blob: 7fd6b9d9e6a280287be448fb7e271a3cbf0116f6 [file] [log] [blame]
/**
* gemver.c: This file is part of the PolyBench/C 3.2 test suite.
*
*
* Contact: Louis-Noel Pouchet <pouchet@cse.ohio-state.edu>
* Web address: http://polybench.sourceforge.net
*/
#include <stdio.h>
#include <unistd.h>
#include <string.h>
#include <math.h>
/* Include polybench common header. */
#include <polybench.h>
/* Include benchmark-specific header. */
/* Default data type is double, default size is 4000. */
#include "gemver.h"
/* Array initialization. */
static
void init_array (int n,
DATA_TYPE *alpha,
DATA_TYPE *beta,
DATA_TYPE POLYBENCH_2D(A,N,N,n,n),
DATA_TYPE POLYBENCH_1D(u1,N,n),
DATA_TYPE POLYBENCH_1D(v1,N,n),
DATA_TYPE POLYBENCH_1D(u2,N,n),
DATA_TYPE POLYBENCH_1D(v2,N,n),
DATA_TYPE POLYBENCH_1D(w,N,n),
#if !FMA_DISABLED
DATA_TYPE POLYBENCH_1D(w_StrictFP,N,n),
#endif
DATA_TYPE POLYBENCH_1D(y,N,n),
DATA_TYPE POLYBENCH_1D(z,N,n))
{
#pragma STDC FP_CONTRACT OFF
int i, j;
*alpha = 43532;
*beta = 12313;
for (i = 0; i < n; i++)
{
u1[i] = i;
u2[i] = (i+1)/n/2.0;
v1[i] = (i+1)/n/4.0;
v2[i] = (i+1)/n/6.0;
y[i] = (i+1)/n/8.0;
z[i] = (i+1)/n/9.0;
#if !FMA_DISABLED
w_StrictFP[i] = w[i] = 0.0;
#endif
for (j = 0; j < n; j++)
A[i][j] = ((DATA_TYPE) i*j) / n;
}
}
/* DCE code. Must scan the entire live-out data.
Can be used also to check the correctness of the output. */
static
void print_array(int n,
DATA_TYPE POLYBENCH_1D(w,N,n))
{
int i;
for (i = 0; i < n; i++) {
fprintf (stderr, DATA_PRINTF_MODIFIER, w[i]);
if (i % 20 == 0) fprintf (stderr, "\n");
}
}
/* Main computational kernel. The whole function will be timed,
including the call and return. */
static
void kernel_gemver(int n,
DATA_TYPE alpha,
DATA_TYPE beta,
DATA_TYPE POLYBENCH_2D(A,N,N,n,n),
DATA_TYPE POLYBENCH_2D(A_tmp,N,N,n,n),
DATA_TYPE POLYBENCH_1D(u1,N,n),
DATA_TYPE POLYBENCH_1D(v1,N,n),
DATA_TYPE POLYBENCH_1D(u2,N,n),
DATA_TYPE POLYBENCH_1D(v2,N,n),
DATA_TYPE POLYBENCH_1D(w,N,n),
DATA_TYPE POLYBENCH_1D(x,N,n),
DATA_TYPE POLYBENCH_1D(y,N,n),
DATA_TYPE POLYBENCH_1D(z,N,n))
{
int i, j;
#pragma scop
for (i = 0; i < _PB_N; i++)
for (j = 0; j < _PB_N; j++)
A_tmp[i][j] = A[i][j] + u1[i] * v1[j] + u2[i] * v2[j];
for (i = 0; i < _PB_N; i++) {
x[i] = 0.0;
for (j = 0; j < _PB_N; j++)
x[i] = x[i] + beta * A_tmp[j][i] * y[j];
}
for (i = 0; i < _PB_N; i++)
x[i] = x[i] + z[i];
for (i = 0; i < _PB_N; i++)
for (j = 0; j < _PB_N; j++)
w[i] = w[i] + alpha * A_tmp[i][j] * x[j];
#pragma endscop
}
#if !FMA_DISABLED
// NOTE: FMA_DISABLED is true for targets where FMA contraction causes
// discrepancies which cause the accuracy checks to fail.
