claude_code/procedural-3d-engine
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Proper barriers for buffer writes and reads between host and device (non-coherent)

Browse source
This commit is contained in:
saschawillems 2017-09-15 20:43:40 +02:00
parent 66a87bbf21
commit 3862b4bcc2
1 changed files with 32 additions and 16 deletions
Download patch file Download diff file Expand all files Collapse all files

48
computeheadless/computeheadless.cpp
View file

@ -247,7 +247,8 @@ public:
// Submit to the queue
VK_CHECK_RESULT(vkQueueSubmit(queue, 1, &submitInfo, fence));
VK_CHECK_RESULT(vkWaitForFences(device, 1, &fence, VK_TRUE, DEFAULT_FENCE_TIMEOUT));
VK_CHECK_RESULT(vkWaitForFences(device, 1, &fence, VK_TRUE, UINT64_MAX));
vkDestroyFence(device, fence, nullptr);
vkFreeCommandBuffers(device, commandPool, 1, &copyCmd);
}
@ -279,9 +280,7 @@ public:
vks::initializers::descriptorSetAllocateInfo(descriptorPool, &descriptorSetLayout, 1);
VK_CHECK_RESULT(vkAllocateDescriptorSets(device, &allocInfo, &descriptorSet));
VkDescriptorBufferInfo bufferDescriptor = {};
bufferDescriptor.buffer = deviceBuffer;
bufferDescriptor.range = VK_WHOLE_SIZE;
VkDescriptorBufferInfo bufferDescriptor = { deviceBuffer, 0, VK_WHOLE_SIZE };
std::vector<VkWriteDescriptorSet> computeWriteDescriptorSets = {
vks::initializers::writeDescriptorSet(descriptorSet, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, 0, &bufferDescriptor),
};
@ -325,17 +324,18 @@ public:
VK_CHECK_RESULT(vkBeginCommandBuffer(commandBuffer, &cmdBufInfo));
// Barrier to ensure that input buffer transfer is finished before compute shader reads from it
VkBufferMemoryBarrier bufferBarrier = vks::initializers::bufferMemoryBarrier();
bufferBarrier.buffer = deviceBuffer;
bufferBarrier.size = VK_WHOLE_SIZE;
bufferBarrier.srcAccessMask = VK_ACCESS_HOST_WRITE_BIT;
bufferBarrier.dstAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
bufferBarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
bufferBarrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
bufferBarrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
bufferBarrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
vkCmdPipelineBarrier(
commandBuffer,
VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_HOST_BIT,
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
VK_FLAGS_NONE,
0, nullptr,
@ -348,9 +348,8 @@ public:
vkCmdDispatch(commandBuffer, BUFFER_ELEMENTS, 1, 1);
// Barrier to ensure that shader writes are finished before buffer is read back from GPU
bufferBarrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
bufferBarrier.dstAccessMask = VK_ACCESS_HOST_READ_BIT;
bufferBarrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
bufferBarrier.buffer = deviceBuffer;
bufferBarrier.size = VK_WHOLE_SIZE;
bufferBarrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
@ -370,19 +369,34 @@ public:
copyRegion.size = bufferSize;
vkCmdCopyBuffer(commandBuffer, deviceBuffer, hostBuffer, 1, &copyRegion);
// Barrier to ensure that buffer copy is finished before host reading from it
bufferBarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
bufferBarrier.dstAccessMask = VK_ACCESS_HOST_READ_BIT;
bufferBarrier.buffer = hostBuffer;
bufferBarrier.size = VK_WHOLE_SIZE;
bufferBarrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
bufferBarrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
vkCmdPipelineBarrier(
commandBuffer,
VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_HOST_BIT,
VK_FLAGS_NONE,
0, nullptr,
1, &bufferBarrier,
0, nullptr);
VK_CHECK_RESULT(vkEndCommandBuffer(commandBuffer));
// Submit compute work
vkResetFences(device, 1, &fence);
const VkPipelineStageFlags waitStageMask = VK_PIPELINE_STAGE_TRANSFER_BIT;
VkSubmitInfo computeSubmitInfo = vks::initializers::submitInfo();
computeSubmitInfo.pWaitDstStageMask = &waitStageMask;
computeSubmitInfo.commandBufferCount = 1;
computeSubmitInfo.pCommandBuffers = &commandBuffer;
VK_CHECK_RESULT(vkQueueSubmit(queue, 1, &computeSubmitInfo, fence));
// Wait for fence
vkWaitForFences(device, 1, &fence, VK_TRUE, UINT64_MAX);
VK_CHECK_RESULT(vkWaitForFences(device, 1, &fence, VK_TRUE, UINT64_MAX));
// Make device writes visible to the host
void *mapped;
@ -393,11 +407,13 @@ public:
mappedRange.size = VK_WHOLE_SIZE;
vkInvalidateMappedMemoryRanges(device, 1, &mappedRange);
void* mapped;
vkMapMemory(device, hostMemory, 0, VK_WHOLE_SIZE, 0, &mapped);
// Copy to output
memcpy(computeOutput.data(), mapped, bufferSize);
vkUnmapMemory(device, hostMemory);
}
vkQueueWaitIdle(queue);
// Output buffer contents
std::cout << "Compute input: " << std::endl;
for (auto v : computeInput) {