Sergey Sharybin
1b4f0bf32a
Initial report was mentioning the Classroom demo scene, but this is probably because the scene was pre-configured to be used with OpenCL. Would expect any OpenCL compositing to be failing prior to this fix. The reason why crash was happening is due to OpenCL queue being released from OpenCLDevice destructor. Is not that obvious, but when Vector (including std::vector) is holding elements by value a destructor will be called on "old" memory when vector capacitance changes. Solved by making forbidding copy semantic for compositor devices and forcing move semantic to be used. Also use emplace semantic in the devices vector initialization.
285 lines
9.3 KiB
C++
285 lines
9.3 KiB
C++
/*
|
|
* This program is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU General Public License
|
|
* as published by the Free Software Foundation; either version 2
|
|
* of the License, or (at your option) any later version.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program; if not, write to the Free Software Foundation,
|
|
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
|
|
*
|
|
* Copyright 2011, Blender Foundation.
|
|
*/
|
|
|
|
#include "COM_OpenCLDevice.h"
|
|
#include "COM_WorkScheduler.h"
|
|
|
|
namespace blender::compositor {
|
|
|
|
enum COM_VendorID { NVIDIA = 0x10DE, AMD = 0x1002 };
|
|
const cl_image_format IMAGE_FORMAT_COLOR = {
|
|
CL_RGBA,
|
|
CL_FLOAT,
|
|
};
|
|
const cl_image_format IMAGE_FORMAT_VECTOR = {
|
|
CL_RGB,
|
|
CL_FLOAT,
|
|
};
|
|
const cl_image_format IMAGE_FORMAT_VALUE = {
|
|
CL_R,
|
|
CL_FLOAT,
|
|
};
|
|
|
|
OpenCLDevice::OpenCLDevice(cl_context context,
|
|
cl_device_id device,
|
|
cl_program program,
|
|
cl_int vendorId)
|
|
{
|
|
this->m_device = device;
|
|
this->m_context = context;
|
|
this->m_program = program;
|
|
this->m_queue = nullptr;
|
|
this->m_vendorID = vendorId;
|
|
|
|
cl_int error;
|
|
this->m_queue = clCreateCommandQueue(this->m_context, this->m_device, 0, &error);
|
|
}
|
|
|
|
OpenCLDevice::OpenCLDevice(OpenCLDevice &&other) noexcept
|
|
: m_context(other.m_context),
|
|
m_device(other.m_device),
|
|
m_program(other.m_program),
|
|
m_queue(other.m_queue),
|
|
m_vendorID(other.m_vendorID)
|
|
{
|
|
other.m_queue = nullptr;
|
|
}
|
|
|
|
OpenCLDevice::~OpenCLDevice()
|
|
{
|
|
if (this->m_queue) {
|
|
clReleaseCommandQueue(this->m_queue);
|
|
}
|
|
}
|
|
|
|
void OpenCLDevice::execute(WorkPackage *work_package)
|
|
{
|
|
const unsigned int chunkNumber = work_package->chunk_number;
|
|
ExecutionGroup *executionGroup = work_package->execution_group;
|
|
|
|
MemoryBuffer **inputBuffers = executionGroup->getInputBuffersOpenCL(chunkNumber);
|
|
MemoryBuffer *outputBuffer = executionGroup->allocateOutputBuffer(work_package->rect);
|
|
|
|
executionGroup->getOutputOperation()->executeOpenCLRegion(
|
|
this, &work_package->rect, chunkNumber, inputBuffers, outputBuffer);
|
