Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-02-05 16:10:42 +01:00
Code Issues Packages Projects Releases Wiki Activity

HIP: Supress transformation warning in softmax.cu

Browse Source 
loops with bounds not known at compile time can not be unrolled.
when ncols_template == 0, the bounds of the loop are not constexpr, thus llvm cant unroll the loops here.
This commit is contained in:
uvos 2025-01-28 23:06:32 +01:00 committed by GitHub
parent cae9fb4361
commit be5ef7963f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
ggml/src/ggml-cuda/softmax.cu
View File

@ -13,6 +13,12 @@ __device__ float __forceinline__ t2f32<half>(half val) {
return __half2float(val);
}
// When ncols_template == 0 the bounds for the loops in this function are not known and can't be unrolled.
// As we want to keep pragma unroll for all other cases we supress the clang transformation warning here.
#ifdef __clang__
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wpass-failed"
#endif
template <bool use_shared, int ncols_template, int block_size_template, typename T>
static __global__ void soft_max_f32(
const float * x, const T * mask, float * dst, const int ncols_par, const int nrows_y,
@ -118,6 +124,9 @@ static __global__ void soft_max_f32(
dst[col] = vals[col] * inv_sum;
}
}
#ifdef __clang__
#pragma clang diagnostic pop
#endif
static __global__ void soft_max_back_f32(
const float * grad, const float * dstf, float * dst, const int ncols, const float scale) {