Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-07-05 Thread James Almer 
On 7/5/2017 7:52 AM, Hendrik Leppkes wrote:
> On Wed, Jul 5, 2017 at 11:40 AM, Paul B Mahol  wrote:
>>
>> I get segfauls in sbrdsp code after some recent commit.
> 
> Indeed, fate also shows almost all AAC tests failing on x86_64 systems.
> 
> - Hendrik

Yeah, got a bit overzealous in an unrelated commit with some
micro-optimization that didn't break checkasm.
Should be reverted now, sorry for dropping the yellow paint bucket on fate.
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-07-05 Thread Hendrik Leppkes 
On Wed, Jul 5, 2017 at 11:40 AM, Paul B Mahol  wrote:
> On 7/5/17, James Almer  wrote:
>> On 7/4/2017 10:45 PM, Michael Niedermayer wrote:
>>> On Tue, Jul 04, 2017 at 03:15:56PM -0300, James Almer wrote:
 On 7/4/2017 2:31 PM, Michael Niedermayer wrote:
> On Mon, Jul 03, 2017 at 02:32:28PM +0200, Matthieu Bouron wrote:
>> On Fri, Jun 30, 2017 at 05:16:37PM +0200, Matthieu Bouron wrote:
>>> On Fri, Jun 30, 2017 at 03:55:52PM +0200, Michael Niedermayer wrote:
 On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
> On 6/29/2017 10:14 PM, Henrik Gramner wrote:
>> On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
>>  wrote:
>>> Program received signal SIGSEGV, Segmentation fault.
>>> 0x00684919 in ff_sbr_hf_gen_sse ()
>>
>>>0x00684909 :   sub%r9,%r8
>>
>>> => 0x00684919 :   movaps
>>> (%rsi,%r8,1),%xmm0
>>
>>> r9 0xdeadbeef0080   -2401053092612145024
>>
>> Another case of a 32-bit int being used as part of a 64-bit
>> operation.
>
> I can't reproduce it on my ArchLinux x86_64 environment for some
> reason,
> but based on what you said i assume the attached patch should fix
> it.

 no crash occurs here with this, so it seems fixed
>>>
>>> Should i push the patchset or wait a little bit longer ?
>>
>> Patchset applied.
>
> it seems theres some issue still in this:
>
> checkasm: using random seed 3655967467
> MMX:
>  - audiodsp.audiodsp [OK]
>  - blockdsp.blockdsp [OK]
>  - h264dsp.idct  [OK]
>  - h264pred.pred4x4  [OK]
>  - h264pred.pred8x8  [OK]
>  - h264pred.pred16x16[OK]
>  - pixblockdsp.get_pixels[OK]
>  - pixblockdsp.diff_pixels   [OK]
>  - vp8dsp.idct   [OK]
>  - vp8dsp.mc [OK]
>  - vp9dsp.ipred  [OK]
>  - vp9dsp.itxfm  [OK]
>  - vp9dsp.mc [OK]
> MMXEXT:
>  - audiodsp.audiodsp [OK]
>  - h264dsp.idct  [OK]
>  - h264pred.pred4x4  [OK]
>  - h264pred.pred8x8  [OK]
>  - h264pred.pred16x16[OK]
>  - h264pred.pred8x8l [OK]
>  - h264qpel.put  [OK]
>  - h264qpel.avg  [OK]
>  - hevc_add_res.add_residual [OK]
>  - hevc_idct.idct_dc [OK]
>  - vp8dsp.mc [OK]
>  - vp9dsp.ipred  [OK]
>  - vp9dsp.itxfm  [OK]
>  - vp9dsp.loopfilter [OK]
>  - vp9dsp.mc [OK]
> SSE:
>  - aacpsdsp.add_squares  [OK]
>  - aacpsdsp.mul_pair_single  [OK]
>  - aacpsdsp.hybrid_analysis  [OK]
>  - sbrdsp.sum64x5[OK]
>  - sbrdsp.sum_square [OK]
>  - sbrdsp.neg_odd_64 [OK]
>  - sbrdsp.qmf_post_shuffle   [OK]
>  - sbrdsp.qmf_deint_neg  [OK]
>  - sbrdsp.qmf_deint_bfly [OK]
>  - sbrdsp.autocorrelate  [OK]
>  - sbrdsp.hf_gen [OK]
>  - sbrdsp.hf_g_filt  [OK]
>  - audiodsp.audiodsp [OK]
>  - blockdsp.blockdsp [OK]
>  - fmtconvert.fmtconvert [OK]
>  - h264pred.pred16x16[OK]
>  - vp8dsp.idct   [OK]
>  - vp8dsp.mc [OK]
>  - vp9dsp.ipred  [OK]
>  - vp9dsp.mc [OK]
>  - float_dsp.vector_fmul [OK]
>  - float_dsp.vector_fmac [OK]
>  - float_dsp.butterflies_float   [OK]
>  - float_dsp.scalarproduct_float [OK]
> SSE2:
>  - sbrdsp.qmf_pre_shuffle[OK]
>  - sbrdsp.qmf_deint_bfly [OK]
>
> Program received signal SIGSEGV, Segmentation fault.
> apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
> 418 movu   m7, [Yq + 2*count + mmsize]
> (gdb) bt
> Python Exception  No module named
> gdb.frames:
> #0  apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
> #1  0x0043659b in checkasm_checked_call () at
> tests/checkasm/x86/checkasm.asm:77
> #2  0xdeadbeefdeadbeef in ?? ()
> #3  0xdeadbeefdeadbeef in ?? ()
> #4  0xdeadbeefdeadbeef in ?? ()
> #5  0xdeadbeefdeadbeef in ?? ()
> #6  0xdeadbeefdeadbeef in ?? ()
> #7  0xdeadbeefdeadbeef in ?? ()
> #8  0xdeadbeefdeadbeef in ?? ()
> #9  0xdeadbeefdeadbeef in ?? ()
> #10 0xdeadbeefdeadbeef in ?? ()
> #11 0xdeadbeefdeadbeef in ?? ()
> #12 0xdeadbeefdeadbeef in ?? ()
>

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-07-05 Thread Paul B Mahol 
On 7/5/17, James Almer  wrote:
> On 7/4/2017 10:45 PM, Michael Niedermayer wrote:
>> On Tue, Jul 04, 2017 at 03:15:56PM -0300, James Almer wrote:
>>> On 7/4/2017 2:31 PM, Michael Niedermayer wrote:
 On Mon, Jul 03, 2017 at 02:32:28PM +0200, Matthieu Bouron wrote:
> On Fri, Jun 30, 2017 at 05:16:37PM +0200, Matthieu Bouron wrote:
>> On Fri, Jun 30, 2017 at 03:55:52PM +0200, Michael Niedermayer wrote:
>>> On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
 On 6/29/2017 10:14 PM, Henrik Gramner wrote:
> On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
>  wrote:
>> Program received signal SIGSEGV, Segmentation fault.
>> 0x00684919 in ff_sbr_hf_gen_sse ()
>
>>0x00684909 :   sub%r9,%r8
>
>> => 0x00684919 :   movaps
>> (%rsi,%r8,1),%xmm0
>
>> r9 0xdeadbeef0080   -2401053092612145024
>
> Another case of a 32-bit int being used as part of a 64-bit
> operation.

