anv: Implement VK_KHR_depth_stencil_resolve

diff --git a/src/intel/vulkan/anv_device.c b/src/intel/vulkan/anv_device.c
index 523f1483e29ea9c7a1b985a79296fbc28b014c95..85212074385e267bab68615c36f8994616c22392 100644 (file)
--- a/src/intel/vulkan/anv_device.c
+++ b/src/intel/vulkan/anv_device.c
@@ -1130,6 +1130,34 @@ void anv_GetPhysicalDeviceProperties2(
 
    vk_foreach_struct(ext, pProperties->pNext) {
       switch (ext->sType) {
+      case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DEPTH_STENCIL_RESOLVE_PROPERTIES_KHR: {
+         VkPhysicalDeviceDepthStencilResolvePropertiesKHR *props =
+            (VkPhysicalDeviceDepthStencilResolvePropertiesKHR *)ext;
+
+         /* We support all of the depth resolve modes */
+         props->supportedDepthResolveModes =
+            VK_RESOLVE_MODE_SAMPLE_ZERO_BIT_KHR |
+            VK_RESOLVE_MODE_AVERAGE_BIT_KHR |
+            VK_RESOLVE_MODE_MIN_BIT_KHR |
+            VK_RESOLVE_MODE_MAX_BIT_KHR;
+
+         /* Average doesn't make sense for stencil so we don't support that */
+         props->supportedStencilResolveModes =
+            VK_RESOLVE_MODE_SAMPLE_ZERO_BIT_KHR;
+         if (pdevice->info.gen >= 8) {
+            /* The advanced stencil resolve modes currently require stencil
+             * sampling be supported by the hardware.
+             */
+            props->supportedStencilResolveModes |=
+               VK_RESOLVE_MODE_MIN_BIT_KHR |
+               VK_RESOLVE_MODE_MAX_BIT_KHR;
+         }
+
+         props->independentResolveNone = VK_TRUE;
+         props->independentResolve = VK_TRUE;
+         break;
+      }
+
       case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES_KHR: {
          VkPhysicalDeviceDriverPropertiesKHR *driver_props =
             (VkPhysicalDeviceDriverPropertiesKHR *) ext;

diff --git a/src/intel/vulkan/anv_extensions.py b/src/intel/vulkan/anv_extensions.py
index 388845003aab68dc031bd4e3b47becacfa151581..2ea4cab0e97c3f4b33b72681b2375376dced3efb 100644 (file)
--- a/src/intel/vulkan/anv_extensions.py
+++ b/src/intel/vulkan/anv_extensions.py
@@ -76,6 +76,7 @@ EXTENSIONS = [
     Extension('VK_KHR_bind_memory2',                      1, True),
     Extension('VK_KHR_create_renderpass2',                1, True),
     Extension('VK_KHR_dedicated_allocation',              1, True),
+    Extension('VK_KHR_depth_stencil_resolve',             1, True),
     Extension('VK_KHR_descriptor_update_template',        1, True),
     Extension('VK_KHR_device_group',                      1, True),
     Extension('VK_KHR_device_group_creation',             1, True),

diff --git a/src/intel/vulkan/anv_pass.c b/src/intel/vulkan/anv_pass.c
index f0eca41935ee05a6dc21edcd79d03ce4b74375bf..02f2be60e0299849e0d7364ccb705ba2c9aa4c74 100644 (file)
--- a/src/intel/vulkan/anv_pass.c
+++ b/src/intel/vulkan/anv_pass.c
@@ -74,6 +74,10 @@ anv_render_pass_compile(struct anv_render_pass *pass)
           subpass->depth_stencil_attachment->attachment == VK_ATTACHMENT_UNUSED)
          subpass->depth_stencil_attachment = NULL;
 
+      if (subpass->ds_resolve_attachment &&
+          subpass->ds_resolve_attachment->attachment == VK_ATTACHMENT_UNUSED)
+         subpass->ds_resolve_attachment = NULL;
+
       for (uint32_t j = 0; j < subpass->attachment_count; j++) {
          struct anv_subpass_attachment *subpass_att = &subpass->attachments[j];
          if (subpass_att->attachment == VK_ATTACHMENT_UNUSED)
@@ -116,6 +120,16 @@ anv_render_pass_compile(struct anv_render_pass *pass)
             color_att->usage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
          }
       }
+
+      if (subpass->ds_resolve_attachment) {
+         struct anv_subpass_attachment *ds_att =
+            subpass->depth_stencil_attachment;
+         UNUSED struct anv_subpass_attachment *resolve_att =
+            subpass->ds_resolve_attachment;
+
+         assert(resolve_att->usage == VK_IMAGE_USAGE_TRANSFER_DST_BIT);
+         ds_att->usage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
+      }
    }
 
