[PATCH] D62414: [InstCombine] canonicalize fcmp+select to minnum/maxnum intrinsics

[Sanjay Patel via Phabricator via llvm-commits]

spatel marked an inline comment as done.
spatel added inline comments.


================
Comment at: llvm/test/Transforms/InstCombine/fast-math.ll:884
 ; CHECK-LABEL: @min3(
-; CHECK-NEXT:    [[TMP1:%.*]] = fcmp nnan nsz olt double [[A:%.*]], [[B:%.*]]
-; CHECK-NEXT:    [[TMP2:%.*]] = select nnan nsz i1 [[TMP1]], double [[A]], double [[B]]
-; CHECK-NEXT:    ret double [[TMP2]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call nnan nsz double @llvm.minnum.f64(double [[A:%.*]], double [[B:%.*]])
+; CHECK-NEXT:    ret double [[TMP1]]
----------------
cameron.mcinally wrote:
> cameron.mcinally wrote:
> > Do we want NSZ here? I see that this was an artifact of the old fcmp+select lowering, but I'm not sure if it is correct. Definitely not IEEE-754 compliant...
> Now that I think about it, maybe it would be better to skip the fcmp+sel step and lower the libm call directly into the intrinsic?
Can you explain/show example for the compliancy problem?

Reference for our current docs:
http://llvm.org/docs/LangRef.html#llvm-minnum-intrinsic


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D62414/new/

https://reviews.llvm.org/D62414