Fix a 4x slowdown in llc -asm-verbose caused by the use of
authorDan Gohman <gohman@apple.com>
Thu, 13 Aug 2009 15:27:57 +0000 (15:27 +0000)
committerDan Gohman <gohman@apple.com>
Thu, 13 Aug 2009 15:27:57 +0000 (15:27 +0000)
commitd6c0f65b532f022e584d8f1e3c044593b4c74a99
tree4161c57241b89e4d97111bdc0bd07f2b797a8291
parent8db5cce0217a2b76a954cb97de95c91e76edf59c
Fix a 4x slowdown in llc -asm-verbose caused by the use of
WriteAsOperand in more places.

Now that more things are using WriteAsOperand, its behavior of
constructing a TypePrinting object and populating it with strings for all
the numbered types in the Module on each call is a significant bottleneck.
Fancier solutions could be pursued here, but for now, just bypass the
TypePrinting overhead in obvious cases.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78906 91177308-0d34-0410-b5e6-96231b3b80d8
lib/VMCore/AsmWriter.cpp