    /* From the Vulkan 1.0.39 spec:
@@ -342,10 +356,15 @@ VkResult anv_CreateRenderPass(
 static unsigned
 num_subpass_attachments2(const VkSubpassDescription2KHR *desc)
 {
+   const VkSubpassDescriptionDepthStencilResolveKHR *ds_resolve =
+      vk_find_struct_const(desc->pNext,
+                           SUBPASS_DESCRIPTION_DEPTH_STENCIL_RESOLVE_KHR);
+
    return desc->inputAttachmentCount +
           desc->colorAttachmentCount +
           (desc->pResolveAttachments ? desc->colorAttachmentCount : 0) +
-          (desc->pDepthStencilAttachment != NULL);
+          (desc->pDepthStencilAttachment != NULL) +
+          (ds_resolve && ds_resolve->pDepthStencilResolveAttachment);
 }
 
 VkResult anv_CreateRenderPass2KHR(
@@ -460,6 +479,22 @@ VkResult anv_CreateRenderPass2KHR(
             .layout =      desc->pDepthStencilAttachment->layout,
          };
       }
+
+      const VkSubpassDescriptionDepthStencilResolveKHR *ds_resolve =
+         vk_find_struct_const(desc->pNext,
+                              SUBPASS_DESCRIPTION_DEPTH_STENCIL_RESOLVE_KHR);
+
+      if (ds_resolve && ds_resolve->pDepthStencilResolveAttachment) {
+         subpass->ds_resolve_attachment = subpass_attachments++;
+
+         *subpass->ds_resolve_attachment = (struct anv_subpass_attachment) {
+            .usage =       VK_IMAGE_USAGE_TRANSFER_DST_BIT,
+            .attachment =  ds_resolve->pDepthStencilResolveAttachment->attachment,
+            .layout =      ds_resolve->pDepthStencilResolveAttachment->layout,
+         };
+         subpass->depth_resolve_mode = ds_resolve->depthResolveMode;
+         subpass->stencil_resolve_mode = ds_resolve->stencilResolveMode;
+      }
    }
 
    for (uint32_t i = 0; i < pCreateInfo->dependencyCount; i++)

diff --git a/src/intel/vulkan/anv_private.h b/src/intel/vulkan/anv_private.h
index 2eeb31695c2fb7acd209bcabc143e4e79e429c42..aaa17a68241ac6eb61271984912087da8c994142 100644 (file)
--- a/src/intel/vulkan/anv_private.h
+++ b/src/intel/vulkan/anv_private.h
@@ -3235,6 +3235,9 @@ struct anv_subpass {
    struct anv_subpass_attachment *              resolve_attachments;
 
    struct anv_subpass_attachment *              depth_stencil_attachment;
+   struct anv_subpass_attachment *              ds_resolve_attachment;
+   VkResolveModeFlagBitsKHR                     depth_resolve_mode;
+   VkResolveModeFlagBitsKHR                     stencil_resolve_mode;
 
    uint32_t                                     view_mask;
 

diff --git a/src/intel/vulkan/genX_cmd_buffer.c b/src/intel/vulkan/genX_cmd_buffer.c
index 11ec153ff88dcbbce761ff70eb6b4ebca693790d..03ea91aa0f9e3c8c26ad2e52a043de16da032715 100644 (file)
--- a/src/intel/vulkan/genX_cmd_buffer.c
+++ b/src/intel/vulkan/genX_cmd_buffer.c
@@ -3886,6 +3886,23 @@ cmd_buffer_begin_subpass(struct anv_cmd_buffer *cmd_buffer,
    cmd_buffer_emit_depth_stencil(cmd_buffer);
 }
 
+static enum blorp_filter
+vk_to_blorp_resolve_mode(VkResolveModeFlagBitsKHR vk_mode)
+{
+   switch (vk_mode) {
+   case VK_RESOLVE_MODE_SAMPLE_ZERO_BIT_KHR:
+      return BLORP_FILTER_SAMPLE_0;
+   case VK_RESOLVE_MODE_AVERAGE_BIT_KHR:
+      return BLORP_FILTER_AVERAGE;
+   case VK_RESOLVE_MODE_MIN_BIT_KHR:
+      return BLORP_FILTER_MIN_SAMPLE;
+   case VK_RESOLVE_MODE_MAX_BIT_KHR:
+      return BLORP_FILTER_MAX_SAMPLE;
+   default:
+      return BLORP_FILTER_NONE;
+   }
+}
+
 static void
 cmd_buffer_end_subpass(struct anv_cmd_buffer *cmd_buffer)
 {
@@ -3953,6 +3970,125 @@ cmd_buffer_end_subpass(struct anv_cmd_buffer *cmd_buffer)
       }
    }
 