 I can't reproduce it on my ArchLinux x86_64 environment for some
 reason,
 but based on what you said i assume the attached patch should fix
 it.
>>>
>>> no crash occurs here with this, so it seems fixed
>>
>> Should i push the patchset or wait a little bit longer ?
>
> Patchset applied.

 it seems theres some issue still in this:

 checkasm: using random seed 3655967467
 MMX:
  - audiodsp.audiodsp [OK]
  - blockdsp.blockdsp [OK]
  - h264dsp.idct  [OK]
  - h264pred.pred4x4  [OK]
  - h264pred.pred8x8  [OK]
  - h264pred.pred16x16[OK]
  - pixblockdsp.get_pixels[OK]
  - pixblockdsp.diff_pixels   [OK]
  - vp8dsp.idct   [OK]
  - vp8dsp.mc [OK]
  - vp9dsp.ipred  [OK]
  - vp9dsp.itxfm  [OK]
  - vp9dsp.mc [OK]
 MMXEXT:
  - audiodsp.audiodsp [OK]
  - h264dsp.idct  [OK]
  - h264pred.pred4x4  [OK]
  - h264pred.pred8x8  [OK]
  - h264pred.pred16x16[OK]
  - h264pred.pred8x8l [OK]
  - h264qpel.put  [OK]
  - h264qpel.avg  [OK]
  - hevc_add_res.add_residual [OK]
  - hevc_idct.idct_dc [OK]
  - vp8dsp.mc [OK]
  - vp9dsp.ipred  [OK]
  - vp9dsp.itxfm  [OK]
  - vp9dsp.loopfilter [OK]
  - vp9dsp.mc [OK]
 SSE:
  - aacpsdsp.add_squares  [OK]
  - aacpsdsp.mul_pair_single  [OK]
  - aacpsdsp.hybrid_analysis  [OK]
  - sbrdsp.sum64x5[OK]
  - sbrdsp.sum_square [OK]
  - sbrdsp.neg_odd_64 [OK]
  - sbrdsp.qmf_post_shuffle   [OK]
  - sbrdsp.qmf_deint_neg  [OK]
  - sbrdsp.qmf_deint_bfly [OK]
  - sbrdsp.autocorrelate  [OK]
  - sbrdsp.hf_gen [OK]
  - sbrdsp.hf_g_filt  [OK]
  - audiodsp.audiodsp [OK]
  - blockdsp.blockdsp [OK]
  - fmtconvert.fmtconvert [OK]
  - h264pred.pred16x16[OK]
  - vp8dsp.idct   [OK]
  - vp8dsp.mc [OK]
  - vp9dsp.ipred  [OK]
  - vp9dsp.mc [OK]
  - float_dsp.vector_fmul [OK]
  - float_dsp.vector_fmac [OK]
  - float_dsp.butterflies_float   [OK]
  - float_dsp.scalarproduct_float [OK]
 SSE2:
  - sbrdsp.qmf_pre_shuffle[OK]
  - sbrdsp.qmf_deint_bfly [OK]

 Program received signal SIGSEGV, Segmentation fault.
 apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
 418 movu   m7, [Yq + 2*count + mmsize]
 (gdb) bt
 Python Exception  No module named
 gdb.frames:
 #0  apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
 #1  0x0043659b in checkasm_checked_call () at
 tests/checkasm/x86/checkasm.asm:77
 #2  0xdeadbeefdeadbeef in ?? ()
 #3  0xdeadbeefdeadbeef in ?? ()
 #4  0xdeadbeefdeadbeef in ?? ()
 #5  0xdeadbeefdeadbeef in ?? ()
 #6  0xdeadbeefdeadbeef in ?? ()
 #7  0xdeadbeefdeadbeef in ?? ()
 #8  0xdeadbeefdeadbeef in ?? ()
 #9  0xdeadbeefdeadbeef in ?? ()
 #10 0xdeadbeefdeadbeef in ?? ()
 #11 0xdeadbeefdeadbeef in ?? ()
 #12 0xdeadbeefdeadbeef in ?? ()
 #13 0xdeadbeefdeadbeef in ?? ()
 #14 0xdeadbeefdeadbeef in ?? ()
 #15 0xdeadbeefdeadbeef in ?? ()
 #16 0xdeadbeefdeadbeef in ?? ()
 #17 0xdeadbeefdeadbeef in ?? ()
 #18

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-07-04 Thread James Almer 
On 7/4/2017 10:45 PM, Michael Niedermayer wrote:
> On Tue, Jul 04, 2017 at 03:15:56PM -0300, James Almer wrote:
>> On 7/4/2017 2:31 PM, Michael Niedermayer wrote:
>>> On Mon, Jul 03, 2017 at 02:32:28PM +0200, Matthieu Bouron wrote:
 On Fri, Jun 30, 2017 at 05:16:37PM +0200, Matthieu Bouron wrote:
> On Fri, Jun 30, 2017 at 03:55:52PM +0200, Michael Niedermayer wrote:
>> On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
>>> On 6/29/2017 10:14 PM, Henrik Gramner wrote:
 On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
  wrote:
> Program received signal SIGSEGV, Segmentation fault.
> 0x00684919 in ff_sbr_hf_gen_sse ()

>0x00684909 :   sub%r9,%r8

> => 0x00684919 :   movaps 
> (%rsi,%r8,1),%xmm0

> r9 0xdeadbeef0080   -2401053092612145024

 Another case of a 32-bit int being used as part of a 64-bit operation.
>>>
>>> I can't reproduce it on my ArchLinux x86_64 environment for some reason,
>>> but based on what you said i assume the attached patch should fix it.
>>
>> no crash occurs here with this, so it seems fixed
>
> Should i push the patchset or wait a little bit longer ?

 Patchset applied.
>>>
>>> it seems theres some issue still in this:
>>>
>>> checkasm: using random seed 3655967467
>>> MMX:
>>>  - audiodsp.audiodsp [OK]
>>>  - blockdsp.blockdsp [OK]
>>>  - h264dsp.idct  [OK]
>>>  - h264pred.pred4x4  [OK]
>>>  - h264pred.pred8x8  [OK]
>>>  - h264pred.pred16x16[OK]
>>>  - pixblockdsp.get_pixels[OK]
>>>  - pixblockdsp.diff_pixels   [OK]
>>>  - vp8dsp.idct   [OK]
>>>  - vp8dsp.mc [OK]
>>>  - vp9dsp.ipred  [OK]
>>>  - vp9dsp.itxfm  [OK]
>>>  - vp9dsp.mc [OK]
>>> MMXEXT:
>>>  - audiodsp.audiodsp [OK]
>>>  - h264dsp.idct  [OK]
>>>  - h264pred.pred4x4  [OK]
>>>  - h264pred.pred8x8  [OK]
>>>  - h264pred.pred16x16[OK]
>>>  - h264pred.pred8x8l [OK]
>>>  - h264qpel.put  [OK]
>>>  - h264qpel.avg  [OK]
>>>  - hevc_add_res.add_residual [OK]
>>>  - hevc_idct.idct_dc [OK]
>>>  - vp8dsp.mc [OK]
>>>  - vp9dsp.ipred  [OK]
>>>  - vp9dsp.itxfm  [OK]
>>>  - vp9dsp.loopfilter [OK]
>>>  - vp9dsp.mc [OK]
>>> SSE:
>>>  - aacpsdsp.add_squares  [OK]
>>>  - aacpsdsp.mul_pair_single  [OK]
>>>  - aacpsdsp.hybrid_analysis  [OK]
>>>  - sbrdsp.sum64x5[OK]
>>>  - sbrdsp.sum_square [OK]
>>>  - sbrdsp.neg_odd_64 [OK]
>>>  - sbrdsp.qmf_post_shuffle   [OK]
>>>  - sbrdsp.qmf_deint_neg  [OK]
>>>  - sbrdsp.qmf_deint_bfly [OK]
>>>  - sbrdsp.autocorrelate  [OK]
>>>  - sbrdsp.hf_gen [OK]
>>>  - sbrdsp.hf_g_filt  [OK]
>>>  - audiodsp.audiodsp [OK]
>>>  - blockdsp.blockdsp [OK]
>>>  - fmtconvert.fmtconvert [OK]
>>>  - h264pred.pred16x16[OK]
>>>  - vp8dsp.idct   [OK]
>>>  - vp8dsp.mc [OK]
>>>  - vp9dsp.ipred  [OK]
>>>  - vp9dsp.mc [OK]
>>>  - float_dsp.vector_fmul [OK]
>>>  - float_dsp.vector_fmac [OK]
>>>  - float_dsp.butterflies_float   [OK]
>>>  - float_dsp.scalarproduct_float [OK]
>>> SSE2:
>>>  - sbrdsp.qmf_pre_shuffle[OK]
>>>  - sbrdsp.qmf_deint_bfly [OK]
>>>
>>> Program received signal SIGSEGV, Segmentation fault.
>>> apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
>>> 418 movu   m7, [Yq + 2*count + mmsize]
>>> (gdb) bt
>>> Python Exception  No module named gdb.frames:
>>> #0  apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
>>> #1  0x0043659b in checkasm_checked_call () at 
>>> tests/checkasm/x86/checkasm.asm:77
>>> #2  0xdeadbeefdeadbeef in ?? ()
>>> #3  0xdeadbeefdeadbeef in ?? ()
>>> #4  0xdeadbeefdeadbeef in ?? ()
>>> #5  0xdeadbeefdeadbeef in ?? ()
>>> #6  0xdeadbeefdeadbeef in ?? ()
>>> #7  0xdeadbeefdeadbeef in ?? ()
>>> #8  0xdeadbeefdeadbeef in ?? ()
>>> #9  0xdeadbeefdeadbeef in ?? ()
>>> #10 0xdeadbeefdeadbeef in ?? ()
>>> #11 0xdeadbeefdeadbeef in ?? ()
>>> #12 0xdeadbeefdeadbeef in ?? ()
>>> #13 0xdeadbeefdeadbeef in ?? ()
>>> #14 0xdeadbeefdeadbeef in ?? ()
>>> #15 0xdeadbeefdeadbeef in ?? ()
>>> #16 0xdeadbeefdeadbeef in ?? ()
>>> #17 0xdeadbeefdeadbeef in ?? ()
>>> #18 0xdeadbeefdeadbeef in ?? ()
>>> #19 0x7fffd870 in ?? ()
>>> #20 0x7fffcc70 in ?? ()
>>> #21 0x7fffce70 in ?? ()
>>> #22 0x in ?? ()
>>> (gdb) info all-registers
>>>

