mirror of
https://github.com/blender/blender
synced 2026-09-27 01:34:15 +03:00
bfb5775388
Local partitioned shader sorting has been used with Metal and oneAPI for a while now. Turns out CUDA/OptiX also benefit, so this change enables it there too. Note that there is no need to operate on shared memory with atomics in CUDA, so the implementation of atomic_store_local/atomic_load_local is kept extremely simple. Pull Request: https://projects.blender.org/blender/blender/pulls/163436
128 lines
2.2 KiB
C++
128 lines
2.2 KiB
C++
/* SPDX-FileCopyrightText: 2011-2022 Blender Foundation
|
|
*
|
|
* SPDX-License-Identifier: Apache-2.0 */
|
|
|
|
#include "util/debug.h"
|
|
|
|
#include <cstdlib>
|
|
|
|
#include "util/log.h"
|
|
|
|
CCL_NAMESPACE_BEGIN
|
|
|
|
DebugFlags::CPU::CPU()
|
|
{
|
|
reset();
|
|
}
|
|
|
|
void DebugFlags::CPU::reset()
|
|
{
|
|
#define STRINGIFY(x) #x
|
|
#define CHECK_CPU_FLAGS(flag, env) \
|
|
do { \
|
|
flag = (getenv(env) == nullptr); \
|
|
if (!flag) { \
|
|
LOG_INFO << "Disabling " << STRINGIFY(flag) << " instruction set."; \
|
|
} \
|
|
} while (0)
|
|
|
|
CHECK_CPU_FLAGS(avx2, "CYCLES_CPU_NO_AVX2");
|
|
|
|
#undef STRINGIFY
|
|
#undef CHECK_CPU_FLAGS
|
|
|
|
bvh_layout = BVH_LAYOUT_AUTO;
|
|
}
|
|
|
|
DebugFlags::CUDA::CUDA()
|
|
{
|
|
reset();
|
|
}
|
|
|
|
DebugFlags::HIP::HIP()
|
|
{
|
|
reset();
|
|
}
|
|
|
|
DebugFlags::Metal::Metal()
|
|
{
|
|
reset();
|
|
}
|
|
|
|
void DebugFlags::CUDA::reset()
|
|
{
|
|
if (getenv("CYCLES_CUDA_ADAPTIVE_COMPILE") != nullptr) {
|
|
adaptive_compile = true;
|
|
}
|
|
|
|
if (const char *str = getenv("CYCLES_CUDA_LOCAL_ATOMIC_SORT")) {
|
|
use_local_atomic_sort = (atoi(str) != 0);
|
|
}
|
|
}
|
|
|
|
void DebugFlags::HIP::reset()
|
|
{
|
|
if (getenv("CYCLES_HIP_ADAPTIVE_COMPILE") != nullptr) {
|
|
adaptive_compile = true;
|
|
}
|
|
}
|
|
|
|
void DebugFlags::Metal::reset()
|
|
{
|
|
if (getenv("CYCLES_METAL_ADAPTIVE_COMPILE") != nullptr) {
|
|
adaptive_compile = true;
|
|
}
|
|
|
|
if (const char *str = getenv("CYCLES_METAL_LOCAL_ATOMIC_SORT")) {
|
|
use_local_atomic_sort = (atoi(str) != 0);
|
|
}
|
|
|
|
if (const char *str = getenv("CYCLES_METAL_NANOVDB")) {
|
|
use_nanovdb = (atoi(str) != 0);
|
|
}
|
|
|
|
if (const char *str = getenv("CYCLES_METAL_ASYNC_PSO_CREATION")) {
|
|
use_async_pso_creation = (atoi(str) != 0);
|
|
}
|
|
|
|
if (const char *str = getenv("CYCLES_METALRT_PCMI")) {
|
|
use_metalrt_pcmi = (atoi(str) != 0);
|
|
}
|
|
|
|
if (const char *str = getenv("CYCLES_METAL_RESIDENCY_SETS")) {
|
|
use_residency_sets_if_available = (atoi(str) != 0);
|
|
}
|
|
}
|
|
|
|
DebugFlags::TextureCache::TextureCache()
|
|
{
|
|
reset();
|
|
}
|
|
|
|
void DebugFlags::TextureCache::reset()
|
|
{
|
|
use_eviction = true;
|
|
preserve_unused = 0;
|
|
min_tile_size = 64;
|
|
}
|
|
|
|
DebugFlags::OptiX::OptiX()
|
|
{
|
|
reset();
|
|
}
|
|
|
|
void DebugFlags::OptiX::reset()
|
|
{
|
|
use_debug = false;
|
|
}
|
|
|
|
void DebugFlags::reset()
|
|
{
|
|
cpu.reset();
|
|
cuda.reset();
|
|
optix.reset();
|
|
metal.reset();
|
|
texture_cache.reset();
|
|
}
|
|
|
|
CCL_NAMESPACE_END
|