+   if (subpass->ds_resolve_attachment) {
+      /* We are about to do some MSAA resolves.  We need to flush so that the
+       * result of writes to the MSAA depth attachments show up in the sampler
+       * when we blit to the single-sampled resolve target.
+       */
+      cmd_buffer->state.pending_pipe_bits |=
+         ANV_PIPE_TEXTURE_CACHE_INVALIDATE_BIT |
+         ANV_PIPE_DEPTH_CACHE_FLUSH_BIT;
+
+      uint32_t src_att = subpass->depth_stencil_attachment->attachment;
+      uint32_t dst_att = subpass->ds_resolve_attachment->attachment;
+
+      assert(src_att < cmd_buffer->state.pass->attachment_count);
+      assert(dst_att < cmd_buffer->state.pass->attachment_count);
+
+      if (cmd_buffer->state.attachments[dst_att].pending_clear_aspects) {
+         /* From the Vulkan 1.0 spec:
+          *
+          *    If the first use of an attachment in a render pass is as a
+          *    resolve attachment, then the loadOp is effectively ignored
+          *    as the resolve is guaranteed to overwrite all pixels in the
+          *    render area.
+          */
+         cmd_buffer->state.attachments[dst_att].pending_clear_aspects = 0;
+      }
+
+      struct anv_image_view *src_iview = fb->attachments[src_att];
+      struct anv_image_view *dst_iview = fb->attachments[dst_att];
+
+      const VkRect2D render_area = cmd_buffer->state.render_area;
+
+      if ((src_iview->image->aspects & VK_IMAGE_ASPECT_DEPTH_BIT) &&
+          subpass->depth_resolve_mode != VK_RESOLVE_MODE_NONE_KHR) {
+
+         struct anv_attachment_state *src_state =
+            &cmd_state->attachments[src_att];
+         struct anv_attachment_state *dst_state =
+            &cmd_state->attachments[dst_att];
+
+         /* MSAA resolves sample from the source attachment.  Transition the
+          * depth attachment first to get rid of any HiZ that we may not be
+          * able to handle.
+          */
+         transition_depth_buffer(cmd_buffer, src_iview->image,
+                                 src_state->current_layout,
+                                 VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
+         src_state->aux_usage =
+            anv_layout_to_aux_usage(&cmd_buffer->device->info, src_iview->image,
+                                    VK_IMAGE_ASPECT_DEPTH_BIT,
+                                    VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
+         src_state->current_layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
+
+         /* MSAA resolves write to the resolve attachment as if it were any
+          * other transfer op.  Transition the resolve attachment accordingly.
+          */
+         VkImageLayout dst_initial_layout = dst_state->current_layout;
+
+         /* If our render area is the entire size of the image, we're going to
+          * blow it all away so we can claim the initial layout is UNDEFINED
+          * and we'll get a HiZ ambiguate instead of a resolve.
+          */
+         if (dst_iview->image->type != VK_IMAGE_TYPE_3D &&
+             render_area.offset.x == 0 && render_area.offset.y == 0 &&
+             render_area.extent.width == dst_iview->extent.width &&
+             render_area.extent.height == dst_iview->extent.height)
+            dst_initial_layout = VK_IMAGE_LAYOUT_UNDEFINED;
+
+         transition_depth_buffer(cmd_buffer, dst_iview->image,
+                                 dst_initial_layout,
+                                 VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
+         dst_state->aux_usage =
+            anv_layout_to_aux_usage(&cmd_buffer->device->info, dst_iview->image,
+                                    VK_IMAGE_ASPECT_DEPTH_BIT,
+                                    VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
+         dst_state->current_layout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
+
+         enum blorp_filter filter =
+            vk_to_blorp_resolve_mode(subpass->depth_resolve_mode);
+
+         anv_image_msaa_resolve(cmd_buffer,
+                                src_iview->image, src_state->aux_usage,
+                                src_iview->planes[0].isl.base_level,
+                                src_iview->planes[0].isl.base_array_layer,
+                                dst_iview->image, dst_state->aux_usage,
+                                dst_iview->planes[0].isl.base_level,
+                                dst_iview->planes[0].isl.base_array_layer,
+                                VK_IMAGE_ASPECT_DEPTH_BIT,
+                                render_area.offset.x, render_area.offset.y,
+                                render_area.offset.x, render_area.offset.y,
+                                render_area.extent.width,
+                                render_area.extent.height,
+                                fb->layers, filter);
+      }
+
+      if ((src_iview->image->aspects & VK_IMAGE_ASPECT_STENCIL_BIT) &&
+          subpass->stencil_resolve_mode != VK_RESOLVE_MODE_NONE_KHR) {
+
+         enum isl_aux_usage src_aux_usage = ISL_AUX_USAGE_NONE;
+         enum isl_aux_usage dst_aux_usage = ISL_AUX_USAGE_NONE;
+
+         enum blorp_filter filter =
+            vk_to_blorp_resolve_mode(subpass->stencil_resolve_mode);
+
+         anv_image_msaa_resolve(cmd_buffer,
+                                src_iview->image, src_aux_usage,
+                                src_iview->planes[0].isl.base_level,
+                                src_iview->planes[0].isl.base_array_layer,
+                                dst_iview->image, dst_aux_usage,
+                                dst_iview->planes[0].isl.base_level,
+                                dst_iview->planes[0].isl.base_array_layer,
+                                VK_IMAGE_ASPECT_STENCIL_BIT,
+                                render_area.offset.x, render_area.offset.y,
+                                render_area.offset.x, render_area.offset.y,
+                                render_area.extent.width,
+                                render_area.extent.height,
+                                fb->layers, filter);
+      }
+   }
+
    for (uint32_t i = 0; i < subpass->attachment_count; ++i) {
       const uint32_t a = subpass->attachments[i].attachment;
       if (a == VK_ATTACHMENT_UNUSED)