vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

cuda : uint -> uint32_t

Browse source
This commit is contained in:
Georgi Gerganov 2024-04-22 19:12:06 +03:00
parent f725ca90fb
commit 5408d55506
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
2 changed files with 9 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ggml-cuda/common.cuh
View file

@ -307,9 +307,9 @@ static __device__ __forceinline__ half2 warp_reduce_max(half2 x) {
}
#if CUDART_VERSION < 12000
static __device__ __forceinline__ uint __hgt2_mask(const half2 a, const half2 b) {
const uint mask_low = 0x0000FFFF * ( __low2half(a) > __low2half(b));
const uint mask_high = 0xFFFF0000 * (__high2half(a) > __high2half(b));
static __device__ __forceinline__ uint32_t __hgt2_mask(const half2 a, const half2 b) {
const uint32_t mask_low = 0x0000FFFF * ( __low2half(a) > __low2half(b));
const uint32_t mask_high = 0xFFFF0000 * (__high2half(a) > __high2half(b));
return mask_low | mask_high;
}
#endif // CUDART_VERSION < 12000