[PATCH] D59070: [builtins] Divide shouldn't underflow if rounded result would be normal.

[Eli Friedman via Phabricator via llvm-commits]

efriedma created this revision.
efriedma added reviewers: compnerd, scanon.
Herald added subscribers: Sanitizers, jdoerfert.
Herald added projects: LLVM, Sanitizers.

We were treating certain edge cases that are actually normal as denormal results, and flushing them to zero; we shouldn't do that. Not sure this is the cleanest way to implement this edge case, but I wanted to avoid adding any code on the common path.

(This doesn't touch the behavior for results that are actually denormal; they're still flushed to zero.)


Repository:
  rCRT Compiler Runtime

https://reviews.llvm.org/D59070

Files:
  lib/builtins/divdf3.c
  lib/builtins/divsf3.c
  lib/builtins/divtf3.c
  test/builtins/Unit/divdf3_test.c
  test/builtins/Unit/divsf3_test.c
  test/builtins/Unit/divtf3_test.c

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D59070.189640.patch
Type: text/x-patch
Size: 6557 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190307/fd6da1bc/attachment-0001.bin>