6 +858,10 @@ use_hw_binning(struct tu_cmd_buffer *cmd)
    const struct tu_framebuffer *fb = cmd->state.framebuffer;
    const struct tu_tiling_config *tiling = &fb->tiling[cmd->state.gmem_layout];
 
+   if (cmd->state.rp.has_gs &&
+       cmd->device->physical_device->info->a7xx.no_gs_hw_binning_quirk)
+      return false;
+
    /* XFB commands are emitted for BINNING || SYSMEM, which makes it
     * incompatible with non-hw binning GMEM rendering. this is required because
     * some of the XFB commands need to only be executed once.
@@ -908,17 +912,21 @@ use_sysmem_rendering(struct tu_cmd_buffer *cmd,
    if (cmd->state.rp.disable_gmem)
       return true;
 
-   /* XFB is incompatible with non-hw binning GMEM rendering, see use_hw_binning */
-   if (cmd->state.rp.xfb_used && !cmd->state.tiling->binning_possible)
-      return true;
+   if (!cmd->state.tiling->binning_possible ||
+       (cmd->state.rp.has_gs &&
+        cmd->device->physical_device->info->a7xx.no_gs_hw_binning_quirk)) {
+      /* XFB is incompatible with non-hw binning GMEM rendering, see
+       * use_hw_binning */
+      if (cmd->state.rp.xfb_used)
+         return true;
 
-   /* QUERY_TYPE_PRIMITIVES_GENERATED is incompatible with non-hw binning
-    * GMEM rendering, see use_hw_binning.
-    */
-   if ((cmd->state.rp.has_prim_generated_query_in_rp ||
-        cmd->state.prim_generated_query_running_before_rp) &&
-       !cmd->state.tiling->binning_possible)
-      return true;
+      /* QUERY_TYPE_PRIMITIVES_GENERATED is incompatible with non-hw binning
+       * GMEM rendering, see use_hw_binning.
+       */
+      if (cmd->state.rp.has_prim_generated_query_in_rp ||
+          cmd->state.prim_generated_query_running_before_rp)
+         return true;
+   }
 
    if (TU_DEBUG(GMEM))
       return false;
@@ -3183,6 +3191,9 @@ tu_pipeline_update_rp_state(struct tu_cmd_state *cmd_state)
    if (cmd_state->pipeline_has_tess) {
       cmd_state->rp.has_tess = true;
    }
+   if (cmd_state->pipeline_has_gs) {
+      cmd_state->rp.has_gs = true;
+   }
 }
 
 VKAPI_ATTR void VKAPI_CALL
@@ -3234,6 +3245,7 @@ tu_CmdBindPipeline(VkCommandBuffer commandBuffer,
    cmd->state.prim_order_gmem = pipeline->prim_order.state_gmem;
    cmd->state.pipeline_sysmem_single_prim_mode = pipeline->prim_order.sysmem_single_prim_mode;
    cmd->state.pipeline_has_tess = pipeline->active_stages & VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT;
+   cmd->state.pipeline_has_gs = pipeline->active_stages & VK_SHADER_STAGE_GEOMETRY_BIT;
    cmd->state.pipeline_disable_gmem = gfx_pipeline->feedback_loop_may_involve_textures;
 
    tu_pipeline_update_rp_state(&cmd->state);
@@ -3723,6 +3735,7 @@ tu_render_pass_state_merge(struct tu_render_pass_state *dst,
 {
    dst->xfb_used |= src->xfb_used;
    dst->has_tess |= src->has_tess;
+   dst->has_gs |= src->has_gs;
    dst->has_prim_generated_query_in_rp |= src->has_prim_generated_query_in_rp;
    dst->disable_gmem |= src->disable_gmem;
    dst->sysmem_single_prim_mode |= src->sysmem_single_prim_mode;

src/freedreno/vulkan/tu_cmd_buffer.h

@@ -276,6 +276,7 @@ struct tu_render_pass_state
 {
    bool xfb_used;
    bool has_tess;
+   bool has_gs;
    bool has_prim_generated_query_in_rp;
    bool disable_gmem;
    bool sysmem_single_prim_mode;
@@ -504,6 +505,7 @@ struct tu_cmd_state
    bool pipeline_feedback_loop_ds;
    bool pipeline_sysmem_single_prim_mode;
    bool pipeline_has_tess;
+   bool pipeline_has_gs;
    bool pipeline_disable_gmem;
 
    bool pipeline_blend_lrz, pipeline_bandwidth;