Jeff Law wrote Wednesday, November 25, 2020 7:11 PM: > On 11/25/20 6:18 AM, Stefan Kanthak wrote: >> Jeff Law <l...@redhat.com> wrote:
[...] >>> My inclination is to leave the overflow checking double-word multiplier >>> as-is. >> See but <https://gcc.gnu.org/pipermail/gcc/2020-October/234048.html> ff. > Already read and considered it. >> >>> Though I guess you could keep the same structure as the existing >>> implementation which tries to avoid unnecessary multiplies and still use >>> the __builtin_{add,mul}_overflow to simplify the code a bit less >>> aggressively. Only relying on GCC to generate efficient code for __builtin_mul_overflow, this should be either DWtype __mulvDI3 (DWtype u, DWtype v) { DWtype w; if (__builtin_mul_overflow (u, v, &w)) abort (); return w; } or DWtype __mulvDI3 (DWtype u, DWtype v) { UDWtype w, s = 0 - (u < 0), t = 0 - (v < 0); if (__builtin_mul_overflow ((u ^ s) - s, (v ^ t) - t, &w)) abort (); s ^= t; w = (w ^ s) - s; if ((DWtype) (w ^ s) < 0) abort (); return w; } if the latter produces better machine code (which it does at least for i386 and AMD64). > Instead keep the tests that detect the special cases that don't need as > many multiplies and use the overflow builtins within that implementation > framework. In cases where we can use the operands directly, that's > helpful as going through the struct/union likely leads to unnecessary > register shuffling. regards Stefan
