personal/time-to-botec
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Revert "tweak: try using array instead of array of pointers"

Browse Source 
it makes code 3x slower
This commit is contained in:
NunoSempere 2023-06-03 10:56:09 -06:00
parent 00e6b67bf6
commit 1ff35f6235
2 changed files with 6 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

BIN
C/out/samples
View File

Binary file not shown.

19
C/samples.c
View File

@ -165,19 +165,12 @@ void mixture(float (*samplers[])(uint32_t*), float* weights, int n_dists, float*
float p1; float p1;
int sample_index, i, split_array_length; int sample_index, i, split_array_length;
uint32_t* seeds = malloc(n_threads * sizeof(uint32_t));
for (uint32_t i = 0; i < n_threads; i++) {
seeds[i] = i + 1;
}
// uint32_t* seeds[n_threads]; // uint32_t* seeds[n_threads];
/* uint32_t** seeds = malloc(n_threads * sizeof(uint32_t*));
uint32_t** seeds = malloc(n_threads * sizeof(uint32_t*));
for (uint32_t i = 0; i < n_threads; i++) { for (uint32_t i = 0; i < n_threads; i++) {
seeds[i] = malloc(sizeof(uint32_t)); seeds[i] = malloc(sizeof(uint32_t));
*seeds[i] = i + 1; // xorshift can't start with 0 *seeds[i] = i + 1; // xorshift can't start with 0
} }
*/
#pragma omp parallel private(i, p1, sample_index, split_array_length) #pragma omp parallel private(i, p1, sample_index, split_array_length)
{ {
@ -185,10 +178,10 @@ void mixture(float (*samplers[])(uint32_t*), float* weights, int n_dists, float*
for (i = 0; i < n_threads; i++) { for (i = 0; i < n_threads; i++) {
split_array_length = split_array_get_length(i, N, n_threads); split_array_length = split_array_get_length(i, N, n_threads);
for (int j = 0; j < split_array_length; j++) { for (int j = 0; j < split_array_length; j++) {
p1 = random_uniform(0, 1, &seeds[i]); p1 = random_uniform(0, 1, seeds[i]);
for (int k = 0; k < n_dists; k++) { for (int k = 0; k < n_dists; k++) {
if (p1 < cumsummed_normalized_weights[k]) { if (p1 < cumsummed_normalized_weights[k]) {
results[i][j] = samplers[k](&seeds[i]); results[i][j] = samplers[k](seeds[i]);
break; break;
} }
} }
@ -198,9 +191,9 @@ void mixture(float (*samplers[])(uint32_t*), float* weights, int n_dists, float*
// free(normalized_weights); // free(normalized_weights);
// free(cummulative_weights); // free(cummulative_weights);
free(cumsummed_normalized_weights); free(cumsummed_normalized_weights);
// for (uint32_t i = 0; i < n_threads; i++) { for (uint32_t i = 0; i < n_threads; i++) {
// free(seeds[i]); free(seeds[i]);
// } }
free(seeds); free(seeds);
} }