On Wed, Mar 04, 2020 at 04:13:43AM +0000, Fu, Linjie wrote: > > -----Original Message----- > > From: ffmpeg-devel <ffmpeg-devel-boun...@ffmpeg.org> On Behalf Of > > Michael Niedermayer > > Sent: Wednesday, March 4, 2020 02:41 > > To: FFmpeg development discussions and patches <ffmpeg- > > de...@ffmpeg.org> > > Subject: Re: [FFmpeg-devel] [PATCH 1/4] tests/checkasm: add overflow test > > for hevc_add_res > > > > On Wed, Dec 11, 2019 at 04:45:36PM +0800, Linjie Fu wrote: > > > Add overflow test for hevc_add_res when int16_t coeff = -32768. > > > The result of C is good, while ASM is not. > > > > > > To verify: > > > make fate-checkasm-hevc_add_res > > > ffmpeg/tests/checkasm/checkasm --test=hevc_add_res > > > > > > ./checkasm --test=hevc_add_res > > > checkasm: using random seed 679391863 > > > MMXEXT: > > > hevc_add_res_4x4_8_mmxext (hevc_add_res.c:69) > > > - hevc_add_res.add_residual [FAILED] > > > SSE2: > > > hevc_add_res_8x8_8_sse2 (hevc_add_res.c:69) > > > hevc_add_res_16x16_8_sse2 (hevc_add_res.c:69) > > > hevc_add_res_32x32_8_sse2 (hevc_add_res.c:69) > > > - hevc_add_res.add_residual [FAILED] > > > AVX: > > > hevc_add_res_8x8_8_avx (hevc_add_res.c:69) > > > hevc_add_res_16x16_8_avx (hevc_add_res.c:69) > > > hevc_add_res_32x32_8_avx (hevc_add_res.c:69) > > > - hevc_add_res.add_residual [FAILED] > > > AVX2: > > > hevc_add_res_32x32_8_avx2 (hevc_add_res.c:69) > > > - hevc_add_res.add_residual [FAILED] > > > checkasm: 8 of 14 tests have failed > > > > > > Signed-off-by: Xu Guangxin <guangxin...@intel.com> > > > Signed-off-by: Linjie Fu <linjie...@intel.com> > > > --- > > > tests/checkasm/hevc_add_res.c | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > diff --git a/tests/checkasm/hevc_add_res.c > > b/tests/checkasm/hevc_add_res.c > > > index e92c6b4..a6e3b8a 100644 > > > --- a/tests/checkasm/hevc_add_res.c > > > +++ b/tests/checkasm/hevc_add_res.c > > > @@ -58,6 +58,7 @@ static void check_add_res(HEVCDSPContext h, int > > bit_depth) > > > > > > randomize_buffers(res0, size); > > > randomize_buffers2(dst0, size); > > > + res0[0] = 0x8000;// overflow test > > > memcpy(res1, res0, sizeof(*res0) * size); > > > memcpy(dst1, dst0, sizeof(int16_t) * size); > > > > unconditionally and always seting this would reduce the space that can > > be tested unless iam missing something > > Thanks and Yes it could be a problem especially for 4x4 tests since it only > has 16 residual tested in total (if we set 1 constant test for overflow) > > To cover the overflow situation without reducing the testing space, one > simple way is to add one more check for checkasm_hevc_add_res to test > 2x cases (no too much cost introduced): > > @@ -80,6 +81,7 @@ void checkasm_check_hevc_add_res(void) > > ff_hevc_dsp_init(&h, bit_depth); > check_add_res(h, bit_depth); > + check_add_res(h, bit_depth); > > If it's suitable, I'll update and resend the patch set. > (also there may be another choice to only apply the fix in patch 2 ~ 4) > > Looking forward to your further comments.
probably ok thx [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB The real ebay dictionary, page 1 "Used only once" - "Some unspecified defect prevented a second use" "In good condition" - "Can be repaird by experienced expert" "As is" - "You wouldnt want it even if you were payed for it, if you knew ..."
signature.asc
Description: PGP signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