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-07-04 Thread Michael Niedermayer 
On Tue, Jul 04, 2017 at 03:15:56PM -0300, James Almer wrote:
> On 7/4/2017 2:31 PM, Michael Niedermayer wrote:
> > On Mon, Jul 03, 2017 at 02:32:28PM +0200, Matthieu Bouron wrote:
> >> On Fri, Jun 30, 2017 at 05:16:37PM +0200, Matthieu Bouron wrote:
> >>> On Fri, Jun 30, 2017 at 03:55:52PM +0200, Michael Niedermayer wrote:
>  On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
> > On 6/29/2017 10:14 PM, Henrik Gramner wrote:
> >> On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
> >>  wrote:
> >>> Program received signal SIGSEGV, Segmentation fault.
> >>> 0x00684919 in ff_sbr_hf_gen_sse ()
> >>
> >>>0x00684909 :   sub%r9,%r8
> >>
> >>> => 0x00684919 :   movaps 
> >>> (%rsi,%r8,1),%xmm0
> >>
> >>> r9 0xdeadbeef0080   -2401053092612145024
> >>
> >> Another case of a 32-bit int being used as part of a 64-bit operation.
> >
> > I can't reproduce it on my ArchLinux x86_64 environment for some reason,
> > but based on what you said i assume the attached patch should fix it.
> 
>  no crash occurs here with this, so it seems fixed
> >>>
> >>> Should i push the patchset or wait a little bit longer ?
> >>
> >> Patchset applied.
> > 
> > it seems theres some issue still in this:
> > 
> > checkasm: using random seed 3655967467
> > MMX:
> >  - audiodsp.audiodsp [OK]
> >  - blockdsp.blockdsp [OK]
> >  - h264dsp.idct  [OK]
> >  - h264pred.pred4x4  [OK]
> >  - h264pred.pred8x8  [OK]
> >  - h264pred.pred16x16[OK]
> >  - pixblockdsp.get_pixels[OK]
> >  - pixblockdsp.diff_pixels   [OK]
> >  - vp8dsp.idct   [OK]
> >  - vp8dsp.mc [OK]
> >  - vp9dsp.ipred  [OK]
> >  - vp9dsp.itxfm  [OK]
> >  - vp9dsp.mc [OK]
> > MMXEXT:
> >  - audiodsp.audiodsp [OK]
> >  - h264dsp.idct  [OK]
> >  - h264pred.pred4x4  [OK]
> >  - h264pred.pred8x8  [OK]
> >  - h264pred.pred16x16[OK]
> >  - h264pred.pred8x8l [OK]
> >  - h264qpel.put  [OK]
> >  - h264qpel.avg  [OK]
> >  - hevc_add_res.add_residual [OK]
> >  - hevc_idct.idct_dc [OK]
> >  - vp8dsp.mc [OK]
> >  - vp9dsp.ipred  [OK]
> >  - vp9dsp.itxfm  [OK]
> >  - vp9dsp.loopfilter [OK]
> >  - vp9dsp.mc [OK]
> > SSE:
> >  - aacpsdsp.add_squares  [OK]
> >  - aacpsdsp.mul_pair_single  [OK]
> >  - aacpsdsp.hybrid_analysis  [OK]
> >  - sbrdsp.sum64x5[OK]
> >  - sbrdsp.sum_square [OK]
> >  - sbrdsp.neg_odd_64 [OK]
> >  - sbrdsp.qmf_post_shuffle   [OK]
> >  - sbrdsp.qmf_deint_neg  [OK]
> >  - sbrdsp.qmf_deint_bfly [OK]
> >  - sbrdsp.autocorrelate  [OK]
> >  - sbrdsp.hf_gen [OK]
> >  - sbrdsp.hf_g_filt  [OK]
> >  - audiodsp.audiodsp [OK]
> >  - blockdsp.blockdsp [OK]
> >  - fmtconvert.fmtconvert [OK]
> >  - h264pred.pred16x16[OK]
> >  - vp8dsp.idct   [OK]
> >  - vp8dsp.mc [OK]
> >  - vp9dsp.ipred  [OK]
> >  - vp9dsp.mc [OK]
> >  - float_dsp.vector_fmul [OK]
> >  - float_dsp.vector_fmac [OK]
> >  - float_dsp.butterflies_float   [OK]
> >  - float_dsp.scalarproduct_float [OK]
> > SSE2:
> >  - sbrdsp.qmf_pre_shuffle[OK]
> >  - sbrdsp.qmf_deint_bfly [OK]
> > 
> > Program received signal SIGSEGV, Segmentation fault.
> > apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
> > 418 movu   m7, [Yq + 2*count + mmsize]
> > (gdb) bt
> > Python Exception  No module named gdb.frames:
> > #0  apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
> > #1  0x0043659b in checkasm_checked_call () at 
> > tests/checkasm/x86/checkasm.asm:77
> > #2  0xdeadbeefdeadbeef in ?? ()
> > #3  0xdeadbeefdeadbeef in ?? ()
> > #4  0xdeadbeefdeadbeef in ?? ()
> > #5  0xdeadbeefdeadbeef in ?? ()
> > #6  0xdeadbeefdeadbeef in ?? ()
> > #7  0xdeadbeefdeadbeef in ?? ()
> > #8  0xdeadbeefdeadbeef in ?? ()
> > #9  0xdeadbeefdeadbeef in ?? ()
> > #10 0xdeadbeefdeadbeef in ?? ()
> > #11 0xdeadbeefdeadbeef in ?? ()
> > #12 0xdeadbeefdeadbeef in ?? ()
> > #13 0xdeadbeefdeadbeef in ?? ()
> > #14 0xdeadbeefdeadbeef in ?? ()
> > #15 0xdeadbeefdeadbeef in ?? ()
> > #16 0xdeadbeefdeadbeef in ?? ()
> > #17 0xdeadbeefdeadbeef in ?? ()
> > #18 0xdeadbeefdeadbeef in ?? ()
> > #19 0x7fffd870 in ?? ()
> > #20 0x7fffcc70 in ?? ()
> > #21 0x7fffce70 in ?? ()
> > #22 0x in ?? ()
> > (gdb) info all-registers
> > rax0x0  0
> > rbx

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-07-04 Thread James Almer 
On 7/4/2017 2:31 PM, Michael Niedermayer wrote:
> On Mon, Jul 03, 2017 at 02:32:28PM +0200, Matthieu Bouron wrote:
>> On Fri, Jun 30, 2017 at 05:16:37PM +0200, Matthieu Bouron wrote:
>>> On Fri, Jun 30, 2017 at 03:55:52PM +0200, Michael Niedermayer wrote:
 On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
> On 6/29/2017 10:14 PM, Henrik Gramner wrote:
>> On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
>>  wrote:
>>> Program received signal SIGSEGV, Segmentation fault.
>>> 0x00684919 in ff_sbr_hf_gen_sse ()
>>
>>>0x00684909 :   sub%r9,%r8
>>
>>> => 0x00684919 :   movaps 
>>> (%rsi,%r8,1),%xmm0
>>
>>> r9 0xdeadbeef0080   -2401053092612145024
>>
>> Another case of a 32-bit int being used as part of a 64-bit operation.
>
> I can't reproduce it on my ArchLinux x86_64 environment for some reason,
> but based on what you said i assume the attached patch should fix it.

