Added testcase for autogeneration of visc IR

llvm/test/VISC/MatrixMultiplication/auto_gemm.c

+#include <stdlib.h>
+#include <stdio.h>
+#include <math.h>
+#include <string.h>
+
+#define WA 1024
+#define HA 1024
+#define WB 1024
+#define HB WA
+#define WC WB
+#define HC HA
+
+
+
+// Thread block size
+#define BLOCK_SIZE 16
+
+// Allocates a matrix with random float entries.
+void randomInit(float* data, int size) {
+  for (int i = 0; i < size; ++i)
+    data[i] = rand() / (float)RAND_MAX;
+}
+
+//////////////////////////////////////////////////////////////////////////////
+//! Loads a Program file.
+//!
+//! @return the source string if succeeded, 0 otherwise
+//! @param cFilename        program filename
+//! @param szFinalLength    returned length of the code string
+//////////////////////////////////////////////////////////////////////////////
+
+// Check bool
+int isEqual(float a, float b) {
+  return (fabs(a-b) < 0.001);
+}
+
+// Check Results
+
+__attribute__ ((noinline)) int checkResults(float* A, float* B, float* C) {
+  unsigned int size_A = WA * HA;
+  unsigned int size_B = WB * HB;
+  unsigned int size_C = WC * HC;
+  unsigned int bytesC = sizeof(float) * size_C;
+  float* goldC = (float*) malloc(bytesC);
+  for (int i=0; i < HC; i++) {
+    for (int j=0; j < WC; j++) {
+      goldC[i*WC + j] = 0;
+      for (int k=0; k < HB; k++) {
+        goldC[i*WC + j] += A[i*WA + k] * B[k*WB + j];
+      }
+      if(!isEqual(goldC[i*WC + j], C[i*WC + j])) {
+        printf("Mismatch at %d,%d --- C = %f and goldC = %f\n", i, j, C[i*WC+j], goldC[i*WC+j]);
+        return 0;
+      }
+    }
+  }
+  return 1; // Success
+}
+
+// Dummy visc node execution call
+//void __visc__node(void kernel (float*, float*, float*, unsigned, unsigned), int numDims, void* dims, int numInputs, void* inputs, int numOutputs, void* outputs);
+
+void matrixMul(float* A, float* B, float* C, unsigned k, unsigned n) {
+
+  printf("Entered function\n");
+  int tx = get_global_id(0); //2D Global Thread ID x
+  int ty = get_global_id(1); //2D Global Thread ID y
+  //int tx = get_global_id(0); //2D Global Thread ID x
+  //int ty = get_global_id(1); //2D Global Thread ID y
+
+  printf("Computing element (%d, %d)\n", tx, ty);
+  // Initialize accumulator
+  float res = 0.0f;
+
+  // Perform dot-product of row-column
+  for (int i = 0; i < k; i++) {
+    printf("Accessing k = %d, A[%d], B[%d]\n", k, ty*k+i, i*n+tx);
+    res += A[ty*k+i] * B[i*n+tx];
+  }
+
+  printf("Result computed\n");
+  // Write in device memory
+  C[ty*n+tx] = res;
+
+  printf("Result written to C\n");
+}
+
+
+// Main
+int main(int argc, char** argv) {
+
+  // seed for rand()
+  srand(2006);
+
+  // Allocate host memory for matrices A and B
+  unsigned int size_A = WA * HA;
+  unsigned int bytes_A = sizeof(float) * size_A;
+  float* h_A = (float*) malloc(bytes_A);
+
+  unsigned int size_B = WB * HB;
+  unsigned int bytes_B = sizeof(float) * size_B;
+  float* h_B = (float*) malloc(bytes_B);
+
+   // Initialize host memory
+   randomInit(h_A, size_A);
+   randomInit(h_B, size_B);
+
+/*
+   // Print A and B
+   printf("\n\nMatrix A\n");
+   for(int i = 0; i < size_A; i++)
+   {
+      printf("%f ", h_A[i]);
+      if(((i + 1) % WA) == 0)
+      printf("\n");
+   }
+
+   printf("\n\nMatrix B\n");
+   for(int i = 0; i < size_B; i++)
+   {
+      printf("%f ", h_B[i]);
+      if(((i + 1) % WB) == 0)
+      printf("\n");
+   }
+*/
+
+  // Allocate host memory for the result matrix C
+  unsigned int size_C = WC * HC;
+  unsigned int bytes_C = sizeof(float) * size_C;
+  float* h_C = (float*) malloc(bytes_C);
+
+   // Compute using OpenCL
+  //matrixMul(h_A, h_B, h_C, WA, WB);
+  //__visc__node(matrixMul, 2, WB, HA, 3, h_A, h_B, h_C, 0);
+  __visc__node(matrixMul, 2, WB, HA, 5, h_A, h_B, h_C, WA, WB, 0);
+  if(checkResults(h_A, h_B, h_C))
+    printf("\nPass!\n");
+  else
+    printf("\nFailed!\n");
+  printf("\nDone!\n");
+
+  // Deallocate memory
+  free(h_A);
+  free(h_B);
+  free(h_C);
+}
+