search for: d60262

...ics. In this proposal the intrinsic is prefixed with a 'v2' for the time being, with the expectation this will be dropped when we remove 'experimental' from the reduction intrinsics in the future. [Option B] Having separate ordered and unordered intrinsics (https://reviews.llvm.org/D60262). declare float @llvm.experimental.vector.reduce.ordered.fadd.f32.v4f32(float %start_value, <4 x float> %vec) declare float @llvm.experimental.vector.reduce.unordered.fadd.f32.v4f32(<4 x float> %vec) This will mean that the behaviour is explicit from the intrinsic and the use of &...

...t ordered and unordered reductions the same (in that they have the same arguments). Fast-math flags determine whether or not the intrinsic needs to be evaluated in strict order. Codegen for non-strict reductions should be able to fold away the identity-value. For Option B (https://reviews.llvm.org/D60262), David made the argument that making the reduction-order explicit (as opposed to deducing this from fast-math flags) would ensure the ordering is always as expected, even when FMF on the call sites are dropped for some reason.
 Is it correct that I sensed a slight preference for Option A? i.e. R...

...is prefixed with a 'v2' for the time being, with the expectation this will be dropped when we remove 'experimental' from the reduction intrinsics in the future. >>>> >>>> [Option B] Having separate ordered and unordered intrinsics (https://reviews.llvm.org/D60262). >>>> >>>> declare float @llvm.experimental.vector.reduce.ordered.fadd.f32.v4f32(float %start_value, <4 x float> %vec) >>>> declare float @llvm.experimental.vector.reduce.unordered.fadd.f32.v4f32(<4 x float> %vec) >>>> >>&gt...

...with a 'v2' for >> the time being, with the expectation this will be dropped when we >> remove 'experimental' from the reduction intrinsics in the future. >> >> [Option B] Having separate ordered and unordered intrinsics >> (https://reviews.llvm.org/D60262). >> >>   declare float >> @llvm.experimental.vector.reduce.ordered.fadd.f32.v4f32(float >> %start_value, <4 x float> %vec) >> >>   declare float >> @llvm.experimental.vector.reduce.unordered.fadd.f32.v4f32(<4 x float> >> %vec) >&gt...