 no crash occurs here with this, so it seems fixed
>>>
>>> Should i push the patchset or wait a little bit longer ?
>>
>> Patchset applied.
> 
> it seems theres some issue still in this:
> 
> checkasm: using random seed 3655967467
> MMX:
>  - audiodsp.audiodsp [OK]
>  - blockdsp.blockdsp [OK]
>  - h264dsp.idct  [OK]
>  - h264pred.pred4x4  [OK]
>  - h264pred.pred8x8  [OK]
>  - h264pred.pred16x16[OK]
>  - pixblockdsp.get_pixels[OK]
>  - pixblockdsp.diff_pixels   [OK]
>  - vp8dsp.idct   [OK]
>  - vp8dsp.mc [OK]
>  - vp9dsp.ipred  [OK]
>  - vp9dsp.itxfm  [OK]
>  - vp9dsp.mc [OK]
> MMXEXT:
>  - audiodsp.audiodsp [OK]
>  - h264dsp.idct  [OK]
>  - h264pred.pred4x4  [OK]
>  - h264pred.pred8x8  [OK]
>  - h264pred.pred16x16[OK]
>  - h264pred.pred8x8l [OK]
>  - h264qpel.put  [OK]
>  - h264qpel.avg  [OK]
>  - hevc_add_res.add_residual [OK]
>  - hevc_idct.idct_dc [OK]
>  - vp8dsp.mc [OK]
>  - vp9dsp.ipred  [OK]
>  - vp9dsp.itxfm  [OK]
>  - vp9dsp.loopfilter [OK]
>  - vp9dsp.mc [OK]
> SSE:
>  - aacpsdsp.add_squares  [OK]
>  - aacpsdsp.mul_pair_single  [OK]
>  - aacpsdsp.hybrid_analysis  [OK]
>  - sbrdsp.sum64x5[OK]
>  - sbrdsp.sum_square [OK]
>  - sbrdsp.neg_odd_64 [OK]
>  - sbrdsp.qmf_post_shuffle   [OK]
>  - sbrdsp.qmf_deint_neg  [OK]
>  - sbrdsp.qmf_deint_bfly [OK]
>  - sbrdsp.autocorrelate  [OK]
>  - sbrdsp.hf_gen [OK]
>  - sbrdsp.hf_g_filt  [OK]
>  - audiodsp.audiodsp [OK]
>  - blockdsp.blockdsp [OK]
>  - fmtconvert.fmtconvert [OK]
>  - h264pred.pred16x16[OK]
>  - vp8dsp.idct   [OK]
>  - vp8dsp.mc [OK]
>  - vp9dsp.ipred  [OK]
>  - vp9dsp.mc [OK]
>  - float_dsp.vector_fmul [OK]
>  - float_dsp.vector_fmac [OK]
>  - float_dsp.butterflies_float   [OK]
>  - float_dsp.scalarproduct_float [OK]
> SSE2:
>  - sbrdsp.qmf_pre_shuffle[OK]
>  - sbrdsp.qmf_deint_bfly [OK]
> 
> Program received signal SIGSEGV, Segmentation fault.
> apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
> 418 movu   m7, [Yq + 2*count + mmsize]
> (gdb) bt
> Python Exception  No module named gdb.frames:
> #0  apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
> #1  0x0043659b in checkasm_checked_call () at 
> tests/checkasm/x86/checkasm.asm:77
> #2  0xdeadbeefdeadbeef in ?? ()
> #3  0xdeadbeefdeadbeef in ?? ()
> #4  0xdeadbeefdeadbeef in ?? ()
> #5  0xdeadbeefdeadbeef in ?? ()
> #6  0xdeadbeefdeadbeef in ?? ()
> #7  0xdeadbeefdeadbeef in ?? ()
> #8  0xdeadbeefdeadbeef in ?? ()
> #9  0xdeadbeefdeadbeef in ?? ()
> #10 0xdeadbeefdeadbeef in ?? ()
> #11 0xdeadbeefdeadbeef in ?? ()
> #12 0xdeadbeefdeadbeef in ?? ()
> #13 0xdeadbeefdeadbeef in ?? ()
> #14 0xdeadbeefdeadbeef in ?? ()
> #15 0xdeadbeefdeadbeef in ?? ()
> #16 0xdeadbeefdeadbeef in ?? ()
> #17 0xdeadbeefdeadbeef in ?? ()
> #18 0xdeadbeefdeadbeef in ?? ()
> #19 0x7fffd870 in ?? ()
> #20 0x7fffcc70 in ?? ()
> #21 0x7fffce70 in ?? ()
> #22 0x in ?? ()
> (gdb) info all-registers
> rax0x0  0
> rbx0xed56bb2dcb3c7736   -1344681633365854410
> rcx0x8e82280
> rdx0x7ab77bbbd070   8842672440749314160
> rsi0x7ab77bbbce70   8842672440749313648
> rdi0xf56edc70   -761539929699263376
> rbp0x8bda43d3fd1a7e06

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-07-04 Thread Michael Niedermayer 
On Mon, Jul 03, 2017 at 02:32:28PM +0200, Matthieu Bouron wrote:
> On Fri, Jun 30, 2017 at 05:16:37PM +0200, Matthieu Bouron wrote:
> > On Fri, Jun 30, 2017 at 03:55:52PM +0200, Michael Niedermayer wrote:
> > > On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
> > > > On 6/29/2017 10:14 PM, Henrik Gramner wrote:
> > > > > On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
> > > > >  wrote:
> > > > >> Program received signal SIGSEGV, Segmentation fault.
> > > > >> 0x00684919 in ff_sbr_hf_gen_sse ()
> > > > > 
> > > > >>0x00684909 :   sub%r9,%r8
> > > > > 
> > > > >> => 0x00684919 :   movaps 
> > > > >> (%rsi,%r8,1),%xmm0
> > > > > 
> > > > >> r9 0xdeadbeef0080   -2401053092612145024
> > > > > 
> > > > > Another case of a 32-bit int being used as part of a 64-bit operation.
> > > > 
> > > > I can't reproduce it on my ArchLinux x86_64 environment for some reason,
> > > > but based on what you said i assume the attached patch should fix it.
> > > 
> > > no crash occurs here with this, so it seems fixed
> > 
> > Should i push the patchset or wait a little bit longer ?
> 
> Patchset applied.

it seems theres some issue still in this:

checkasm: using random seed 3655967467
MMX:
 - audiodsp.audiodsp [OK]
 - blockdsp.blockdsp [OK]
 - h264dsp.idct  [OK]
 - h264pred.pred4x4  [OK]
 - h264pred.pred8x8  [OK]
 - h264pred.pred16x16[OK]
 - pixblockdsp.get_pixels[OK]
 - pixblockdsp.diff_pixels   [OK]
 - vp8dsp.idct   [OK]
 - vp8dsp.mc [OK]
 - vp9dsp.ipred  [OK]
 - vp9dsp.itxfm  [OK]
 - vp9dsp.mc [OK]
MMXEXT:
 - audiodsp.audiodsp [OK]
 - h264dsp.idct  [OK]
 - h264pred.pred4x4  [OK]
 - h264pred.pred8x8  [OK]
 - h264pred.pred16x16[OK]
 - h264pred.pred8x8l [OK]
 - h264qpel.put  [OK]
 - h264qpel.avg  [OK]
 - hevc_add_res.add_residual [OK]
 - hevc_idct.idct_dc [OK]
 - vp8dsp.mc [OK]
 - vp9dsp.ipred  [OK]
 - vp9dsp.itxfm  [OK]
 - vp9dsp.loopfilter [OK]
 - vp9dsp.mc [OK]
SSE:
 - aacpsdsp.add_squares  [OK]
 - aacpsdsp.mul_pair_single  [OK]
 - aacpsdsp.hybrid_analysis  [OK]
 - sbrdsp.sum64x5[OK]
 - sbrdsp.sum_square [OK]
 - sbrdsp.neg_odd_64 [OK]
 - sbrdsp.qmf_post_shuffle   [OK]
 - sbrdsp.qmf_deint_neg  [OK]
 - sbrdsp.qmf_deint_bfly [OK]
 - sbrdsp.autocorrelate  [OK]
 - sbrdsp.hf_gen [OK]
 - sbrdsp.hf_g_filt  [OK]
 - audiodsp.audiodsp [OK]
 - blockdsp.blockdsp [OK]
 - fmtconvert.fmtconvert [OK]
 - h264pred.pred16x16[OK]
 - vp8dsp.idct   [OK]
 - vp8dsp.mc [OK]
 - vp9dsp.ipred  [OK]
 - vp9dsp.mc [OK]
 - float_dsp.vector_fmul [OK]
 - float_dsp.vector_fmac [OK]
 - float_dsp.butterflies_float   [OK]
 - float_dsp.scalarproduct_float [OK]
SSE2:
 - sbrdsp.qmf_pre_shuffle[OK]
 - sbrdsp.qmf_deint_bfly [OK]

