[LLVMbugs] [Bug 5022] setz+mozxb -> xor+setz optimization

bugzilla-daemon at cs.uiuc.edu bugzilla-daemon at cs.uiuc.edu
Wed Sep 23 06:40:13 PDT 2009


http://llvm.org/bugs/show_bug.cgi?id=5022


Török Edwin <edwintorok at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




--- Comment #8 from Török Edwin <edwintorok at gmail.com>  2009-09-23 08:40:13 ---
(In reply to comment #6)
> In your numbers, the 95% confidence interval for the "rax -> rcx" case is
> entirely contained within the 95% confidence interval for the "original" case,
> so this isn't showing a significant difference.
> 

Indeed, I was timing foo1 always. I fixed that, and also used rdtsc directly,
results are:

Timing original:         60300165 average, 189461 stddev, 95% confidence
interval: 59921243 - 60679087;
Timing rax -> rcx:       60303412 average, 214691 stddev, 95% confidence
interval: 59874030 - 60732794;
Timing rax -> rcx, setz+movz -> xor+setz:        70366981 average, 287094
stddev, 95% confidence interval: 69792793 - 70941169;


So first two are the same, and the transform proposed here is actually slower.


Closing this bug.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.



More information about the llvm-bugs mailing list