upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-10-02 02:47:26 -05:00
Code Issues Packages Projects Releases Wiki Activity

opencl: mark vec subgroup bcast as supproted for Adreno E17 compiler (#29698)

Browse Source
This commit is contained in:
lhez
2026-10-01 08:33:23 +03:00
committed by GitHub
parent db33d3cb89
commit 3ec4df42d9
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-opencl/ggml-opencl.cpp
+2 -1
View File
@@ -6700,7 +6700,8 @@ static ggml_backend_opencl_context * ggml_cl_init(ggml_backend_dev_t dev) {
backend_ctx->adreno_cl_compiler_version = get_adreno_cl_compiler_version(driver_version.data());
backend_ctx->has_vector_subgroup_broadcast =
(backend_ctx->adreno_cl_compiler_version.type == E031 && backend_ctx->adreno_cl_compiler_version.major >= 47) ||
(backend_ctx->adreno_cl_compiler_version.type == DX && backend_ctx->adreno_cl_compiler_version.major >= 17);
(backend_ctx->adreno_cl_compiler_version.type == DX && backend_ctx->adreno_cl_compiler_version.major >= 17) ||
(backend_ctx->adreno_cl_compiler_version.type == E17);
// The q6_K flat mul_mat miscompile is a defect of the older E031 compilers, not a
// property of any GPU generation: it reproduces on E031.38 (Adreno 642L) and E031.41