assigned to @rotateright

%23 = call <32 x i8> @​llvm.abs.v32i8(<32 x i8> %20, i1 false)

What are the implications of clearing the poison flag as part of such a fold?

The lowering may be less efficient in some targets. But it won't be worse than the original code (worst case the target can just emit the old code).
So I think there's no problem (in theory). Any issue that may come up should be fixable.

It's okay to drop the poison flag on abs during transforms, it's not particularly important. It's not like ctlz etc where it influences lowering. It only exists so we can infer that abs(x) >= 0.

Regressed with introduction of new abs intrinsic?

Regressed with introduction of new abs intrinsic?

Yes.

int8_t transform_abs_epi8(int8_t x) {
return abs(x);
}

Trunk:

define signext i8 @​transform_abs_epi8(i8 signext %0) {
%2 = sext i8 %0 to i32
%3 = tail call i32 @​llvm.abs.i32(i32 %2, i1 true)
%4 = trunc i32 %3 to i8
ret i8 %4
}
declare i32 @​llvm.abs.i32(i32, i1 immarg)

Clang 11:

define signext i8 @​transform_abs_epi8(i8 signext %0) {
%2 = icmp slt i8 %0, 0
%3 = sub i8 0, %0
%4 = select i1 %2, i8 %3, i8 %0
ret i8 %4
}

I can work on an instcombine for this. I'll see if min/max need a similar fold.

Minimally, this combine starts from abs (don't need the trunc):
https://alive2.llvm.org/ce/z/ECaz-p

define i32 @​src(i8 %0) {
%a = sext i8 %0 to i32
%m = call i32 @​llvm.abs.i32(i32 %a, i1 1)
ret i32 %m
}

define i32 @​tgt(i8 %0) {
%m = call i8 @​llvm.abs.i8(i8 %0, i1 0)
%r = zext i8 %m to i32
ret i32 %r
}

https://reviews.llvm.org/D95235

https://reviews.llvm.org/rG411c144e4c99

min/max intrinsics should have a similar fold, but we'll need to account for binop variations like a constant operand vs. 2 {s/z}exts.