|
|
|
delete outputBuffer;
|
|
|
|
executionGroup->finalizeChunkExecution(chunkNumber, inputBuffers);
|
|
}
|
|
cl_mem OpenCLDevice::COM_clAttachMemoryBufferToKernelParameter(cl_kernel kernel,
|
|
int parameterIndex,
|
|
int offsetIndex,
|
|
std::list<cl_mem> *cleanup,
|
|
MemoryBuffer **inputMemoryBuffers,
|
|
SocketReader *reader)
|
|
{
|
|
return COM_clAttachMemoryBufferToKernelParameter(kernel,
|
|
parameterIndex,
|
|
offsetIndex,
|
|
cleanup,
|
|
inputMemoryBuffers,
|
|
(ReadBufferOperation *)reader);
|
|
}
|
|
|
|
const cl_image_format *OpenCLDevice::determineImageFormat(MemoryBuffer *memoryBuffer)
|
|
{
|
|
switch (memoryBuffer->get_num_channels()) {
|
|
case 1:
|
|
return &IMAGE_FORMAT_VALUE;
|
|
break;
|
|
case 3:
|
|
return &IMAGE_FORMAT_VECTOR;
|
|
break;
|
|
case 4:
|
|
return &IMAGE_FORMAT_COLOR;
|
|
break;
|
|
default:
|
|
BLI_assert(!"Unsupported num_channels.");
|
|
}
|
|
|
|
return &IMAGE_FORMAT_COLOR;
|
|
}
|
|
|
|
cl_mem OpenCLDevice::COM_clAttachMemoryBufferToKernelParameter(cl_kernel kernel,
|
|
int parameterIndex,
|
|
int offsetIndex,
|
|
std::list<cl_mem> *cleanup,
|
|
MemoryBuffer **inputMemoryBuffers,
|
|
ReadBufferOperation *reader)
|
|
{
|
|
cl_int error;
|
|
|
|
MemoryBuffer *result = reader->getInputMemoryBuffer(inputMemoryBuffers);
|
|
|
|
const cl_image_format *imageFormat = determineImageFormat(result);
|
|
|
|
cl_mem clBuffer = clCreateImage2D(this->m_context,
|
|
CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR,
|
|
imageFormat,
|
|
result->getWidth(),
|
|
result->getHeight(),
|
|
0,
|
|
result->getBuffer(),
|
|
&error);
|
|
|
|
if (error != CL_SUCCESS) {
|
|
printf("CLERROR[%d]: %s\n", error, clewErrorString(error));
|
|
}
|
|
if (error == CL_SUCCESS) {
|
|
cleanup->push_back(clBuffer);
|
|
}
|
|
|
|
error = clSetKernelArg(kernel, parameterIndex, sizeof(cl_mem), &clBuffer);
|
|
if (error != CL_SUCCESS) {
|
|
printf("CLERROR[%d]: %s\n", error, clewErrorString(error));
|
|
}
|
|
|
|
COM_clAttachMemoryBufferOffsetToKernelParameter(kernel, offsetIndex, result);
|
|
return clBuffer;
|
|
}
|
|
|
|
void OpenCLDevice::COM_clAttachMemoryBufferOffsetToKernelParameter(cl_kernel kernel,
|
|
int offsetIndex,
|
|
MemoryBuffer *memoryBuffer)
|
|
{
|
|
if (offsetIndex != -1) {
|
|
cl_int error;
|
|
const rcti &rect = memoryBuffer->get_rect();
|
|
cl_int2 offset = {{rect.xmin, rect.ymin}};
|
|
|
|
error = clSetKernelArg(kernel, offsetIndex, sizeof(cl_int2), &offset);
|
|
if (error != CL_SUCCESS) {
|
|
printf("CLERROR[%d]: %s\n", error, clewErrorString(error));
|
|
}
|
|
}
|
|
}
|
|
|
|
void OpenCLDevice::COM_clAttachSizeToKernelParameter(cl_kernel kernel,
|
|
int offsetIndex,
|
|
NodeOperation *operation)
|
|
{
|
|
if (offsetIndex != -1) {
|
|
cl_int error;
|
|
cl_int2 offset = {{(cl_int)operation->getWidth(), (cl_int)operation->getHeight()}};
|
|
|
|
error = clSetKernelArg(kernel, offsetIndex, sizeof(cl_int2), &offset);