Program received signal SIGSEGV, Segmentation fault.
apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
418 movu   m7, [Yq + 2*count + mmsize]
(gdb) bt
Python Exception  No module named gdb.frames:
#0  apply_noise_main.loop () at libavcodec/x86/sbrdsp.asm:418
#1  0x0043659b in checkasm_checked_call () at 
tests/checkasm/x86/checkasm.asm:77
#2  0xdeadbeefdeadbeef in ?? ()
#3  0xdeadbeefdeadbeef in ?? ()
#4  0xdeadbeefdeadbeef in ?? ()
#5  0xdeadbeefdeadbeef in ?? ()
#6  0xdeadbeefdeadbeef in ?? ()
#7  0xdeadbeefdeadbeef in ?? ()
#8  0xdeadbeefdeadbeef in ?? ()
#9  0xdeadbeefdeadbeef in ?? ()
#10 0xdeadbeefdeadbeef in ?? ()
#11 0xdeadbeefdeadbeef in ?? ()
#12 0xdeadbeefdeadbeef in ?? ()
#13 0xdeadbeefdeadbeef in ?? ()
#14 0xdeadbeefdeadbeef in ?? ()
#15 0xdeadbeefdeadbeef in ?? ()
#16 0xdeadbeefdeadbeef in ?? ()
#17 0xdeadbeefdeadbeef in ?? ()
#18 0xdeadbeefdeadbeef in ?? ()
#19 0x7fffd870 in ?? ()
#20 0x7fffcc70 in ?? ()
#21 0x7fffce70 in ?? ()
#22 0x in ?? ()
(gdb) info all-registers
rax0x0  0
rbx0xed56bb2dcb3c7736   -1344681633365854410
rcx0x8e82280
rdx0x7ab77bbbd070   8842672440749314160
rsi0x7ab77bbbce70   8842672440749313648
rdi0xf56edc70   -761539929699263376
rbp0x8bda43d3fd1a7e06   0x8bda43d3fd1a7e06
rsp0x7fffcae8   0x7fffcae8
r8 0xdeadbeef   -2401053092612145152
r9 0x8549044409c0   -8842531703260968512
r10

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-07-03 Thread Matthieu Bouron 
On Fri, Jun 30, 2017 at 05:16:37PM +0200, Matthieu Bouron wrote:
> On Fri, Jun 30, 2017 at 03:55:52PM +0200, Michael Niedermayer wrote:
> > On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
> > > On 6/29/2017 10:14 PM, Henrik Gramner wrote:
> > > > On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
> > > >  wrote:
> > > >> Program received signal SIGSEGV, Segmentation fault.
> > > >> 0x00684919 in ff_sbr_hf_gen_sse ()
> > > > 
> > > >>0x00684909 :   sub%r9,%r8
> > > > 
> > > >> => 0x00684919 :   movaps 
> > > >> (%rsi,%r8,1),%xmm0
> > > > 
> > > >> r9 0xdeadbeef0080   -2401053092612145024
> > > > 
> > > > Another case of a 32-bit int being used as part of a 64-bit operation.
> > > 
> > > I can't reproduce it on my ArchLinux x86_64 environment for some reason,
> > > but based on what you said i assume the attached patch should fix it.
> > 
> > no crash occurs here with this, so it seems fixed
> 
> Should i push the patchset or wait a little bit longer ?

Patchset applied.

-- 
Matthieu B.
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-06-30 Thread Matthieu Bouron 
On Fri, Jun 30, 2017 at 03:55:52PM +0200, Michael Niedermayer wrote:
> On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
> > On 6/29/2017 10:14 PM, Henrik Gramner wrote:
> > > On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
> > >  wrote:
> > >> Program received signal SIGSEGV, Segmentation fault.
> > >> 0x00684919 in ff_sbr_hf_gen_sse ()
> > > 
> > >>0x00684909 :   sub%r9,%r8
> > > 
> > >> => 0x00684919 :   movaps (%rsi,%r8,1),%xmm0
> > > 
> > >> r9 0xdeadbeef0080   -2401053092612145024
> > > 
> > > Another case of a 32-bit int being used as part of a 64-bit operation.
> > 
> > I can't reproduce it on my ArchLinux x86_64 environment for some reason,
> > but based on what you said i assume the attached patch should fix it.
> 
> no crash occurs here with this, so it seems fixed

Should i push the patchset or wait a little bit longer ?

[...]

-- 
Matthieu B.
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-06-30 Thread James Almer 
On 6/30/2017 10:55 AM, Michael Niedermayer wrote:
> On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
>> On 6/29/2017 10:14 PM, Henrik Gramner wrote:
>>> On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
>>>  wrote:
 Program received signal SIGSEGV, Segmentation fault.
 0x00684919 in ff_sbr_hf_gen_sse ()
>>>
0x00684909 :   sub%r9,%r8
>>>
 => 0x00684919 :   movaps (%rsi,%r8,1),%xmm0
>>>
 r9 0xdeadbeef0080   -2401053092612145024
>>>
>>> Another case of a 32-bit int being used as part of a 64-bit operation.
>>
>> I can't reproduce it on my ArchLinux x86_64 environment for some reason,
>> but based on what you said i assume the attached patch should fix it.
> 
> no crash occurs here with this, so it seems fixed
> 
> thx

Applied then.
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-06-30 Thread Michael Niedermayer 
On Thu, Jun 29, 2017 at 10:53:06PM -0300, James Almer wrote:
> On 6/29/2017 10:14 PM, Henrik Gramner wrote:
> > On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
> >  wrote:
> >> Program received signal SIGSEGV, Segmentation fault.
> >> 0x00684919 in ff_sbr_hf_gen_sse ()
> > 
> >>0x00684909 :   sub%r9,%r8
> > 
> >> => 0x00684919 :   movaps (%rsi,%r8,1),%xmm0
> > 
> >> r9 0xdeadbeef0080   -2401053092612145024
> > 
> > Another case of a 32-bit int being used as part of a 64-bit operation.
> 
> I can't reproduce it on my ArchLinux x86_64 environment for some reason,
> but based on what you said i assume the attached patch should fix it.

no crash occurs here with this, so it seems fixed

thx

[...]

-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Let us carefully observe those good qualities wherein our enemies excel us
and endeavor to excel them, by avoiding what is faulty, and imitating what
is excellent in them. -- Plutarch


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-06-29 Thread James Almer 
On 6/29/2017 10:14 PM, Henrik Gramner wrote:
> On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
>  wrote:
>> Program received signal SIGSEGV, Segmentation fault.
>> 0x00684919 in ff_sbr_hf_gen_sse ()
> 
>>0x00684909 :   sub%r9,%r8
> 
>> => 0x00684919 :   movaps (%rsi,%r8,1),%xmm0
> 
>> r9 0xdeadbeef0080   -2401053092612145024
> 
> Another case of a 32-bit int being used as part of a 64-bit operation.

I can't reproduce it on my ArchLinux x86_64 environment for some reason,
but based on what you said i assume the attached patch should fix it.
From f4646091b450b7c4c5479fbb4163ef89615a4a8d Mon Sep 17 00:00:00 2001
From: James Almer 
Date: Thu, 29 Jun 2017 22:51:04 -0300
Subject: [PATCH] x86/sbrdsp: zero extend start and end gprs in
 ff_sbr_hf_gen_sse

Signed-off-by: James Almer 
---
 libavcodec/x86/sbrdsp.asm | 26 +-
 1 file changed, 13 insertions(+), 13 deletions(-)

diff --git a/libavcodec/x86/sbrdsp.asm b/libavcodec/x86/sbrdsp.asm
index d0f774b277..c716184b14 100644
--- a/libavcodec/x86/sbrdsp.asm
+++ b/libavcodec/x86/sbrdsp.asm
@@ -149,19 +149,19 @@ cglobal sbr_hf_gen, 4,4,8, X_high, X_low, alpha0, alpha1, 
BW, S, E
 ; start and end 6th and 7th args on stack
 movr2d, Sm
 movr3d, Em
-%define  start r2q
-%define  end   r3q
+DEFINE_ARGS X_high, X_low, start, end
 %else
 ; BW does not actually occupy a register, so shift by 1
-%define  start BWq
-%define  end   Sq
+DEFINE_ARGS X_high, X_low, alpha0, alpha1, start, end
+movsxd  startq, startd
+movsxdendq, endd
 %endif
-sub  start, end  ; neg num of loops
-leaX_highq, [X_highq + end*2*4]
-lea X_lowq, [X_lowq  + end*2*4 - 2*2*4]
-shl  start, 3; offset from num loops
+sub startq, endq ; neg num of loops
+leaX_highq, [X_highq + endq*2*4]
+lea X_lowq, [X_lowq  + endq*2*4 - 2*2*4]
+shl startq, 3; offset from num loops
 
