nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 25 Packages Projects Releases Wiki Activity

vulkan: Workaround compiler bug in conv2d coopmat2 path (#24924)

Browse Source 
* vulkan: Workaround compiler bug in conv2d coopmat2 path

* apply same workaround to CONV_3D

* Apply suggestion from @jeffbolznv
This commit is contained in:
Jeff Bolz
2026-06-26 11:53:32 +02:00
committed by GitHub
parent 2f18fe13c5
commit 9df06805ee
3 changed files with 5 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/vulkan-shaders/conv2d_mm.comp
+1 -1
View File
@@ -158,7 +158,7 @@ const uint32_t Csh_stride = BS_NPQ;
#ifdef COOPMAT #ifdef COOPMAT
const uint32_t Csh_len = BS_K * Csh_stride; const uint32_t Csh_len = BS_K * Csh_stride;
#else #else
const uint32_t Csh_len = csh_store != 0 ? BS_K * Csh_stride : 1; const uint32_t Csh_len = csh_store != 0 ? BS_K * Csh_stride : 8; // 8 to workaround compiler bug
#endif #endif
shared SHMEM_TYPE Csh[Csh_len]; // K x NPQ shared SHMEM_TYPE Csh[Csh_len]; // K x NPQ
#endif #endif
ggml/src/ggml-vulkan/vulkan-shaders/conv3d_mm.comp
+1 -1
View File
@@ -144,7 +144,7 @@ const uint32_t Csh_stride = BS_NPQ;
#ifdef COOPMAT #ifdef COOPMAT
const uint32_t Csh_len = BS_K * Csh_stride; const uint32_t Csh_len = BS_K * Csh_stride;
#else #else
const uint32_t Csh_len = csh_store != 0 ? BS_K * Csh_stride : 1; const uint32_t Csh_len = csh_store != 0 ? BS_K * Csh_stride : 8; // 8 to workaround compiler bug
#endif #endif
shared SHMEM_TYPE Csh[Csh_len]; // K x NPQ shared SHMEM_TYPE Csh[Csh_len]; // K x NPQ
#endif #endif
tests/test-backend-ops.cpp
+3
View File
@@ -7973,6 +7973,9 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
} }
} }
} }
for (auto kernel_type : {GGML_TYPE_F32, GGML_TYPE_F16}) {
test_cases.emplace_back(new test_conv_2d({ 256, 256, 192, 1 }, { 3, 3, 192, 96 }, kernel_type, 1, 1, 1, 1, 1, 1, false));
}
// sycl backend will limit task global_range < MAX_INT // sycl backend will limit task global_range < MAX_INT
// test cases for 2D im2col with large input W and H (occurs in stable-diffusion) // test cases for 2D im2col with large input W and H (occurs in stable-diffusion)