[llvm-commits] [llvm] r134199 - /llvm/trunk/lib/CodeGen/RegisterCoalescer.cpp

[Rafael Ávila de Espíndola]

On 2011-07-01 03:25, Duncan Sands wrote:
>> Don't give up on coalescing A and B when we find
>>
>> A = X
>> B = X
>>
>> Instead, proceed as if we had found
>>
>> A = X
>> B = A
> 
> Testcase?

I have attached the closest thing I have to a
testcase that could be added. The problem is that
the result is fairly unstable. I guess I could
check for:

	leaq _ZZN2js9InterpretEP9...
	cmpq	$71, %rax
	sete	%r8b
	movb	%r8b, %al
	jmpq	*%rdx

which in the original had extra moves before the
jmpq:

	movq	%rcx, %r10
	movq	%rcx, %rsi

Do you think it is worth it?

> Ciao, Duncan.

Cheers,
Rafael
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: test5.ll
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20110702/68a7d61b/attachment.ksh>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: old.bz2
Type: application/x-bzip
Size: 2866 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20110702/68a7d61b/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: new.bz2
Type: application/x-bzip
Size: 2759 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20110702/68a7d61b/attachment-0001.bin>