Revert "warp size fixes"

It seems like 32 is faster for me, at least and it won't cause so many conflicts.

This reverts commit 5d6eb72.

ggml-cuda.cu

@@ -182,11 +182,7 @@ typedef struct {
 } block_q6_k;
 static_assert(sizeof(block_q6_k) == sizeof(ggml_fp16_t) + 13*QK_K/16, "wrong q6_k block size/padding");
 
-#if defined(GGML_USE_HIPBLAS)
-#define WARP_SIZE warpSize
-#else
 #define WARP_SIZE 32
-#endif
 
 #define CUDA_MUL_BLOCK_SIZE 256
 
@@ -683,8 +679,8 @@ static __global__ void dequantize_mul_mat_vec(const void * vx, const float * y,
     // sum up partial sums and write back result
     __syncthreads();
 #pragma unroll
-    for (int mask = WARP_SIZE/2; mask > 0; mask >>= 1) {
-        tmp += __shfl_xor_sync(0xffffffff, tmp, mask, WARP_SIZE);
+    for (int mask = 16; mask > 0; mask >>= 1) {
+        tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
     }
 
     if (tid == 0) {