llvm dev - Jan 2013 - [LLVMdev] [Patch][Review Requested][Compilation Time] Avoid frequent copy of elements in LoopStrengthReduce

Hello,
This patch aims to improve compile time performance by increasing the SCEV
vector size in LoopStrengthReduce. It is observed that the BaseRegs vector size
is 4 in most cases, and elements are frequently copied when it is initialized as
SmallVector<const SCEV *, 2> BaseRegs.
Our benchmark results show that the compilation time performance improved by
~0.5%.

Patch by Wan Xiaofei.

Thanks
Sriram
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130129/137e5b0b/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: loopStrengthReduce.patch
Type: application/octet-stream
Size: 3944 bytes
Desc: loopStrengthReduce.patch
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130129/137e5b0b/attachment.obj>

On 1/29/13 2:59 PM, Murali, Sriram wrote:
>
> Hello,
>
> This patch aims to improve compile time performance by increasing the 
> SCEV vector size in LoopStrengthReduce. It is observed that the 
> BaseRegs vector size is 4 in most cases, and elements are frequently 
> copied when it is initialized as SmallVector<const SCEV *, 2>
BaseRegs.
>
> Our benchmark results show that the compilation time performance 
> improved by ~0.5%.
>
Just an FYI that I think patches are supposed to be sent to llvm-commits.

-- John T.
> Patch by Wan Xiaofei.
>
> Thanks
>
> Sriram
>
>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130129/7c857b56/attachment.html>

On Tue, Jan 29, 2013 at 3:59 PM, Murali, Sriram <sriram.murali at
intel.com> wrote:
> Our benchmark results show that the compilation time performance improved
by
> ~0.5%.
That's fairly small; what was the standard deviation, confidence interval,
etc?

-- Sean Silva

The compilation time is measured for different benchmarks while compiling a .bc
file into a shared object. The improvement across the range of benchmarks is
listed in following table. If the reason behind the need for other performance
metrics is to identify possible measurement errors, then I think this table
would be of some help. However, we do not have the standard deviation and
confidence interval at the moment.
Benchmark\ Improvement (%)

Trunk

BaseRegs.Size

401.bzip2

100.00

101.14

403.gcc

100.00

100.33

429.mcf

100.00

100.37

433.milc

100.00

99.85

444.namd

100.00

100.28

445.gobmk

100.00

100.11

450.soplex

100.00

100.46

456.hmmer

100.00

100.29

458.sjeng

100.00

100.70

464.h264ref

100.00

102.10

470.lbm

100.00

100.07

471.omnetpp

100.00

100.36

bitmnp01

100.00

100.44

cjpegv2data6

100.00

100.76

idctrn01

100.00

99.83

libquake2

100.00

100.08

libquake_portable

100.00

100.02

libxcsoar

100.00

100.08

linpack

100.00

101.07

matrix01

100.00

100.28

nbench

100.00

100.76

tblook01

100.00

101.52

ttsprk01

100.00

100.82

Geomean

100.00

100.51





-----Original Message-----
From: Sean Silva [mailto:silvas at purdue.edu]
Sent: Tuesday, January 29, 2013 6:41 PM
To: Murali, Sriram
Cc: llvmdev at cs.uiuc.edu
Subject: Re: [LLVMdev] [Patch][Review Requested][Compilation Time] Avoid
frequent copy of elements in LoopStrengthReduce



On Tue, Jan 29, 2013 at 3:59 PM, Murali, Sriram <sriram.murali at
intel.com<mailto:sriram.murali at intel.com>> wrote:
> Our benchmark results show that the compilation time performance
> improved by ~0.5%.


That's fairly small; what was the standard deviation, confidence interval,
etc?



-- Sean Silva
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130130/6faf3223/attachment.html>

Sriram,

This patch looks good. Please commit.

...and thanks for the data.

-Andy

On Jan 29, 2013, at 12:59 PM, "Murali, Sriram" <sriram.murali at
intel.com> wrote:
> Hello,
> This patch aims to improve compile time performance by increasing the SCEV
vector size in LoopStrengthReduce. It is observed that the BaseRegs vector size
is 4 in most cases, and elements are frequently copied when it is initialized as
SmallVector<const SCEV *, 2> BaseRegs.
> Our benchmark results show that the compilation time performance improved
by ~0.5%.
>  
> Patch by Wan Xiaofei.
>  
> Thanks
> Sriram
>
<loopStrengthReduce.patch>_______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130201/74aa8c9c/attachment.html>