CUDA: fix shared memory access condition for mmv (llama/10740)

diff --git a/ggml/src/ggml-cuda/mmv.cu b/ggml/src/ggml-cuda/mmv.cu
index cfe91f4283fa6f8f15d5b90245da64f61277f804..a4b4f6bc10d989ce18bfa551e67d49dbec4ba153 100644 (file)
--- a/ggml/src/ggml-cuda/mmv.cu
+++ b/ggml/src/ggml-cuda/mmv.cu
@@ -57,7 +57,7 @@ static __global__ void mul_mat_vec(
     if (block_size > WARP_SIZE) {
         buf_iw[tid/WARP_SIZE] = sumf;
         __syncthreads();
-        if (tid > WARP_SIZE) {
+        if (tid >= WARP_SIZE) {
             return;
         }
         sumf = buf_iw[tid];