michaelmaitland wrote: Adding a data point here: changing the small data limit to 0 is causing significant regressions on dhrystone, spec2006, spec2017, and geekbench5 on the sifive-x280 and sifive-p470.
Separate from this data point, a default of 8 bytes may be justified because most scalar primitives are 8 bytes or smaller on most architectures (boolean, char, int, unsigned, long, float, double, half). Data larger than this is likely something like arrays or structs. https://github.com/llvm/llvm-project/pull/83093 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
