Refs EA31337#738. Matrix multiplication via OpenCL. Ready for testing.

nseam · May 1, 2024 · ad24e5b · ad24e5b
1 parent dce2921
commit ad24e5b
Show file tree

Hide file tree

Showing 5 changed files with 148 additions and 98 deletions.
diff --git a/Matrix.matmul.naive.cl b/Matrix.matmul.naive.cl
@@ -1,23 +1,14 @@
 #pragma OPENCL EXTENSION cl_khr_fp64 : enable
 
-__kernel void matmul(
-  __global double* A,
-  __global double* B,
-  __global double* C,
-  int rowsA,
-  int colsA,
-  int colsB
-  )
-{
-  int row = get_global_id(0);
-  int col = get_global_id(1);
-
-  double sum = 0.0;
-
-  for(int k = 0; k < colsA; ++k) {
-    sum += A[row * colsA + k] * B[k * colsB + col];
-    //sum += col;
+__kernel void matmul(__global double* A, __global double* B, __global double* C, int rowsA, int colsA, int colsB) {
+  int i = get_global_id(0);
+  int j = get_global_id(1);
+
+  if (i < rowsA && j < colsB) {
+    float sum = 0.0f;
+    for (int k = 0; k < colsA; ++k) {
+      sum += A[i * colsA + k] * B[k * colsB + j];
+    }
+    C[i * colsB + j] = sum;
   }
-
-  C[row * colsB + col] = sum;
 }