archive/rpcs3
1
0
Fork 0
mirror of https://github.com/RPCS3/rpcs3.git synced 2025-07-14 02:38:37 +12:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

rsx: Some shader compiler threads tuning

Browse source 
- Allow more threads for wide CPUs
- Simplify 'auto' selection a bit
This commit is contained in:
kd-11 2020-11-14 14:58:51 +03:00 • committed by kd-11
parent 14358d5e84
commit cab4c78b7b
4 changed files with 16 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

4
rpcs3/Emu/RSX/GL/GLGSRender.cpp
View file

@ -76,9 +76,7 @@ void GLGSRender::on_init_thread()
m_frame->delete_context(ctx); m_frame->delete_context(ctx);
}; };
int thread_count = g_cfg.video.shader_compiler_threads_count; gl::initialize_pipe_compiler(context_create_func, context_bind_func, context_destroy_func, g_cfg.video.shader_compiler_threads_count);
if (!thread_count) thread_count = -1;
gl::initialize_pipe_compiler(context_create_func, context_bind_func, context_destroy_func, thread_count);
} }
else else
{ {

10
rpcs3/Emu/RSX/GL/GLPipelineCompiler.cpp
View file

@ -94,15 +94,19 @@ namespace gl
std::function<void(draw_context_t)> context_destroy_func, std::function<void(draw_context_t)> context_destroy_func,
int num_worker_threads) int num_worker_threads)
{ {
if (num_worker_threads == -1) if (num_worker_threads == 0)
{ {
// Select optimal number of compiler threads // Select optimal number of compiler threads
const auto hw_threads = std::thread::hardware_concurrency(); const auto hw_threads = std::thread::hardware_concurrency();
if (hw_threads >= 12) if (hw_threads > 12)
{
num_worker_threads = 6;
}
else if (hw_threads > 8)
{ {
num_worker_threads = 4; num_worker_threads = 4;
} }
else if (hw_threads >= 8) else if (hw_threads == 8)
{ {
num_worker_threads = 2; num_worker_threads = 2;
} }

4
rpcs3/Emu/RSX/VK/VKGSRender.cpp
View file

@ -473,10 +473,8 @@ VKGSRender::VKGSRender() : GSRender()
null_buffer = std::make_unique<vk::buffer>(*m_device, 32, memory_map.device_local, VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT, VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT, 0); null_buffer = std::make_unique<vk::buffer>(*m_device, 32, memory_map.device_local, VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT, VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT, 0);
null_buffer_view = std::make_unique<vk::buffer_view>(*m_device, null_buffer->value, VK_FORMAT_R8_UINT, 0, 32); null_buffer_view = std::make_unique<vk::buffer_view>(*m_device, null_buffer->value, VK_FORMAT_R8_UINT, 0, 32);
int thread_count = g_cfg.video.shader_compiler_threads_count;
if (!thread_count) thread_count = -1;
vk::initialize_compiler_context(); vk::initialize_compiler_context();
vk::initialize_pipe_compiler(thread_count); vk::initialize_pipe_compiler(g_cfg.video.shader_compiler_threads_count);
if (g_cfg.video.overlay) if (g_cfg.video.overlay)
{ {

10
rpcs3/Emu/RSX/VK/VKPipelineCompiler.cpp
View file

@ -182,15 +182,19 @@ namespace vk
void initialize_pipe_compiler(int num_worker_threads) void initialize_pipe_compiler(int num_worker_threads)
{ {
if (num_worker_threads == -1) if (num_worker_threads == 0)
{ {
// Select optimal number of compiler threads // Select optimal number of compiler threads
const auto hw_threads = std::thread::hardware_concurrency(); const auto hw_threads = std::thread::hardware_concurrency();
if (hw_threads >= 12) if (hw_threads > 12)
{
num_worker_threads = 6;
}
else if (hw_threads > 8)
{ {
num_worker_threads = 4; num_worker_threads = 4;
} }
else if (hw_threads >= 8) else if (hw_threads == 8)
{ {
num_worker_threads = 2; num_worker_threads = 2;
} }