|
|
if (error != CL_SUCCESS) {
|
|
printf("CLERROR[%d]: %s\n", error, clewErrorString(error));
|
|
}
|
|
}
|
|
}
|
|
|
|
void OpenCLDevice::COM_clAttachOutputMemoryBufferToKernelParameter(cl_kernel kernel,
|
|
int parameterIndex,
|
|
cl_mem clOutputMemoryBuffer)
|
|
{
|
|
cl_int error;
|
|
error = clSetKernelArg(kernel, parameterIndex, sizeof(cl_mem), &clOutputMemoryBuffer);
|
|
if (error != CL_SUCCESS) {
|
|
printf("CLERROR[%d]: %s\n", error, clewErrorString(error));
|
|
}
|
|
}
|
|
|
|
void OpenCLDevice::COM_clEnqueueRange(cl_kernel kernel, MemoryBuffer *outputMemoryBuffer)
|
|
{
|
|
cl_int error;
|
|
const size_t size[] = {
|
|
(size_t)outputMemoryBuffer->getWidth(),
|
|
(size_t)outputMemoryBuffer->getHeight(),
|
|
};
|
|
|
|
error = clEnqueueNDRangeKernel(
|
|
this->m_queue, kernel, 2, nullptr, size, nullptr, 0, nullptr, nullptr);
|
|
if (error != CL_SUCCESS) {
|
|
printf("CLERROR[%d]: %s\n", error, clewErrorString(error));
|
|
}
|
|
}
|
|
|
|
void OpenCLDevice::COM_clEnqueueRange(cl_kernel kernel,
|
|
MemoryBuffer *outputMemoryBuffer,
|
|
int offsetIndex,
|
|
NodeOperation *operation)
|
|
{
|
|
cl_int error;
|
|
const int width = outputMemoryBuffer->getWidth();
|
|
const int height = outputMemoryBuffer->getHeight();
|
|
int offsetx;
|
|
int offsety;
|
|
int localSize = 1024;
|
|
size_t size[2];
|
|
cl_int2 offset;
|
|
|
|
if (this->m_vendorID == NVIDIA) {
|
|
localSize = 32;
|
|
}
|
|
|
|
bool breaked = false;
|
|
for (offsety = 0; offsety < height && (!breaked); offsety += localSize) {
|
|
offset.s[1] = offsety;
|
|
if (offsety + localSize < height) {
|
|
size[1] = localSize;
|
|
}
|
|
else {
|
|
size[1] = height - offsety;
|
|
}
|
|
|
|
for (offsetx = 0; offsetx < width && (!breaked); offsetx += localSize) {
|
|
if (offsetx + localSize < width) {
|
|
size[0] = localSize;
|
|
}
|
|
else {
|
|
size[0] = width - offsetx;
|
|
}
|
|
offset.s[0] = offsetx;
|
|
|
|
error = clSetKernelArg(kernel, offsetIndex, sizeof(cl_int2), &offset);
|
|
if (error != CL_SUCCESS) {
|
|
printf("CLERROR[%d]: %s\n", error, clewErrorString(error));
|
|
}
|
|
error = clEnqueueNDRangeKernel(
|
|
this->m_queue, kernel, 2, nullptr, size, nullptr, 0, nullptr, nullptr);
|
|
if (error != CL_SUCCESS) {
|
|
printf("CLERROR[%d]: %s\n", error, clewErrorString(error));
|
|
}
|
|
clFlush(this->m_queue);
|
|
if (operation->isBraked()) {
|
|
breaked = false;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
cl_kernel OpenCLDevice::COM_clCreateKernel(const char *kernelname,
|
|
std::list<cl_kernel> *clKernelsToCleanUp)
|
|
{
|
|
cl_int error;
|
|
cl_kernel kernel = clCreateKernel(this->m_program, kernelname, &error);
|
|
if (error != CL_SUCCESS) {
|
|
printf("CLERROR[%d]: %s\n", error, clewErrorString(error));
|
|
}
|
|
else {
|
|
if (clKernelsToCleanUp) {
|
|
clKernelsToCleanUp->push_back(kernel);
|
|
}
|
|
}
|
|
return kernel;
|
|
}
|
|
|
|
} // namespace blender::compositor
|