unimore/hpc-2022-g3
unimore/hpc-2022-g3
1
Fork 0
mirror of https://github.com/Steffo99/unimore-hpc-assignments.git synced 2024-11-26 01:54:22 +00:00
Code Activity

kernel_atax: Remove useless tmp array, use a local variable instead

Browse source 
Local improvement: 30ms

Co-authored-by: Caterina Gazzotti <gazzotti.caterina@gmail.com>
This commit is contained in:
Steffo 2022-11-16 14:21:57 +01:00
parent aefa038d44
commit 7fc2506cc7
Signed by: steffo
GPG key ID: 6965406171929D01
1 changed files with 5 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
OpenMP/linear-algebra/kernels/atax/atax.c
View file

@ -60,8 +60,7 @@ static void print_array(int nx,
static void kernel_atax(int nx, int ny, static void kernel_atax(int nx, int ny,
DATA_TYPE POLYBENCH_2D(A, NX, NY, nx, ny), DATA_TYPE POLYBENCH_2D(A, NX, NY, nx, ny),
DATA_TYPE POLYBENCH_1D(x, NY, ny), DATA_TYPE POLYBENCH_1D(x, NY, ny),
DATA_TYPE POLYBENCH_1D(y, NY, ny), DATA_TYPE POLYBENCH_1D(y, NY, ny))
DATA_TYPE POLYBENCH_1D(tmp, NX, nx))
{ {
int i, j; int i, j;
@ -74,15 +73,15 @@ static void kernel_atax(int nx, int ny,
for (i = 0; i < _PB_NX; i++) for (i = 0; i < _PB_NX; i++)
{ {
/// Every iteration has its own tmp variable /// Every iteration has its own tmp variable
tmp[i] = 0; DATA_TYPE tmp = 0;
for (j = 0; j < _PB_NY; j++) for (j = 0; j < _PB_NY; j++)
/// Which gets increased by a bit on every iteration /// Which gets increased by a bit on every iteration
tmp[i] += A[i][j] * x[j]; tmp += A[i][j] * x[j];
for (j = 0; j < _PB_NY; j++) for (j = 0; j < _PB_NY; j++)
/// Which is later used for [something else] /// Which is later used for [something else]
y[j] = y[j] + A[i][j] * tmp[i]; y[j] = y[j] + A[i][j] * tmp;
} }
} }
@ -96,7 +95,6 @@ int main(int argc, char **argv)
POLYBENCH_2D_ARRAY_DECL(A, DATA_TYPE, NX, NY, nx, ny); POLYBENCH_2D_ARRAY_DECL(A, DATA_TYPE, NX, NY, nx, ny);
POLYBENCH_1D_ARRAY_DECL(x, DATA_TYPE, NY, ny); POLYBENCH_1D_ARRAY_DECL(x, DATA_TYPE, NY, ny);
POLYBENCH_1D_ARRAY_DECL(y, DATA_TYPE, NY, ny); POLYBENCH_1D_ARRAY_DECL(y, DATA_TYPE, NY, ny);
POLYBENCH_1D_ARRAY_DECL(tmp, DATA_TYPE, NX, nx);
/* Initialize array(s). */ /* Initialize array(s). */
init_array(nx, ny, POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(x)); init_array(nx, ny, POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(x));
@ -108,8 +106,7 @@ int main(int argc, char **argv)
kernel_atax(nx, ny, kernel_atax(nx, ny,
POLYBENCH_ARRAY(A), POLYBENCH_ARRAY(A),
POLYBENCH_ARRAY(x), POLYBENCH_ARRAY(x),
POLYBENCH_ARRAY(y), POLYBENCH_ARRAY(y));
POLYBENCH_ARRAY(tmp));
/* Stop and print timer. */ /* Stop and print timer. */
polybench_stop_instruments; polybench_stop_instruments;
@ -123,7 +120,6 @@ int main(int argc, char **argv)
POLYBENCH_FREE_ARRAY(A); POLYBENCH_FREE_ARRAY(A);
POLYBENCH_FREE_ARRAY(x); POLYBENCH_FREE_ARRAY(x);
POLYBENCH_FREE_ARRAY(y); POLYBENCH_FREE_ARRAY(y);
POLYBENCH_FREE_ARRAY(tmp);
return 0; return 0;
} }