-movam0, [X_lowq + start]
+movam0, [X_lowq + startq]
 shufps  m3, m3, q
 shufps  m4, m4, q
 xorps   m3, [ps_mask]
@@ -169,7 +169,7 @@ cglobal sbr_hf_gen, 4,4,8, X_high, X_low, alpha0, alpha1, 
BW, S, E
 shufps  m2, m2, q
 xorps   m4, [ps_mask]
 .loop2:
-movum7, [X_lowq + start + 8]; BbCc
+movum7, [X_lowq + startq + 8]   ; BbCc
 movam6, m0
 movam5, m7
 shufps  m0, m0, q2301   ; aAbB
@@ -179,12 +179,12 @@ cglobal sbr_hf_gen, 4,4,8, X_high, X_low, alpha0, alpha1, 
BW, S, E
 mulps   m6, m2
 mulps   m5, m1
 addps   m7, m0
-movam0, [X_lowq + start +16]; CcDd
+movam0, [X_lowq + startq + 16]  ; CcDd
 addps   m7, m0
 addps   m6, m5
 addps   m7, m6
-mova  [X_highq + start], m7
-add start, 16
+mova  [X_highq + startq], m7
+add startq, 16
 jnz .loop2
 RET
 
-- 
2.13.0

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-06-29 Thread Henrik Gramner 
On Fri, Jun 30, 2017 at 1:58 AM, Michael Niedermayer
 wrote:
> Program received signal SIGSEGV, Segmentation fault.
> 0x00684919 in ff_sbr_hf_gen_sse ()

>0x00684909 :   sub%r9,%r8

> => 0x00684919 :   movaps (%rsi,%r8,1),%xmm0

> r9 0xdeadbeef0080   -2401053092612145024

Another case of a 32-bit int being used as part of a 64-bit operation.
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-06-29 Thread Michael Niedermayer 
On Wed, Jun 28, 2017 at 02:48:55PM +0200, Matthieu Bouron wrote:
> On Fri, Jun 23, 2017 at 05:01:35PM +0200, Matthieu Bouron wrote:
> > On Thu, Jun 22, 2017 at 12:53:52PM -0300, James Almer wrote:
> > > On 6/22/2017 9:56 AM, Matthieu Bouron wrote:
> > > > ---
> > > > 
> > > > The following patchset applies on top of Clément's aacpsdsp patchset.
> > > > 
> > > > ---
> > > >  tests/checkasm/Makefile   |   3 +-
> > > >  tests/checkasm/checkasm.c |   1 +
> > > >  tests/checkasm/checkasm.h |   1 +
> > > >  tests/checkasm/sbrdsp.c   | 297 
> > > > ++
> > > >  4 files changed, 301 insertions(+), 1 deletion(-)
> > > >  create mode 100644 tests/checkasm/sbrdsp.c
> > > > 
> > > > diff --git a/tests/checkasm/Makefile b/tests/checkasm/Makefile
> > > > index 638e811931..60e80ab738 100644
> > > > --- a/tests/checkasm/Makefile
> > > > +++ b/tests/checkasm/Makefile
> > > > @@ -13,7 +13,8 @@ AVCODECOBJS-$(CONFIG_VP8DSP)+= vp8dsp.o
> > > >  AVCODECOBJS-$(CONFIG_VIDEODSP)  += videodsp.o
> > > >  
> > > >  # decoders/encoders
> > > > -AVCODECOBJS-$(CONFIG_AAC_DECODER)   += aacpsdsp.o
> > > > +AVCODECOBJS-$(CONFIG_AAC_DECODER)   += aacpsdsp.o \
> > > > +   sbrdsp.o
> > > >  AVCODECOBJS-$(CONFIG_ALAC_DECODER)  += alacdsp.o
> > > >  AVCODECOBJS-$(CONFIG_DCA_DECODER)   += synth_filter.o
> > > >  AVCODECOBJS-$(CONFIG_JPEG2000_DECODER)  += jpeg2000dsp.o
> > > > diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
> > > > index e66744b162..29f201b1b3 100644
> > > > --- a/tests/checkasm/checkasm.c
> > > > +++ b/tests/checkasm/checkasm.c
> > > > @@ -67,6 +67,7 @@ static const struct {
> > > >  #if CONFIG_AVCODEC
> > > >  #if CONFIG_AAC_DECODER
> > > >  { "aacpsdsp", checkasm_check_aacpsdsp },
> > > > +{ "sbrdsp",   checkasm_check_sbrdsp },
> > > >  #endif
> > > >  #if CONFIG_ALAC_DECODER
> > > >  { "alacdsp", checkasm_check_alacdsp },
> > > > diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h
> > > > index dfb0ce561c..fa51e71e4b 100644
> > > > --- a/tests/checkasm/checkasm.h
> > > > +++ b/tests/checkasm/checkasm.h
> > > > @@ -50,6 +50,7 @@ void checkasm_check_hevc_idct(void);
> > > >  void checkasm_check_jpeg2000dsp(void);
> > > >  void checkasm_check_llviddsp(void);
> > > >  void checkasm_check_pixblockdsp(void);
> > > > +void checkasm_check_sbrdsp(void);
> > > >  void checkasm_check_synth_filter(void);
> > > >  void checkasm_check_v210enc(void);
> > > >  void checkasm_check_vp8dsp(void);
> > > > diff --git a/tests/checkasm/sbrdsp.c b/tests/checkasm/sbrdsp.c
> > > > new file mode 100644
> > > > index 00..8333510c6b
> > > > --- /dev/null
> > > > +++ b/tests/checkasm/sbrdsp.c
> > > > @@ -0,0 +1,297 @@
> > > > +/*
> > > > + * This file is part of FFmpeg.
> > > > + *
> > > > + * FFmpeg is free software; you can redistribute it and/or modify
> > > > + * it under the terms of the GNU General Public License as published by
> > > > + * the Free Software Foundation; either version 2 of the License, or
> > > > + * (at your option) any later version.
> > > > + *
> > > > + * FFmpeg is distributed in the hope that it will be useful,
> > > > + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> > > > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > > > + * GNU General Public License for more details.
> > > > + *
> > > > + * You should have received a copy of the GNU General Public License 
> > > > along
> > > > + * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
> > > > + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
> > > > + */
> > > > +
> > > > +#include "libavcodec/sbrdsp.h"
> > > > +
> > > > +#include "checkasm.h"
> > > > +
> > > > +#define randomize(buf, len) do {\
> > > > +int i;  \
> > > > +for (i = 0; i < len; i++) { \
> > > > +const INTFLOAT f = (INTFLOAT)rnd() / UINT_MAX;  \
> > > > +(buf)[i] = f;   \
> > > > +}   \
> > > > +} while (0)
> > > > +
> > > > +#define EPS 0.0001
> > > > +
> > > > +static void test_sum64x5(void)
> > > > +{
> > > > +LOCAL_ALIGNED_16(INTFLOAT, dst0, [64 + 256]);
> > > > +LOCAL_ALIGNED_16(INTFLOAT, dst1, [64 + 256]);
> > > > +
> > > > +declare_func(void, INTFLOAT *z);
> > > > +
> > > > +randomize((INTFLOAT *)dst0, 64 + 256);
> > > > +memcpy(dst1, dst0, (64 + 256) * sizeof(INTFLOAT));
> > > > +call_ref(dst0);
> > > > +call_new(dst1);
> > > > +if (!float_near_abs_eps_array(dst0, dst1, EPS, 64 + 256))
> > > > +fail();
> > > > +bench_new(dst1);
> > > > +}
> > > > +
> > > > +static void test_sum_square(void)
> > > > +{
> > > > +INTFLOAT res0;
> > > > +INTFLOAT

