intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-05 09:09:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

Handle printf output when implicit args are required

Browse Source 
Related-To: NEO-5081
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
This commit is contained in:
Mateusz Jablonski
2021-09-22 13:16:09 +00:00
committed by Compute-Runtime-Automation
parent 8b36473a9a
commit 34ad95ae12
7 changed files with 141 additions and 58 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
opencl/source/helpers/task_information.h
View File

@@ -139,6 +139,7 @@ class CommandComputeKernel : public Command {
LinearStream *getCommandStream() override { return kernelOperation->commandStream.get(); }
Kernel *peekKernel() const { return kernel; }
PrintfHandler *peekPrintfHandler() const { return printfHandler.get(); }
protected:
std::vector<Surface *> surfaces;

46
opencl/source/program/printf_handler.cpp
View File

@@ -61,16 +61,19 @@ void PrintfHandler::prepareDispatch(const MultiDispatchInfo &multiDispatchInfo)
device.getDevice(), printfSurface, 0, printfSurfaceInitialDataSizePtr.get(),
sizeof(*printfSurfaceInitialDataSizePtr.get()));
const auto &printfSurfaceArg = kernel->getKernelInfo().kernelDescriptor.payloadMappings.implicitArgs.printfSurfaceAddress;
auto printfPatchAddress = ptrOffset(reinterpret_cast<uintptr_t *>(kernel->getCrossThreadData()), printfSurfaceArg.stateless);
patchWithRequiredSize(printfPatchAddress, printfSurfaceArg.pointerSize, (uintptr_t)printfSurface->getGpuAddressToPatch());
if (isValidOffset(printfSurfaceArg.bindful)) {
auto surfaceState = ptrOffset(reinterpret_cast<uintptr_t *>(kernel->getSurfaceStateHeap()), printfSurfaceArg.bindful);
void *addressToPatch = printfSurface->getUnderlyingBuffer();
size_t sizeToPatch = printfSurface->getUnderlyingBufferSize();
Buffer::setSurfaceState(&device.getDevice(), surfaceState, false, false, sizeToPatch, addressToPatch, 0, printfSurface, 0, 0,
kernel->getKernelInfo().kernelDescriptor.kernelAttributes.flags.useGlobalAtomics,
kernel->areMultipleSubDevicesInContext());
if (kernel->getKernelInfo().kernelDescriptor.kernelAttributes.flags.usesPrintf) {
const auto &printfSurfaceArg = kernel->getKernelInfo().kernelDescriptor.payloadMappings.implicitArgs.printfSurfaceAddress;
auto printfPatchAddress = ptrOffset(reinterpret_cast<uintptr_t *>(kernel->getCrossThreadData()), printfSurfaceArg.stateless);
patchWithRequiredSize(printfPatchAddress, printfSurfaceArg.pointerSize, (uintptr_t)printfSurface->getGpuAddressToPatch());
if (isValidOffset(printfSurfaceArg.bindful)) {
auto surfaceState = ptrOffset(reinterpret_cast<uintptr_t *>(kernel->getSurfaceStateHeap()), printfSurfaceArg.bindful);
void *addressToPatch = printfSurface->getUnderlyingBuffer();
size_t sizeToPatch = printfSurface->getUnderlyingBufferSize();
Buffer::setSurfaceState(&device.getDevice(), surfaceState, false, false, sizeToPatch, addressToPatch, 0, printfSurface, 0, 0,
kernel->getKernelInfo().kernelDescriptor.kernelAttributes.flags.useGlobalAtomics,
kernel->areMultipleSubDevicesInContext());
}
}
auto pImplicitArgs = kernel->getImplicitArgs();
if (pImplicitArgs) {
@@ -83,31 +86,28 @@ void PrintfHandler::makeResident(CommandStreamReceiver &commandStreamReceiver) {
}
void PrintfHandler::printEnqueueOutput() {
auto usesStringMap = kernel->getDescriptor().kernelAttributes.flags.usesStringMapForPrintf || nullptr != kernel->getImplicitArgs();
const auto &hwInfoConfig = *HwInfoConfig::get(device.getHardwareInfo().platform.eProductFamily);
auto printfOutputBuffer = reinterpret_cast<const uint8_t *>(printfSurface->getUnderlyingBuffer());
auto printfOutputSize = static_cast<uint32_t>(printfSurface->getUnderlyingBufferSize());
std::unique_ptr<uint8_t[]> printfOutputDecompressed;
if (hwInfoConfig.allowStatelessCompression(device.getHardwareInfo())) {
auto printOutputSize = static_cast<uint32_t>(printfSurface->getUnderlyingBufferSize());
auto printOutputDecompressed = std::make_unique<uint8_t[]>(printOutputSize);
printfOutputDecompressed = std::make_unique<uint8_t[]>(printfOutputSize);
printfOutputBuffer = printfOutputDecompressed.get();
auto &bcsEngine = device.getEngine(EngineHelpers::getBcsEngineType(device.getHardwareInfo(), device.getDeviceBitfield(), device.getSelectorCopyEngine(), true), EngineUsage::Regular);
BlitPropertiesContainer blitPropertiesContainer;
blitPropertiesContainer.push_back(
BlitProperties::constructPropertiesForReadWrite(BlitterConstants::BlitDirection::BufferToHostPtr,
*bcsEngine.commandStreamReceiver, printfSurface, nullptr,
printOutputDecompressed.get(),
printfOutputDecompressed.get(),
printfSurface->getGpuAddress(),
0, 0, 0, Vec3<size_t>(printOutputSize, 0, 0), 0, 0, 0, 0));
0, 0, 0, Vec3<size_t>(printfOutputSize, 0, 0), 0, 0, 0, 0));
bcsEngine.commandStreamReceiver->blitBuffer(blitPropertiesContainer, true, false, device.getDevice());
PrintFormatter printFormatter(printOutputDecompressed.get(), printOutputSize,
kernel->is32Bit(),
kernel->getDescriptor().kernelAttributes.flags.usesStringMapForPrintf ? &kernel->getDescriptor().kernelMetadata.printfStringsMap : nullptr);
printFormatter.printKernelOutput();
return;
}
PrintFormatter printFormatter(reinterpret_cast<const uint8_t *>(printfSurface->getUnderlyingBuffer()), static_cast<uint32_t>(printfSurface->getUnderlyingBufferSize()),
kernel->is32Bit(),
kernel->getDescriptor().kernelAttributes.flags.usesStringMapForPrintf ? &kernel->getDescriptor().kernelMetadata.printfStringsMap : nullptr);
PrintFormatter printFormatter(printfOutputBuffer, printfOutputSize, kernel->is32Bit(),
usesStringMap ? &kernel->getDescriptor().kernelMetadata.printfStringsMap : nullptr);
printFormatter.printKernelOutput();
}
} // namespace NEO