don't vectorize tiny loop

could theoretically maybe be beneficial for functions with a hilarious
number of arguments, but whatever. Improves perf on few-argument
functions from not unrolling/vectorizing.
This commit is contained in:
dzaima 2024-07-04 21:52:56 +03:00
parent 3e24442011
commit 352f22f8f6

View File

@ -940,7 +940,7 @@ B libffiFn_c2(B t, B w, B x) {
genObj(e.o, o, tmpAlloc + e.staticOffset, &ffiObjs);
}
for (usz i = 0; i < argn; i++) argPtrs[i] = tmpAlloc + ents[i+1].staticOffset;
PLAINLOOP for (usz i = 0; i < argn; i++) argPtrs[i] = tmpAlloc + ents[i+1].staticOffset;
void* res = tmpAlloc + ents[0].staticOffset;
// for (usz i = 0; i < argObj->staticAllocTotal; i++) { // simple hexdump of the allocation