zink: force unordered_write=false when binding image descriptors

there's no way to link up image layouts between the unordered cmdbuf
and the main one, so if an op is promoted to unordered after an image
is used as a descriptor, the layout will be broken

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22027>

diff --git a/src/gallium/drivers/zink/zink_context.c b/src/gallium/drivers/zink/zink_context.c
index c1adf58..6a7abe9 100644 (file)
--- a/src/gallium/drivers/zink/zink_context.c
+++ b/src/gallium/drivers/zink/zink_context.c
@@ -1738,8 +1738,9 @@ finalize_image_bind(struct zink_context *ctx, struct zink_resource *res, bool is
       update_binds_for_samplerviews(ctx, res, is_compute);
    if (!check_for_layout_update(ctx, res, is_compute)) {
       /* no deferred barrier: unset unordered usage immediately */
-      if (zink_resource_access_is_write(res->barrier_access[is_compute]))
-         res->obj->unordered_write = false;
+      // TODO: figure out a way to link up layouts between unordered and main cmdbuf
+      // if (zink_resource_access_is_write(res->barrier_access[is_compute]))
+      res->obj->unordered_write = false;
       res->obj->unordered_read = false;
    }
 }
@@ -2039,13 +2040,15 @@ zink_set_sampler_views(struct pipe_context *pctx,
             if (b->cube_array) {
                ctx->di.cubes[shader_type] |= BITFIELD_BIT(start_slot + i);
             }
-            if (!check_for_layout_update(ctx, res, shader_type == MESA_SHADER_COMPUTE) && !ctx->unordered_blitting)
+            if (!check_for_layout_update(ctx, res, shader_type == MESA_SHADER_COMPUTE) && !ctx->unordered_blitting) {
                /* no deferred barrier: unset unordered usage immediately */
                res->obj->unordered_read = false;
+               // TODO: figure out a way to link up layouts between unordered and main cmdbuf
+               res->obj->unordered_write = false;
+            }
             if (!a)
                update = true;
             zink_batch_resource_usage_set(&ctx->batch, res, false, false);
-            res->obj->unordered_write = false;
             if (b->zs_view) {
                assert(start_slot + i < 32); //bitfield size
                ctx->di.zs_swizzle[shader_type].mask |= BITFIELD_BIT(start_slot + i);
@@ -2271,10 +2274,16 @@ zink_make_texture_handle_resident(struct pipe_context *pctx, uint64_t handle, bo
          ii->imageView = ds->surface->image_view;
          ii->imageLayout = zink_descriptor_util_image_layout_eval(ctx, res, false);
          flush_pending_clears(ctx, res);
-         if (!check_for_layout_update(ctx, res, false))
+         if (!check_for_layout_update(ctx, res, false)) {
             res->obj->unordered_read = false;
-         if (!check_for_layout_update(ctx, res, true))
+            // TODO: figure out a way to link up layouts between unordered and main cmdbuf
+            res->obj->unordered_write = false;
+         }
+         if (!check_for_layout_update(ctx, res, true)) {
             res->obj->unordered_read = false;
+            // TODO: figure out a way to link up layouts between unordered and main cmdbuf
+            res->obj->unordered_write = false;
+         }
          zink_batch_resource_usage_set(&ctx->batch, res, false, false);
          res->obj->unordered_write = false;
       }
@@ -5725,10 +5734,11 @@ zink_update_barriers(struct zink_context *ctx, bool is_compute,
             if (is_feedback)
                update_res_sampler_layouts(ctx, res);
          }
-         if (zink_resource_access_is_write(res->barrier_access[is_compute]))
-            res->obj->unordered_read = res->obj->unordered_write = false;
-         else
-            res->obj->unordered_read = false;
+         if (zink_resource_access_is_write(res->barrier_access[is_compute]) ||
+             // TODO: figure out a way to link up layouts between unordered and main cmdbuf
+             res->base.b.target != PIPE_BUFFER)
+            res->obj->unordered_write = false;
+         res->obj->unordered_read = false;
          /* always barrier on draw if this resource has either multiple image write binds or
           * image write binds and image read binds
           */