compute-runtime/runtime/command_queue/enqueue_read_image.h

121 lines
4.1 KiB
C++

/*
* Copyright (C) 2017-2018 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#pragma once
#include "runtime/built_ins/built_ins.h"
#include "hw_cmds.h"
#include "runtime/command_queue/command_queue_hw.h"
#include "runtime/command_stream/command_stream_receiver.h"
#include "runtime/event/event.h"
#include "runtime/helpers/surface_formats.h"
#include "runtime/helpers/cache_policy.h"
#include "runtime/helpers/kernel_commands.h"
#include "runtime/helpers/basic_math.h"
#include "runtime/helpers/mipmap.h"
#include "runtime/mem_obj/image.h"
#include "runtime/memory_manager/surface.h"
#include <algorithm>
#include <new>
namespace OCLRT {
template <typename GfxFamily>
cl_int CommandQueueHw<GfxFamily>::enqueueReadImage(
Image *srcImage,
cl_bool blockingRead,
const size_t *origin,
const size_t *region,
size_t inputRowPitch,
size_t inputSlicePitch,
void *ptr,
cl_uint numEventsInWaitList,
const cl_event *eventWaitList,
cl_event *event) {
notifyEnqueueReadImage(srcImage, !!blockingRead);
MultiDispatchInfo di;
auto isMemTransferNeeded = true;
if (srcImage->isMemObjZeroCopy()) {
size_t hostOffset;
Image::calculateHostPtrOffset(&hostOffset, origin, region, inputRowPitch, inputSlicePitch, srcImage->getImageDesc().image_type, srcImage->getSurfaceFormatInfo().ImageElementSizeInBytes);
isMemTransferNeeded = srcImage->checkIfMemoryTransferIsRequired(hostOffset, 0, ptr, CL_COMMAND_READ_IMAGE);
}
if (!isMemTransferNeeded) {
NullSurface s;
Surface *surfaces[] = {&s};
enqueueHandler<CL_COMMAND_MARKER>(
surfaces,
blockingRead == CL_TRUE,
di,
numEventsInWaitList,
eventWaitList,
event);
if (event) {
auto pEvent = castToObjectOrAbort<Event>(*event);
pEvent->setCmdType(CL_COMMAND_READ_IMAGE);
}
if (context->isProvidingPerformanceHints()) {
context->providePerformanceHint(CL_CONTEXT_DIAGNOSTICS_LEVEL_GOOD_INTEL, CL_ENQUEUE_READ_IMAGE_DOESNT_REQUIRES_COPY_DATA, static_cast<cl_mem>(srcImage));
}
return CL_SUCCESS;
}
auto &builder = getDevice().getExecutionEnvironment()->getBuiltIns()->getBuiltinDispatchInfoBuilder(EBuiltInOps::CopyImage3dToBuffer,
this->getContext(), this->getDevice());
BuiltInOwnershipWrapper builtInLock(builder, this->context);
size_t hostPtrSize = calculateHostPtrSizeForImage(region, inputRowPitch, inputSlicePitch, srcImage);
void *dstPtr = ptr;
MemObjSurface srcImgSurf(srcImage);
HostPtrSurface hostPtrSurf(dstPtr, hostPtrSize);
Surface *surfaces[] = {&srcImgSurf, &hostPtrSurf};
if (region[0] != 0 &&
region[1] != 0 &&
region[2] != 0) {
bool status = getCommandStreamReceiver().createAllocationForHostSurface(hostPtrSurf, getDevice(), true);
if (!status) {
return CL_OUT_OF_RESOURCES;
}
dstPtr = reinterpret_cast<void *>(hostPtrSurf.getAllocation()->getGpuAddressToPatch());
}
BuiltinDispatchInfoBuilder::BuiltinOpParams dc;
dc.srcMemObj = srcImage;
dc.dstPtr = dstPtr;
dc.srcOffset = origin;
dc.size = region;
dc.srcRowPitch = inputRowPitch;
dc.srcSlicePitch = inputSlicePitch;
if (srcImage->getImageDesc().num_mip_levels > 0) {
dc.srcMipLevel = findMipLevel(srcImage->getImageDesc().image_type, origin);
}
builder.buildDispatchInfos(di, dc);
enqueueHandler<CL_COMMAND_READ_IMAGE>(
surfaces,
blockingRead == CL_TRUE,
di,
numEventsInWaitList,
eventWaitList,
event);
if (context->isProvidingPerformanceHints()) {
if (!isL3Capable(ptr, hostPtrSize)) {
context->providePerformanceHint(CL_CONTEXT_DIAGNOSTICS_LEVEL_BAD_INTEL, CL_ENQUEUE_READ_IMAGE_DOESNT_MEET_ALIGNMENT_RESTRICTIONS, ptr, hostPtrSize, MemoryConstants::pageSize, MemoryConstants::pageSize);
}
}
return CL_SUCCESS;
}
} // namespace OCLRT