tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-07-29 13:43:38 -04:00
Code Issues Packages Projects Releases Wiki Activity

musa: fix compilation warnings in mp_22/31 (#12780)

Browse Source 
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
This commit is contained in:
R0CKSTAR
2025-04-06 21:23:54 +08:00
committed by GitHub
parent 0c74b04376
commit 916c83bfe7
4 changed files with 35 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
ggml/src/ggml-cuda/fattn-common.cuh
View File

@@ -62,7 +62,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_0(
T sum = 0.0f;
#pragma unroll
for (int k_KQ_0 = 0; k_KQ_0 < D/sizeof(int); k_KQ_0 += warp_size) {
for (int k_KQ_0 = 0; k_KQ_0 < int(D/sizeof(int)); k_KQ_0 += warp_size) {
const int k_KQ = k_KQ_0 + threadIdx.x;
const int ib = k_KQ / QI8_1;
@@ -102,7 +102,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_1(
T sum = 0.0f;
#pragma unroll
for (int k_KQ_0 = 0; k_KQ_0 < D/sizeof(int); k_KQ_0 += warp_size) {
for (int k_KQ_0 = 0; k_KQ_0 < int(D/sizeof(int)); k_KQ_0 += warp_size) {
const int k_KQ = k_KQ_0 + threadIdx.x;
const int ib = k_KQ / QI8_1;
@@ -146,7 +146,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_0(
T sum = 0.0f;
#pragma unroll
for (int k_KQ_0 = 0; k_KQ_0 < D/sizeof(int); k_KQ_0 += warp_size) {
for (int k_KQ_0 = 0; k_KQ_0 < int(D/sizeof(int)); k_KQ_0 += warp_size) {
const int k_KQ = k_KQ_0 + threadIdx.x;
const int ib = k_KQ / QI8_1;
@@ -193,7 +193,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_1(
T sum = 0.0f;
#pragma unroll
for (int k_KQ_0 = 0; k_KQ_0 < D/sizeof(int); k_KQ_0 += warp_size) {
for (int k_KQ_0 = 0; k_KQ_0 < int(D/sizeof(int)); k_KQ_0 += warp_size) {
const int k_KQ = k_KQ_0 + threadIdx.x;
const int ib = k_KQ / QI8_1;
@@ -244,7 +244,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q8_0(
T sum = 0.0f;
#pragma unroll
for (int k_KQ_0 = 0; k_KQ_0 < D/sizeof(int); k_KQ_0 += warp_size) {
for (int k_KQ_0 = 0; k_KQ_0 < int(D/sizeof(int)); k_KQ_0 += warp_size) {
const int k_KQ = k_KQ_0 + threadIdx.x;
const int ib = k_KQ / QI8_0;