From: Eve <redacted>
Date: Sat, 4 Oct 2025 20:04:27 +0000 (+0000)
Subject: vulkan: use a more appropriate amount of threads when generating shaders (#16418)
X-Git-Tag: upstream/0.0.6764~74
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=86df2c9ae4f2f1ee63d2558a9dc797b98524639b;p=pkg%2Fggml%2Fsources%2Fllama.cpp

vulkan: use a more appropriate amount of threads when generating shaders (#16418)

* use a more flexible amount of threads

* fix windows compile and 0 thread case

* nominmax
---

diff --git a/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp b/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp
index e2726f1f..f0cc24ff 100644
--- a/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp
+++ b/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp
@@ -1,5 +1,3 @@
-
-
 #include <iostream>
 #include <fstream>
 #include <sstream>
@@ -22,6 +20,7 @@
 #include <sys/types.h>
 
 #ifdef _WIN32
+    #define NOMINMAX
     #include <windows.h>
     #include <direct.h> // For _mkdir on Windows
 #else
@@ -306,7 +305,7 @@ using compile_count_guard = std::unique_ptr<uint32_t, decltype(&decrement_compil
 compile_count_guard acquire_compile_slot() {
     // wait until fewer than N compiles are in progress.
     // 16 is an arbitrary limit, the goal is to avoid "failed to create pipe" errors.
-    uint32_t N = 16;
+    uint32_t N = std::max(1u, std::min(16u, std::thread::hardware_concurrency()));
     std::unique_lock<std::mutex> guard(compile_count_mutex);
     compile_count_cond.wait(guard, [N] { return compile_count < N; });
     compile_count++;