Displaying 6 results from an estimated 6 matches for "r187728".
2013 Aug 09
2
[LLVMdev] [Polly] Summary of some expensive compiler passes, especially PollyDependence
...ollyScopInfo run with a PollyDependences run (same source code?) and
>the Dependences run shows compile time speedup over ScopInfo? This is
>surprising.
In fact, there are three runs:
pollyOpt (run id = 18): the baseline of Polly
pollyScopInfo (run id = 25): Polly with the ScopInfo patch r187728 that split start value from SCEVAddRecExpr
pollyDependence (run id = 26): Polly with both ScopInfo patch r187728 and PollyDependence patch r187981 that simplify the Read/Write/MayWrite before feeding them into ISL.
As a result, I compare pollyDependence (run id = 26) with pollyScopInfo (run id = 2...
2013 Aug 09
0
[LLVMdev] [Polly] Summary of some expensive compiler passes, especially PollyDependence
...ependences run (same source code?) and
>> the Dependences run shows compile time speedup over ScopInfo? This is
>> surprising.
>
> In fact, there are three runs:
>
> pollyOpt (run id = 18): the baseline of Polly
> pollyScopInfo (run id = 25): Polly with the ScopInfo patch r187728 that split start value from SCEVAddRecExpr
> pollyDependence (run id = 26): Polly with both ScopInfo patch r187728 and PollyDependence patch r187981 that simplify the Read/Write/MayWrite before feeding them into ISL.
>
> As a result, I compare pollyDependence (run id = 26) with pollyScopIn...
2013 Aug 09
0
[LLVMdev] [Polly] Summary of some expensive compiler passes, especially PollyDependence
On 08/08/2013 06:27 PM, Star Tan wrote:
> At 2013-08-08 22:28:33,"Tobias Grosser" <tobias at grosser.es> wrote:
>> On 08/08/2013 01:29 AM, Star Tan wrote:
>>> Hi all,
>>>
>>>
>>> I have summarized the top 10 compiler passes for Polly when compiling LLVM test-ssuite. Results can be viewed on:
>>>
2013 Aug 09
2
[LLVMdev] [Polly] Summary of some expensive compiler passes, especially PollyDependence
... multiplies: 54.3% (Poly - Calculate dependence)
>> Puzzle: 22.8% (Poly - Calculate dependence)
>>
>>
>> As a result, it is critical to improve the PollyDependence pass. I have previously committed a patch file to split start value from SCEVAddRecExpr (r187728), which can improve the PollyDependence pass for some benchmarks as follows:
>> http://188.40.87.11:8000/db_default/v4/nts/25?baseline=18&compare_to=18
>
>Those are very nice results. Removing 80% of the compile time for the lu
>benchmark is a very nice outcome.
>
>&...
2013 Aug 08
2
[LLVMdev] [Polly] Summary of some expensive compiler passes, especially PollyDependence
....1% (Polly - Calculate dependence)
multiplies: 54.3% (Poly - Calculate dependence)
Puzzle: 22.8% (Poly - Calculate dependence)
As a result, it is critical to improve the PollyDependence pass. I have previously committed a patch file to split start value from SCEVAddRecExpr (r187728), which can improve the PollyDependence pass for some benchmarks as follows:
http://188.40.87.11:8000/db_default/v4/nts/25?baseline=18&compare_to=18
Obviously, it is not enough.
I have investigated the salsa20 benchmark, in which the PollyDependence pass consumes 98.5% of total compile-...
2013 Aug 08
0
[LLVMdev] [Polly] Summary of some expensive compiler passes, especially PollyDependence
...pendence)
> multiplies: 54.3% (Poly - Calculate dependence)
> Puzzle: 22.8% (Poly - Calculate dependence)
>
>
> As a result, it is critical to improve the PollyDependence pass. I have previously committed a patch file to split start value from SCEVAddRecExpr (r187728), which can improve the PollyDependence pass for some benchmarks as follows:
> http://188.40.87.11:8000/db_default/v4/nts/25?baseline=18&compare_to=18
Those are very nice results. Removing 80% of the compile time for the lu
benchmark is a very nice outcome.
> Obviously, it is not...