[PATCH] Add glretrace option '--snapshot-format=MD5' that allow write image MD5 to a text format file.

[Meng, Mengmeng]

Hi,
> Do the same .png files mismatch, corresponding to the md5 hashes?
I don't checked png files now. But I check .264 files and the .264 files are the same(by md5sum or diff) between two tests. 
So I may do more investigation on the MD5 patch.

----------------------------------
glretrace --snapshot-format=RGB -s - tmp.trace  | gst-launch-0.10 fdsrc blocksize=409600 ! queue ! videoparse format=rgb width=800 height=600 ! queue ! ffmpegcolorspace ! queue ! vaapiupload  ! queue ! vaapiencodeh264 ! filesink location=3.264


Thanks,
Meng


-----Original Message-----
From: Nigel Stewart [mailto:nigels.com at gmail.com] 
Sent: Wednesday, October 30, 2013 12:06 PM
To: Meng, Mengmeng
Cc: Alexander Monakov; apitrace at lists.freedesktop.org
Subject: Re: [PATCH] Add glretrace option '--snapshot-format=MD5' that allow write image MD5 to a text format file.

Ah, good.  A 3x speedup for testing throughput is all goodness...
Do the same .png files mismatch, corresponding to the md5 hashes?

- Nigel

On Tue, Oct 29, 2013 at 10:31 PM, Meng, Mengmeng <mengmeng.meng at intel.com> wrote:
> Oh, sorry. I make an error, MD5 should be faster than PNG . It should be:
> MD5: Rendered 3435 frames in 143.251 secs, average of 23.9789 fps   md5.file 4K
> PNG: Rendered 3435 frames in 478.734 secs, average of 7.17518 fps   png.file 1.7G
>
> Thanks,
> Meng
>
>
> -----Original Message-----
> From: apitrace-bounces at lists.freedesktop.org 
> [mailto:apitrace-bounces at lists.freedesktop.org] On Behalf Of Nigel 
> Stewart
> Sent: Wednesday, October 30, 2013 11:23 AM
> To: Meng, Mengmeng
> Cc: Alexander Monakov; apitrace at lists.freedesktop.org
> Subject: Re: [PATCH] Add glretrace option '--snapshot-format=MD5' that allow write image MD5 to a text format file.
>
> Is that right?  md5sbeing 3x slower than (uncompressed?) .png files?
> That your md5s don't match is a concern - Intel GPU and driver?
> Does it help at all to throw away the least significant RGBA bits?
>
> - Nigel
>
> On Tue, Oct 29, 2013 at 8:34 PM, Meng, Mengmeng <mengmeng.meng at intel.com> wrote:
>> Hi, I do a test comparing MD5 with PNG, Open Arena at 800x600 on IVB
>> MD5:  Rendered 3435 frames in 478.734 secs, average of 7.17518 fps  md5.file 4K
>> PNG:  Rendered 3435 frames in 143.251 secs, average of 23.9789 fps   pnp.file 1.7G
>>
>> But there is a verification problem. It means that with the same 
>> driver, MD5s are not all same between two tests. E.g. there are ~263 different frames in 3435 frames So we may consider "a threshold value"?
>>
>>
>>
>> Thanks,
>> Meng
>>
>>
>> -----Original Message-----
>> From: apitrace-bounces at lists.freedesktop.org
>> [mailto:apitrace-bounces at lists.freedesktop.org] On Behalf Of Nigel 
>> Stewart
>> Sent: Tuesday, October 29, 2013 11:48 PM
>> To: Alexander Monakov
>> Cc: Meng, Mengmeng; apitrace at lists.freedesktop.org
>> Subject: Re: [PATCH] Add glretrace option '--snapshot-format=MD5' that allow write image MD5 to a text format file.
>>
>> Alex,
>>
>> Sure, we can all happily agree about having clean and tidy patches for consideration.
>>
>> - Nigel
>>
>> On Tue, Oct 29, 2013 at 10:25 AM, Alexander Monakov <amonakov at ispras.ru> wrote:
>>> Let's hear what more experienced (than myself) maintainers say.
>>>
>>> Regarding the patch itself, it should avoid adding irrelevant files 
>>> in thirdparty (such as VS project files), and references to 
>>> libcrypto should be removed from CMakeLists.txt.
>>>
>>> Alexander
>> _______________________________________________
>> apitrace mailing list
>> apitrace at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/apitrace
> _______________________________________________
> apitrace mailing list
> apitrace at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/apitrace