[llvm-commits] [llvm] r122627 - /llvm/trunk/lib/CodeGen/StrongPHIElimination.cpp

[Jakob Stoklund Olesen]

On Dec 29, 2010, at 7:23 PM, Cameron Zwarich wrote:

> On Dec 29, 2010, at 6:39 PM, Jakob Stoklund Olesen <stoklund at 2pi.dk> wrote:
> 
>> On Dec 29, 2010, at 4:36 PM, Cameron Zwarich wrote:
>>>> How does strong phi elimination affect the runtime of coalescing?
>>> 
>>> It doesn't reduce it as much as I would like. On 403.gcc, coalescing takes up 4.0-4.1% of compile time with normal PHIElimination, and it goes down to 3.9% with StrongPHIElimination. StrongPHIElimination is getting fast enough that total compile time is just noise, but I'd like to investigate why there isn't as much of a speedup in coalescing as I would expect.
>> 
>> You may want to get some numbers from the ARM target as well. X86 has lots of copies from two-address instructions.
> 
> How useful would this sort of dominator-based coalescing be for copies introduced by two-address instructions?

Probably not very.

The advantage of strong phi elimination lies in avoiding creating a bunch of virtual registers and copy instructions only to delete them again. Those are not very expensive operations, so whatever you do must be fast.

You could handle 2-address instructions in a similar way, avoiding the copy instructions. I don't think it would be worth it.

/jakob

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1929 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20101229/e10fdb30/attachment.bin>