// In this case, the test runs with the option -ffp-contract=off
static
void kernel_gemver_StrictFP(int n,
DATA_TYPE alpha,
DATA_TYPE beta,
DATA_TYPE POLYBENCH_2D(A,N,N,n,n),
DATA_TYPE POLYBENCH_2D(A_tmp,N,N,n,n),
DATA_TYPE POLYBENCH_1D(u1,N,n),
DATA_TYPE POLYBENCH_1D(v1,N,n),
DATA_TYPE POLYBENCH_1D(u2,N,n),
DATA_TYPE POLYBENCH_1D(v2,N,n),
DATA_TYPE POLYBENCH_1D(w,N,n),
DATA_TYPE POLYBENCH_1D(x,N,n),
DATA_TYPE POLYBENCH_1D(y,N,n),
DATA_TYPE POLYBENCH_1D(z,N,n))
{
#pragma STDC FP_CONTRACT OFF
int i, j;
for (i = 0; i < _PB_N; i++)
for (j = 0; j < _PB_N; j++)
A_tmp[i][j] = A[i][j] + u1[i] * v1[j] + u2[i] * v2[j];
for (i = 0; i < _PB_N; i++) {
x[i] = 0.0;
for (j = 0; j < _PB_N; j++)
x[i] = x[i] + beta * A_tmp[j][i] * y[j];
}
for (i = 0; i < _PB_N; i++)
x[i] = x[i] + z[i];
for (i = 0; i < _PB_N; i++)
for (j = 0; j < _PB_N; j++)
w[i] = w[i] + alpha * A_tmp[i][j] * x[j];
}
/* Return 0 when one of the elements of arrays A and B do not match within the
allowed FP_ABSTOLERANCE. Return 1 when all elements match. */
static int
check_FP(int n,
DATA_TYPE POLYBENCH_1D(A,N,n),
DATA_TYPE POLYBENCH_1D(B,N,n)) {
int i;
double AbsTolerance = FP_ABSTOLERANCE;
for (i = 0; i < _PB_N; i++)
{
double V1 = A[i];
double V2 = B[i];
double Diff = fabs(V1 - V2);
if (Diff > AbsTolerance) {
fprintf(stderr, "A[%d] = %lf and B[%d] = %lf differ more than"
" FP_ABSTOLERANCE = %lf\n", i, V1, i, V2, AbsTolerance);
return 0;
}
}
return 1;
}
#endif
int main(int argc, char** argv)
{
/* Retrieve problem size. */
int n = N;
/* Variable declaration/allocation. */
DATA_TYPE alpha;
DATA_TYPE beta;
POLYBENCH_2D_ARRAY_DECL(A, DATA_TYPE, N, N, n, n);
POLYBENCH_2D_ARRAY_DECL(A_tmp, DATA_TYPE, N, N, n, n);
POLYBENCH_1D_ARRAY_DECL(u1, DATA_TYPE, N, n);
POLYBENCH_1D_ARRAY_DECL(v1, DATA_TYPE, N, n);
POLYBENCH_1D_ARRAY_DECL(u2, DATA_TYPE, N, n);
POLYBENCH_1D_ARRAY_DECL(v2, DATA_TYPE, N, n);
POLYBENCH_1D_ARRAY_DECL(w, DATA_TYPE, N, n);
#if !FMA_DISABLED
POLYBENCH_1D_ARRAY_DECL(w_StrictFP, DATA_TYPE, N, n);
#endif
POLYBENCH_1D_ARRAY_DECL(x, DATA_TYPE, N, n);
POLYBENCH_1D_ARRAY_DECL(y, DATA_TYPE, N, n);
POLYBENCH_1D_ARRAY_DECL(z, DATA_TYPE, N, n);
/* Initialize array(s). */
init_array (n, &alpha, &beta,
POLYBENCH_ARRAY(A),
POLYBENCH_ARRAY(u1),
POLYBENCH_ARRAY(v1),
POLYBENCH_ARRAY(u2),
POLYBENCH_ARRAY(v2),
POLYBENCH_ARRAY(w),
#if !FMA_DISABLED
POLYBENCH_ARRAY(w_StrictFP),
#endif
POLYBENCH_ARRAY(y),
POLYBENCH_ARRAY(z));
/* Start timer. */
polybench_start_instruments;
/* Run kernel. */
kernel_gemver (n, alpha, beta,
POLYBENCH_ARRAY(A),
POLYBENCH_ARRAY(A_tmp),
POLYBENCH_ARRAY(u1),
POLYBENCH_ARRAY(v1),
POLYBENCH_ARRAY(u2),
POLYBENCH_ARRAY(v2),
POLYBENCH_ARRAY(w),
POLYBENCH_ARRAY(x),
POLYBENCH_ARRAY(y),
POLYBENCH_ARRAY(z));
/* Stop and print timer. */
polybench_stop_instruments;
polybench_print_instruments;
#if FMA_DISABLED
/* Prevent dead-code elimination. All live-out data must be printed
by the function call in argument. */
polybench_prevent_dce(print_array(n, POLYBENCH_ARRAY(w)));
#else
kernel_gemver_StrictFP(n, alpha, beta,
POLYBENCH_ARRAY(A),
POLYBENCH_ARRAY(A_tmp),
POLYBENCH_ARRAY(u1),
POLYBENCH_ARRAY(v1),
POLYBENCH_ARRAY(u2),
POLYBENCH_ARRAY(v2),
POLYBENCH_ARRAY(w_StrictFP),
POLYBENCH_ARRAY(x),
POLYBENCH_ARRAY(y),
POLYBENCH_ARRAY(z));
if (!check_FP(n, POLYBENCH_ARRAY(w), POLYBENCH_ARRAY(w_StrictFP)))
return 1;
/* Prevent dead-code elimination. All live-out data must be printed
by the function call in argument. */
polybench_prevent_dce(print_array(n, POLYBENCH_ARRAY(w_StrictFP)));
#endif
/* Be clean. */
POLYBENCH_FREE_ARRAY(A);
POLYBENCH_FREE_ARRAY(A_tmp);
POLYBENCH_FREE_ARRAY(u1);
POLYBENCH_FREE_ARRAY(v1);
POLYBENCH_FREE_ARRAY(u2);
POLYBENCH_FREE_ARRAY(v2);
POLYBENCH_FREE_ARRAY(w);
#if !FMA_DISABLED
POLYBENCH_FREE_ARRAY(w_StrictFP);
#endif
POLYBENCH_FREE_ARRAY(x);
POLYBENCH_FREE_ARRAY(y);
POLYBENCH_FREE_ARRAY(z);
return 0;
}