[PATCH] D87115: [FPEnv][X86][SystemZ] Use different algorithms for i64->double uint_to_fp under strictfp to avoid producing -0.0 when rounding to negative infinity

[Eli Friedman via Phabricator via llvm-commits]

efriedma added inline comments.


================
Comment at: llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp:2440
+      (SrcVT == MVT::i64 && DestVT == MVT::f64)) {
+    LLVM_DEBUG(dbgs() << "Converting unsigned i32/i64 to f32/f64\n");
     // For unsigned conversions, convert them to signed conversions using the
----------------
What types is this correct for?  I guess the integer type has to be larger than the mantissa by some number of bits, so the divide by 2 doesn't affect the rounding?


================
Comment at: llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp:6469
+  if (Node->isStrictFPOpcode())
+    return false;
+
----------------
Should we drop the isStrictFPOpcode check later in the function?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D87115/new/

https://reviews.llvm.org/D87115