tidy up a comment, as suggested by duncan
llvm-svn: 123590
This commit is contained in:
parent
cba4c33949
commit
7c9f4c9c2b
|
@ -1868,8 +1868,8 @@ bool SROA::isSafeAllocaToScalarRepl(AllocaInst *AI) {
|
|||
HasPadding(AI->getAllocatedType(), *TD))
|
||||
return false;
|
||||
|
||||
// If the alloca is never has an access to just *part* of it, but is accessed
|
||||
// with loads and stores, then we should use ConvertToScalarInfo to promote
|
||||
// If the alloca never has an access to just *part* of it, but is accessed
|
||||
// via loads and stores, then we should use ConvertToScalarInfo to promote
|
||||
// the alloca instead of promoting each piece at a time and inserting fission
|
||||
// and fusion code.
|
||||
if (!Info.hasSubelementAccess && Info.hasALoadOrStore) {
|
||||
|
|
Loading…
Reference in New Issue