Differential D749 Diff 2456 intern/cycles/device/device_cuda.cpp

Changeset View

Standalone View

intern/cycles/device/device_cuda.cpp

Show First 20 Lines • Show All 670 Lines • ▼ Show 20 Lines	void shader(DeviceTask& task)
else {		else {
cuda_assert(cuModuleGetFunction(&cuShader, cuModule, "kernel_cuda_shader"));		cuda_assert(cuModuleGetFunction(&cuShader, cuModule, "kernel_cuda_shader"));
}		}

/* do tasks in smaller chunks, so we can cancel it */		/* do tasks in smaller chunks, so we can cancel it */
const int shader_chunk_size = 65536;		const int shader_chunk_size = 65536;
const int start = task.shader_x;		const int start = task.shader_x;
const int end = task.shader_x + task.shader_w;		const int end = task.shader_x + task.shader_w;
		int offset = task.offset;

bool canceled = false;		bool canceled = false;
for(int sample = 0; sample < task.num_samples && !canceled; sample++) {		for(int sample = 0; sample < task.num_samples && !canceled; sample++) {
for(int shader_x = start; shader_x < end; shader_x += shader_chunk_size) {		for(int shader_x = start; shader_x < end; shader_x += shader_chunk_size) {
int shader_w = min(shader_chunk_size, end - shader_x);		int shader_w = min(shader_chunk_size, end - shader_x);

/* pass in parameters */		/* pass in parameters */
void *args[] = {&d_input,		void *args[] = {&d_input,
&d_output,		&d_output,
&task.shader_eval_type,		&task.shader_eval_type,
&shader_x,		&shader_x,
&shader_w,		&shader_w,
		&offset,
&sample};		&sample};

/* launch kernel */		/* launch kernel */
int threads_per_block;		int threads_per_block;
cuda_assert(cuFuncGetAttribute(&threads_per_block, CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK, cuShader));		cuda_assert(cuFuncGetAttribute(&threads_per_block, CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK, cuShader));

int xblocks = (shader_w + threads_per_block - 1)/threads_per_block;		int xblocks = (shader_w + threads_per_block - 1)/threads_per_block;

▲ Show 20 Lines • Show All 381 Lines • Show Last 20 Lines