blob: cbc5b24fc1e0e08f8dcddb8aa7ead311d6ba0a40 [file] [log] [blame]
* Copyright 2022 Google LLC
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
#ifndef skgpu_graphite_VulkanBuffer_DEFINED
#define skgpu_graphite_VulkanBuffer_DEFINED
#include "include/gpu/vk/VulkanMemoryAllocator.h"
#include "src/gpu/graphite/Buffer.h"
#include "src/gpu/graphite/vk/VulkanSharedContext.h"
namespace skgpu::graphite {
class VulkanBuffer final : public Buffer {
static sk_sp<Buffer> Make(const VulkanSharedContext*, size_t, BufferType, PrioritizeGpuReads);
void freeGpuData() override {}
VkBuffer vkBuffer() const { return fBuffer; }
VulkanBuffer(const VulkanSharedContext*, size_t, BufferType, PrioritizeGpuReads, VkBuffer,
const skgpu::VulkanAlloc&);
void onMap() override;
void onUnmap() override;
void internalMap(size_t readOffset, size_t readSize);
void internalUnmap(size_t flushOffset, size_t flushSize);
bool isMappable() const { return fAlloc.fFlags & skgpu::VulkanAlloc::kMappable_Flag; }
const VulkanSharedContext* vulkanSharedContext() const {
return static_cast<const VulkanSharedContext*>(this->sharedContext());
VkBuffer fBuffer;
skgpu::VulkanAlloc fAlloc;
* Buffers can either be mapped for:
* 1) Reading from the CPU (The effect of writing would be undefined)
* 2) Writing from the CPU (The existing contents are discarded. Even in the case where the
* initial contents are overwritten, CPU reads should be avoided for performance reasons as
* the memory may not be cached).
bool fBufferUsedForCPURead = false;
} // namespace skgpu::graphite
#endif // skgpu_graphite_VulkanBuffer_DEFINED