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-06-28 Thread Matthieu Bouron 
On Fri, Jun 23, 2017 at 05:01:35PM +0200, Matthieu Bouron wrote:
> On Thu, Jun 22, 2017 at 12:53:52PM -0300, James Almer wrote:
> > On 6/22/2017 9:56 AM, Matthieu Bouron wrote:
> > > ---
> > > 
> > > The following patchset applies on top of Clément's aacpsdsp patchset.
> > > 
> > > ---
> > >  tests/checkasm/Makefile   |   3 +-
> > >  tests/checkasm/checkasm.c |   1 +
> > >  tests/checkasm/checkasm.h |   1 +
> > >  tests/checkasm/sbrdsp.c   | 297 
> > > ++
> > >  4 files changed, 301 insertions(+), 1 deletion(-)
> > >  create mode 100644 tests/checkasm/sbrdsp.c
> > > 
> > > diff --git a/tests/checkasm/Makefile b/tests/checkasm/Makefile
> > > index 638e811931..60e80ab738 100644
> > > --- a/tests/checkasm/Makefile
> > > +++ b/tests/checkasm/Makefile
> > > @@ -13,7 +13,8 @@ AVCODECOBJS-$(CONFIG_VP8DSP)+= vp8dsp.o
> > >  AVCODECOBJS-$(CONFIG_VIDEODSP)  += videodsp.o
> > >  
> > >  # decoders/encoders
> > > -AVCODECOBJS-$(CONFIG_AAC_DECODER)   += aacpsdsp.o
> > > +AVCODECOBJS-$(CONFIG_AAC_DECODER)   += aacpsdsp.o \
> > > +   sbrdsp.o
> > >  AVCODECOBJS-$(CONFIG_ALAC_DECODER)  += alacdsp.o
> > >  AVCODECOBJS-$(CONFIG_DCA_DECODER)   += synth_filter.o
> > >  AVCODECOBJS-$(CONFIG_JPEG2000_DECODER)  += jpeg2000dsp.o
> > > diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
> > > index e66744b162..29f201b1b3 100644
> > > --- a/tests/checkasm/checkasm.c
> > > +++ b/tests/checkasm/checkasm.c
> > > @@ -67,6 +67,7 @@ static const struct {
> > >  #if CONFIG_AVCODEC
> > >  #if CONFIG_AAC_DECODER
> > >  { "aacpsdsp", checkasm_check_aacpsdsp },
> > > +{ "sbrdsp",   checkasm_check_sbrdsp },
> > >  #endif
> > >  #if CONFIG_ALAC_DECODER
> > >  { "alacdsp", checkasm_check_alacdsp },
> > > diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h
> > > index dfb0ce561c..fa51e71e4b 100644
> > > --- a/tests/checkasm/checkasm.h
> > > +++ b/tests/checkasm/checkasm.h
> > > @@ -50,6 +50,7 @@ void checkasm_check_hevc_idct(void);
> > >  void checkasm_check_jpeg2000dsp(void);
> > >  void checkasm_check_llviddsp(void);
> > >  void checkasm_check_pixblockdsp(void);
> > > +void checkasm_check_sbrdsp(void);
> > >  void checkasm_check_synth_filter(void);
> > >  void checkasm_check_v210enc(void);
> > >  void checkasm_check_vp8dsp(void);
> > > diff --git a/tests/checkasm/sbrdsp.c b/tests/checkasm/sbrdsp.c
> > > new file mode 100644
> > > index 00..8333510c6b
> > > --- /dev/null
> > > +++ b/tests/checkasm/sbrdsp.c
> > > @@ -0,0 +1,297 @@
> > > +/*
> > > + * This file is part of FFmpeg.
> > > + *
> > > + * FFmpeg is free software; you can redistribute it and/or modify
> > > + * it under the terms of the GNU General Public License as published by
> > > + * the Free Software Foundation; either version 2 of the License, or
> > > + * (at your option) any later version.
> > > + *
> > > + * FFmpeg is distributed in the hope that it will be useful,
> > > + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> > > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > > + * GNU General Public License for more details.
> > > + *
> > > + * You should have received a copy of the GNU General Public License 
> > > along
> > > + * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
> > > + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
> > > + */
> > > +
> > > +#include "libavcodec/sbrdsp.h"
> > > +
> > > +#include "checkasm.h"
> > > +
> > > +#define randomize(buf, len) do {\
> > > +int i;  \
> > > +for (i = 0; i < len; i++) { \
> > > +const INTFLOAT f = (INTFLOAT)rnd() / UINT_MAX;  \
> > > +(buf)[i] = f;   \
> > > +}   \
> > > +} while (0)
> > > +
> > > +#define EPS 0.0001
> > > +
> > > +static void test_sum64x5(void)
> > > +{
> > > +LOCAL_ALIGNED_16(INTFLOAT, dst0, [64 + 256]);
> > > +LOCAL_ALIGNED_16(INTFLOAT, dst1, [64 + 256]);
> > > +
> > > +declare_func(void, INTFLOAT *z);
> > > +
> > > +randomize((INTFLOAT *)dst0, 64 + 256);
> > > +memcpy(dst1, dst0, (64 + 256) * sizeof(INTFLOAT));
> > > +call_ref(dst0);
> > > +call_new(dst1);
> > > +if (!float_near_abs_eps_array(dst0, dst1, EPS, 64 + 256))
> > > +fail();
> > > +bench_new(dst1);
> > > +}
> > > +
> > > +static void test_sum_square(void)
> > > +{
> > > +INTFLOAT res0;
> > > +INTFLOAT res1;
> > > +LOCAL_ALIGNED_16(INTFLOAT, src, [256], [2]);
> > > +
> > > +declare_func(INTFLOAT, INTFLOAT (*x)[2], int n);
> > > +
> > > +randomize((INTFLOAT *)src, 256 * 2);
> > > +res0 = call_ref(src, 256);
> > > +res1 = call_new(src, 256);
> > 
> > This one is failing on

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-06-23 Thread Matthieu Bouron 
On Thu, Jun 22, 2017 at 12:53:52PM -0300, James Almer wrote:
> On 6/22/2017 9:56 AM, Matthieu Bouron wrote:
> > ---
> > 
> > The following patchset applies on top of Clément's aacpsdsp patchset.
> > 
> > ---
> >  tests/checkasm/Makefile   |   3 +-
> >  tests/checkasm/checkasm.c |   1 +
> >  tests/checkasm/checkasm.h |   1 +
> >  tests/checkasm/sbrdsp.c   | 297 
> > ++
> >  4 files changed, 301 insertions(+), 1 deletion(-)
> >  create mode 100644 tests/checkasm/sbrdsp.c
> > 
> > diff --git a/tests/checkasm/Makefile b/tests/checkasm/Makefile
> > index 638e811931..60e80ab738 100644
> > --- a/tests/checkasm/Makefile
> > +++ b/tests/checkasm/Makefile
> > @@ -13,7 +13,8 @@ AVCODECOBJS-$(CONFIG_VP8DSP)+= vp8dsp.o
> >  AVCODECOBJS-$(CONFIG_VIDEODSP)  += videodsp.o
> >  
> >  # decoders/encoders
> > -AVCODECOBJS-$(CONFIG_AAC_DECODER)   += aacpsdsp.o
> > +AVCODECOBJS-$(CONFIG_AAC_DECODER)   += aacpsdsp.o \
> > +   sbrdsp.o
> >  AVCODECOBJS-$(CONFIG_ALAC_DECODER)  += alacdsp.o
> >  AVCODECOBJS-$(CONFIG_DCA_DECODER)   += synth_filter.o
> >  AVCODECOBJS-$(CONFIG_JPEG2000_DECODER)  += jpeg2000dsp.o
> > diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
> > index e66744b162..29f201b1b3 100644
> > --- a/tests/checkasm/checkasm.c
> > +++ b/tests/checkasm/checkasm.c
> > @@ -67,6 +67,7 @@ static const struct {
> >  #if CONFIG_AVCODEC
> >  #if CONFIG_AAC_DECODER
> >  { "aacpsdsp", checkasm_check_aacpsdsp },
> > +{ "sbrdsp",   checkasm_check_sbrdsp },
> >  #endif
> >  #if CONFIG_ALAC_DECODER
> >  { "alacdsp", checkasm_check_alacdsp },
> > diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h
> > index dfb0ce561c..fa51e71e4b 100644
> > --- a/tests/checkasm/checkasm.h
> > +++ b/tests/checkasm/checkasm.h
> > @@ -50,6 +50,7 @@ void checkasm_check_hevc_idct(void);
> >  void checkasm_check_jpeg2000dsp(void);
> >  void checkasm_check_llviddsp(void);
> >  void checkasm_check_pixblockdsp(void);
> > +void checkasm_check_sbrdsp(void);
> >  void checkasm_check_synth_filter(void);
> >  void checkasm_check_v210enc(void);
> >  void checkasm_check_vp8dsp(void);
> > diff --git a/tests/checkasm/sbrdsp.c b/tests/checkasm/sbrdsp.c
> > new file mode 100644
> > index 00..8333510c6b
> > --- /dev/null
> > +++ b/tests/checkasm/sbrdsp.c
> > @@ -0,0 +1,297 @@
> > +/*
> > + * This file is part of FFmpeg.
> > + *
> > + * FFmpeg is free software; you can redistribute it and/or modify
> > + * it under the terms of the GNU General Public License as published by
> > + * the Free Software Foundation; either version 2 of the License, or
> > + * (at your option) any later version.
> > + *
> > + * FFmpeg is distributed in the hope that it will be useful,
> > + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > + * GNU General Public License for more details.
> > + *
> > + * You should have received a copy of the GNU General Public License along
> > + * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
> > + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
> > + */
> > +
> > +#include "libavcodec/sbrdsp.h"
> > +
> > +#include "checkasm.h"
> > +
> > +#define randomize(buf, len) do {\
> > +int i;  \
> > +for (i = 0; i < len; i++) { \
> > +const INTFLOAT f = (INTFLOAT)rnd() / UINT_MAX;  \
> > +(buf)[i] = f;   \
> > +}   \
> > +} while (0)
> > +
> > +#define EPS 0.0001
> > +
> > +static void test_sum64x5(void)
> > +{
> > +LOCAL_ALIGNED_16(INTFLOAT, dst0, [64 + 256]);
> > +LOCAL_ALIGNED_16(INTFLOAT, dst1, [64 + 256]);
> > +
> > +declare_func(void, INTFLOAT *z);
> > +
> > +randomize((INTFLOAT *)dst0, 64 + 256);
> > +memcpy(dst1, dst0, (64 + 256) * sizeof(INTFLOAT));
> > +call_ref(dst0);
> > +call_new(dst1);
> > +if (!float_near_abs_eps_array(dst0, dst1, EPS, 64 + 256))
> > +fail();
> > +bench_new(dst1);
> > +}
> > +
> > +static void test_sum_square(void)
> > +{
> > +INTFLOAT res0;
> > +INTFLOAT res1;
> > +LOCAL_ALIGNED_16(INTFLOAT, src, [256], [2]);
> > +
> > +declare_func(INTFLOAT, INTFLOAT (*x)[2], int n);
> > +
> > +randomize((INTFLOAT *)src, 256 * 2);
> > +res0 = call_ref(src, 256);
> > +res1 = call_new(src, 256);
> 
> This one is failing on x86 because the second argument has garbage in
> the upper 32 bits of the grp.
> 
> The solution is to either change n from int to ptrdiff_t, or zero the
> upper bits of n in ff_sbr_sum_square_sse(), which can be done implicitly
> in the first instruction.
> 
> > +if (!float_near_abs_eps(res0, res1, EPS))

