JJSierraM/squiggle.c
1
0
Fork 0
forked from personal/squiggle.c
Code Issues Pull Requests Packages Projects Releases Wiki Activity

remove loop unrolling again

Browse Source
This commit is contained in:
NunoSempere 2024-01-13 00:50:51 +01:00
parent 3bb2804ccf
commit b497b5b399
1 changed files with 0 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
squiggle_more.c
View File

@ -75,23 +75,6 @@ void sampler_parallel(double (*sampler)(uint64_t* seed), double* results, int n_
// b) trying to unroll loops actually makes the code slower // b) trying to unroll loops actually makes the code slower
// c) 8 results[j] are 8 doubles, which fit a cache line. If n_samples/n_threads // c) 8 results[j] are 8 doubles, which fit a cache line. If n_samples/n_threads
} }
// Failed loop unrolling
/*
for (int j = lower_bound_inclusive; j < 4*(upper_bound_not_inclusive/4); j++) {
results[j+0] = sampler(&(cache_box[i].seed));
results[j+1] = sampler(&(cache_box[i].seed));
results[j+2] = sampler(&(cache_box[i].seed));
results[j+3] = sampler(&(cache_box[i].seed));
// In principle, these results[j] could also result in two threads competing for the same cache line.
// In practice, though,
// a) this would happen infrequently
// b) trying to unroll loops actually makes the code slower
// c) 8 results[j] are 8 doubles, which fit a cache line. If n_samples/n_threads
}
for (int j = 4*(upper_bound_not_inclusive/4); j < upper_bound_not_inclusive; j++) {
results[j] = sampler(&(cache_box[i].seed));
}
*/
} }
} }
for (int j = divisor_multiple; j < n_samples; j++) { for (int j = divisor_multiple; j < n_samples; j++) {