Merge pull request #905 from LucQueen/outofbounds

fix array index out of bounds in kgetColRowStats

Merge pull request #905 from LucQueen/outofbounds
fix array index out of bounds in kgetColRowStats
3e706031 · Tim Dettmers · GitHub · 947db7cf · 51d30913 · 3e706031
Unverified Commit 3e706031 authored Jan 07, 2024 by Tim Dettmers Committed by GitHub Jan 07, 2024
Hide whitespace changes
Inline Side-by-side

Showing with 6 additions and 1 deletion

csrc/kernels.cu csrc/kernels.cu +6 -1

No files found.
--- a/csrc/kernels.cu
+++ b/csrc/kernels.cu
@@ -2178,7 +2178,12 @@ template<typename T, int THREADS, int ITEMS_PER_THREAD, int TILE_ROWS, int TILE_
  {
    //smem_col_absmax_values[threadIdx.x + (j*THREADS)] = -FLT_MAX;
    smem_row_absmax_values[threadIdx.x + (j*THREADS)] = -FLT_MAX;
-    smem_row_nnz_values[threadIdx.x + (j*THREADS)] = 0;
+    // smem_row_nnz_values[threadIdx.x + (j*THREADS)] = 0;
+  }
+
+  #pragma unroll TILE_ROWS
+  for (int j = 0; j < TILE_ROWS; j++) {
+    smem_row_nnz_values[j] = 0;
  }

  #pragma unroll ITEMS_PER_THREAD