Re: [FFmpeg-devel] [PATCH 1/2] checkasm: add sbrdsp tests

2017-06-22 Thread James Almer 
On 6/22/2017 9:56 AM, Matthieu Bouron wrote:
> ---
> 
> The following patchset applies on top of Clément's aacpsdsp patchset.
> 
> ---
>  tests/checkasm/Makefile   |   3 +-
>  tests/checkasm/checkasm.c |   1 +
>  tests/checkasm/checkasm.h |   1 +
>  tests/checkasm/sbrdsp.c   | 297 
> ++
>  4 files changed, 301 insertions(+), 1 deletion(-)
>  create mode 100644 tests/checkasm/sbrdsp.c
> 
> diff --git a/tests/checkasm/Makefile b/tests/checkasm/Makefile
> index 638e811931..60e80ab738 100644
> --- a/tests/checkasm/Makefile
> +++ b/tests/checkasm/Makefile
> @@ -13,7 +13,8 @@ AVCODECOBJS-$(CONFIG_VP8DSP)+= vp8dsp.o
>  AVCODECOBJS-$(CONFIG_VIDEODSP)  += videodsp.o
>  
>  # decoders/encoders
> -AVCODECOBJS-$(CONFIG_AAC_DECODER)   += aacpsdsp.o
> +AVCODECOBJS-$(CONFIG_AAC_DECODER)   += aacpsdsp.o \
> +   sbrdsp.o
>  AVCODECOBJS-$(CONFIG_ALAC_DECODER)  += alacdsp.o
>  AVCODECOBJS-$(CONFIG_DCA_DECODER)   += synth_filter.o
>  AVCODECOBJS-$(CONFIG_JPEG2000_DECODER)  += jpeg2000dsp.o
> diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
> index e66744b162..29f201b1b3 100644
> --- a/tests/checkasm/checkasm.c
> +++ b/tests/checkasm/checkasm.c
> @@ -67,6 +67,7 @@ static const struct {
>  #if CONFIG_AVCODEC
>  #if CONFIG_AAC_DECODER
>  { "aacpsdsp", checkasm_check_aacpsdsp },
> +{ "sbrdsp",   checkasm_check_sbrdsp },
>  #endif
>  #if CONFIG_ALAC_DECODER
>  { "alacdsp", checkasm_check_alacdsp },
> diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h
> index dfb0ce561c..fa51e71e4b 100644
> --- a/tests/checkasm/checkasm.h
> +++ b/tests/checkasm/checkasm.h
> @@ -50,6 +50,7 @@ void checkasm_check_hevc_idct(void);
>  void checkasm_check_jpeg2000dsp(void);
>  void checkasm_check_llviddsp(void);
>  void checkasm_check_pixblockdsp(void);
> +void checkasm_check_sbrdsp(void);
>  void checkasm_check_synth_filter(void);
>  void checkasm_check_v210enc(void);
>  void checkasm_check_vp8dsp(void);
> diff --git a/tests/checkasm/sbrdsp.c b/tests/checkasm/sbrdsp.c
> new file mode 100644
> index 00..8333510c6b
> --- /dev/null
> +++ b/tests/checkasm/sbrdsp.c
> @@ -0,0 +1,297 @@
> +/*
> + * This file is part of FFmpeg.
> + *
> + * FFmpeg is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
> + *
> + * FFmpeg is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License along
> + * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
> + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
> + */
> +
> +#include "libavcodec/sbrdsp.h"
> +
> +#include "checkasm.h"
> +
> +#define randomize(buf, len) do {\
> +int i;  \
> +for (i = 0; i < len; i++) { \
> +const INTFLOAT f = (INTFLOAT)rnd() / UINT_MAX;  \
> +(buf)[i] = f;   \
> +}   \
> +} while (0)
> +
> +#define EPS 0.0001
> +
> +static void test_sum64x5(void)
> +{
> +LOCAL_ALIGNED_16(INTFLOAT, dst0, [64 + 256]);
> +LOCAL_ALIGNED_16(INTFLOAT, dst1, [64 + 256]);
> +
> +declare_func(void, INTFLOAT *z);
> +
> +randomize((INTFLOAT *)dst0, 64 + 256);
> +memcpy(dst1, dst0, (64 + 256) * sizeof(INTFLOAT));
> +call_ref(dst0);
> +call_new(dst1);
> +if (!float_near_abs_eps_array(dst0, dst1, EPS, 64 + 256))
> +fail();
> +bench_new(dst1);
> +}
> +
> +static void test_sum_square(void)
> +{
> +INTFLOAT res0;
> +INTFLOAT res1;
> +LOCAL_ALIGNED_16(INTFLOAT, src, [256], [2]);
> +
> +declare_func(INTFLOAT, INTFLOAT (*x)[2], int n);
> +
> +randomize((INTFLOAT *)src, 256 * 2);
> +res0 = call_ref(src, 256);
> +res1 = call_new(src, 256);

This one is failing on x86 because the second argument has garbage in
the upper 32 bits of the grp.

The solution is to either change n from int to ptrdiff_t, or zero the
upper bits of n in ff_sbr_sum_square_sse(), which can be done implicitly
in the first instruction.

> +if (!float_near_abs_eps(res0, res1, EPS))
> +fail();
> +bench_new(src, 256);
> +}
> +
> +static void test_neg_odd_64(void)
> +{
> +LOCAL_ALIGNED_16(INTFLOAT, dst0, [64]);
> +LOCAL_ALIGNED_16(INTFLOAT, dst1, [64]);
> +
> +declare_func(void, INTFLOAT *x);
> +
> +randomize((INTFLOAT *)dst0, 64);
> +memcpy(dst1, dst0, (64) *