Re: [FFmpeg-devel] [PATCH] tests/checkasm/float_dsp: Increase allowed difference for float_dsp.vector_dmul
On 11/26/2017 8:09 PM, Michael Niedermayer wrote: > On Sun, Nov 26, 2017 at 12:47:22AM +0100, Michael Niedermayer wrote: >> On Sun, Nov 26, 2017 at 12:10:38AM +0100, Michael Niedermayer wrote: >>> On Fri, Nov 24, 2017 at 11:37:36PM -0300, James Almer wrote: On 10/29/2017 11:57 AM, Michael Niedermayer wrote: > The choosen value is the lowest power of 2 that allows 1000 iterations of > fate-checkasm-float_dsp > to pass on x86-32 Ticket #6848 reports this value is still not enough. Maybe something like 1.0e-12 or 1.0e-13 instead? >>> >>> ok, ill push it with 1e-12 >> >> Or do people prefer this: (this should be more correct) >> >> commit 67ba87a320faba623c0b35a0692adb916860ac40 (HEAD -> master) >> Author: Michael Niedermayer >> Date: Sun Oct 29 15:26:50 2017 +0100 >> >> tests/checkasm/float_dsp: Increase allowed difference for >> float_dsp.vector_dmul >> >> Tested for 1 iterations on x86-32 >> >> Fixes: Ticket6848 >> >> Signed-off-by: Michael Niedermayer > > ill push a variant of this so that gets fixed. We can change it > later if people prefer somthing else > leaving it open is bad ... Any solution is fine with me. Thanks for fixing it. ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] tests/checkasm/float_dsp: Increase allowed difference for float_dsp.vector_dmul
On Sun, Nov 26, 2017 at 12:47:22AM +0100, Michael Niedermayer wrote: > On Sun, Nov 26, 2017 at 12:10:38AM +0100, Michael Niedermayer wrote: > > On Fri, Nov 24, 2017 at 11:37:36PM -0300, James Almer wrote: > > > On 10/29/2017 11:57 AM, Michael Niedermayer wrote: > > > > The choosen value is the lowest power of 2 that allows 1000 iterations > > > > of fate-checkasm-float_dsp > > > > to pass on x86-32 > > > > > > Ticket #6848 reports this value is still not enough. Maybe something > > > like 1.0e-12 or 1.0e-13 instead? > > > > ok, ill push it with 1e-12 > > Or do people prefer this: (this should be more correct) > > commit 67ba87a320faba623c0b35a0692adb916860ac40 (HEAD -> master) > Author: Michael Niedermayer > Date: Sun Oct 29 15:26:50 2017 +0100 > > tests/checkasm/float_dsp: Increase allowed difference for > float_dsp.vector_dmul > > Tested for 1 iterations on x86-32 > > Fixes: Ticket6848 > > Signed-off-by: Michael Niedermayer ill push a variant of this so that gets fixed. We can change it later if people prefer somthing else leaving it open is bad ... [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB If you drop bombs on a foreign country and kill a hundred thousand innocent people, expect your government to call the consequence "unprovoked inhuman terrorist attacks" and use it to justify dropping more bombs and killing more people. The technology changed, the idea is old. signature.asc Description: Digital signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] tests/checkasm/float_dsp: Increase allowed difference for float_dsp.vector_dmul
On Sun, Nov 26, 2017 at 12:10:38AM +0100, Michael Niedermayer wrote: > On Fri, Nov 24, 2017 at 11:37:36PM -0300, James Almer wrote: > > On 10/29/2017 11:57 AM, Michael Niedermayer wrote: > > > The choosen value is the lowest power of 2 that allows 1000 iterations of > > > fate-checkasm-float_dsp > > > to pass on x86-32 > > > > Ticket #6848 reports this value is still not enough. Maybe something > > like 1.0e-12 or 1.0e-13 instead? > > ok, ill push it with 1e-12 Or do people prefer this: (this should be more correct) commit 67ba87a320faba623c0b35a0692adb916860ac40 (HEAD -> master) Author: Michael Niedermayer Date: Sun Oct 29 15:26:50 2017 +0100 tests/checkasm/float_dsp: Increase allowed difference for float_dsp.vector_dmul Tested for 1 iterations on x86-32 Fixes: Ticket6848 Signed-off-by: Michael Niedermayer diff --git a/tests/checkasm/float_dsp.c b/tests/checkasm/float_dsp.c index 9b0a221c25..32cac45a17 100644 --- a/tests/checkasm/float_dsp.c +++ b/tests/checkasm/float_dsp.c @@ -165,7 +165,8 @@ static void test_vector_dmul_scalar(const double *src0, const double *src1) call_ref(cdst, src0, src1[0], LEN); call_new(odst, src0, src1[0], LEN); for (i = 0; i < LEN; i++) { -if (!double_near_abs_eps(cdst[i], odst[i], DBL_EPSILON)) { +double t = FFMAX3(fabs(src1[0]), fabs(src0[i]), fabs(src1[0] * src0[i])); +if (!double_near_abs_eps(cdst[i], odst[i], FFMAX(t, 1.0) * 2 * DBL_EPSILON)) { fprintf(stderr, "%d: %- .12f - %- .12f = % .12g\n", i, cdst[i], odst[i], cdst[i] - odst[i]); fail(); [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB "I am not trying to be anyone's saviour, I'm trying to think about the future and not be sad" - Elon Musk signature.asc Description: Digital signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] tests/checkasm/float_dsp: Increase allowed difference for float_dsp.vector_dmul
On Fri, Nov 24, 2017 at 11:37:36PM -0300, James Almer wrote: > On 10/29/2017 11:57 AM, Michael Niedermayer wrote: > > The choosen value is the lowest power of 2 that allows 1000 iterations of > > fate-checkasm-float_dsp > > to pass on x86-32 > > Ticket #6848 reports this value is still not enough. Maybe something > like 1.0e-12 or 1.0e-13 instead? ok, ill push it with 1e-12 thx [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB If you fake or manipulate statistics in a paper in physics you will never get a job again. If you fake or manipulate statistics in a paper in medicin you will get a job for life at the pharma industry. signature.asc Description: Digital signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] tests/checkasm/float_dsp: Increase allowed difference for float_dsp.vector_dmul
On 10/29/2017 11:57 AM, Michael Niedermayer wrote: > The choosen value is the lowest power of 2 that allows 1000 iterations of > fate-checkasm-float_dsp > to pass on x86-32 Ticket #6848 reports this value is still not enough. Maybe something like 1.0e-12 or 1.0e-13 instead? > > Signed-off-by: Michael Niedermayer > --- > tests/checkasm/float_dsp.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/tests/checkasm/float_dsp.c b/tests/checkasm/float_dsp.c > index 9b0a221c25..676c341f45 100644 > --- a/tests/checkasm/float_dsp.c > +++ b/tests/checkasm/float_dsp.c > @@ -165,7 +165,7 @@ static void test_vector_dmul_scalar(const double *src0, > const double *src1) > call_ref(cdst, src0, src1[0], LEN); > call_new(odst, src0, src1[0], LEN); > for (i = 0; i < LEN; i++) { > -if (!double_near_abs_eps(cdst[i], odst[i], DBL_EPSILON)) { > +if (!double_near_abs_eps(cdst[i], odst[i], 256*DBL_EPSILON)) { > fprintf(stderr, "%d: %- .12f - %- .12f = % .12g\n", i, > cdst[i], odst[i], cdst[i] - odst[i]); > fail(); > ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel