[PATCH] D49125: [ARM] ParallelDSP: multiple reduction stmts in loop

[Sjoerd Meijer via Phabricator via llvm-commits]

SjoerdMeijer updated this revision to Diff 154809.
SjoerdMeijer added a comment.

Thanks for the suggestion.

This update makes explicit that the different MAC chains don't interfere/alias with each other. I.e. MAC-chains are a simple chains that starts with a reducing integer add statement, and then a chain of adds and muls, which only has sext and loads as operands. Thus, we don't write memory, and we set a ReadOnly flag.


https://reviews.llvm.org/D49125

Files:
  lib/Target/ARM/ARMParallelDSP.cpp
  test/CodeGen/ARM/smlad0.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D49125.154809.patch
Type: text/x-patch
Size: 13468 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180710/80d35ea4/attachment.bin>