Optimizing WriteTexture offset on Vulkan

Added an additional alignment on the staging buffer offset in
WriteTexture on Vulkan to match
VkPhysicalDeviceLimits::optimalBufferCopyOffsetAlignment.
We now allocate some additional memory in the staging buffer for that
purpose.

Bug: dawn:483
Change-Id: Ic5b95aa57f6c241e0ab9f9853b4c3152e52e5630
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24982
Commit-Queue: Tomek Ponitka <tommek@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>

src/dawn_native/vulkan/QueueVk.cpp

@@ -41,9 +41,15 @@ namespace dawn_native { namespace vulkan {
             uint32_t newDataSize = ComputeRequiredBytesInCopy(
                 textureFormat, *writeSize, optimallyAlignedBytesPerRow, alignedRowsPerImage);
 
+            uint64_t optimalOffsetAlignment =
+                ToBackend(device)
+                    ->GetDeviceInfo()
+                    .properties.limits.optimalBufferCopyOffsetAlignment;
+
             UploadHandle uploadHandle;
             DAWN_TRY_ASSIGN(uploadHandle, device->GetDynamicUploader()->Allocate(
-                                              newDataSize, device->GetPendingCommandSerial()));
+                                              newDataSize + optimalOffsetAlignment - 1,
+                                              device->GetPendingCommandSerial()));
             ASSERT(uploadHandle.mappedBuffer != nullptr);
 
             // TODO(tommek@google.com): Add an optimization to do a single memcpy if the data
@@ -58,6 +64,11 @@ namespace dawn_native { namespace vulkan {
                 dataRowsPerImageInBlock = writeSize->height / textureFormat.blockHeight;
             }
 
+            uint64_t additionalOffset =
+                Align(uploadHandle.startOffset, optimalOffsetAlignment) - uploadHandle.startOffset;
+            uploadHandle.startOffset += additionalOffset;
+            dstPointer += additionalOffset;
+
             ASSERT(dataRowsPerImageInBlock >= alignedRowsPerImageInBlock);
             uint64_t imageAdditionalStride =
                 dataLayout->bytesPerRow * (dataRowsPerImageInBlock - alignedRowsPerImageInBlock);