[llvm-commits] [llvm] r41207 - in /llvm/trunk: include/llvm/Analysis/LoopInfo.h lib/Analysis/LoopInfo.cpp lib/Analysis/ScalarEvolution.cpp lib/Transforms/IPO/LoopExtractor.cpp lib/Transforms/Scalar/IndVarSimplify.cpp lib/Transforms/Scalar/LICM.cpp lib/Transforms/Scalar/LoopRotation.cpp lib/Transforms/Scalar/LoopUnswitch.cpp lib/Transforms/Utils/LCSSA.cpp lib/Transforms/Utils/LoopSimplify.cpp

[Devang Patel]

On Aug 20, 2007, at 8:43 PM, Chris Lattner wrote:

>> URL: http://llvm.org/viewvc/llvm-project?rev=41207&view=rev
>> Log:
>> Use SmallVector instead of std::vector.
>
> Thanks Devang,
>
> Two minor tweaks:
>
>>    /// outside of the loop.  These are the blocks _inside of the
>> current loop_
>>    /// which branch out.  The returned list is always unique.
>>    ///
>> -  void getExitingBlocks(std::vector<BasicBlock*> &Blocks) const;
>> +  void getExitingBlocks(SmallVector<BasicBlock *, 8> &Blocks) const;
>
> One of the nifty but non-obvious features of smallvector is that
> libraries who operate on a smallvector don't have to know the size
> that the client declared it as.  Instead of hardcoding 8 in here,
> just declare these as taking "SmallVectorImpl<BasicBlock*>&".  This
> will let clients decide "how small small should be". :)

nifty.

>
>> +void Loop::getExitingBlocks(SmallVector<BasicBlock*, 8>
>> &ExitingBlocks) const {
>>    std::vector<BasicBlock*> LoopBBs(block_begin(), block_end());
>
>> +void Loop::getExitBlocks(SmallVector<BasicBlock*, 8> &ExitBlocks)
>> const {
>>    std::vector<BasicBlock*> LoopBBs(block_begin(), block_end());
>
>> +void Loop::getUniqueExitBlocks(SmallVector<BasicBlock*, 8>
>> &ExitBlocks) const {
>>    std::vector<BasicBlock*> LoopBBs(block_begin(), block_end());
>
> Since you're in here, these are good candidates for (large)
> smallvectors also.  Try making them SmallVector<, 128> for example.

Yes. I intended to do it as a separate check-in.

-
Devang