| Message ID | 20211115062221.1650-1-mindmark@gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | [FFmpeg-devel,v3,1/2] libavutil/common: clip nan value to amin | expand |
| Context | Check | Description |
|---|---|---|
| andriy/make_x86 | success | Make finished |
| andriy/make_fate_x86 | success | Make fate finished |
| andriy/make_ppc | success | Make finished |
| andriy/make_fate_ppc | success | Make fate finished |
Quoting mindmark@gmail.com (2021-11-15 07:22:20) > From: Mark Reid <mindmark@gmail.com> > > Changes av_clipf to return amin if a is nan. > Before if a is nan av_clipf_c returned nan and > av_clipf_sse would return amax. Now the both > should behave the same. > > This works because nan > amin is false. > The max(nan, amin) will be amin. This seems like it goes against the point of nan. I would expect any float operation involving nan to return nan.
On Tue, Nov 16, 2021 at 1:57 PM Anton Khirnov <anton@khirnov.net> wrote: > Quoting mindmark@gmail.com (2021-11-15 07:22:20) > > From: Mark Reid <mindmark@gmail.com> > > > > Changes av_clipf to return amin if a is nan. > > Before if a is nan av_clipf_c returned nan and > > av_clipf_sse would return amax. Now the both > > should behave the same. > > > > This works because nan > amin is false. > > The max(nan, amin) will be amin. > > This seems like it goes against the point of nan. I would expect any > float operation involving nan to return nan. > fmaxf,fminf does not return nan, unless both values are nan. So to me patch is fine. > > -- > Anton Khirnov > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > To unsubscribe, visit link above, or email > ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe". >
diff --git a/libavutil/common.h b/libavutil/common.h index 3cc1f07566..9338bda7d5 100644 --- a/libavutil/common.h +++ b/libavutil/common.h @@ -379,6 +379,8 @@ static av_always_inline int64_t av_sat_sub64_c(int64_t a, int64_t b) { /** * Clip a float value into the amin-amax range. + * If a is nan or -inf amin will be returned. + * If a is +inf amax will be returned. * @param a value to clip * @param amin minimum value of the clip range * @param amax maximum value of the clip range @@ -389,13 +391,13 @@ static av_always_inline av_const float av_clipf_c(float a, float amin, float ama #if defined(HAVE_AV_CONFIG_H) && defined(ASSERT_LEVEL) && ASSERT_LEVEL >= 2 if (amin > amax) abort(); #endif - if (a < amin) return amin; - else if (a > amax) return amax; - else return a; + return FFMIN(FFMAX(a, amin), amax); } /** * Clip a double value into the amin-amax range. + * If a is nan or -inf amin will be returned. + * If a is +inf amax will be returned. * @param a value to clip * @param amin minimum value of the clip range * @param amax maximum value of the clip range @@ -406,9 +408,7 @@ static av_always_inline av_const double av_clipd_c(double a, double amin, double #if defined(HAVE_AV_CONFIG_H) && defined(ASSERT_LEVEL) && ASSERT_LEVEL >= 2 if (amin > amax) abort(); #endif - if (a < amin) return amin; - else if (a > amax) return amax; - else return a; + return FFMIN(FFMAX(a, amin), amax); } /** Compute ceil(log2(x)). diff --git a/libavutil/x86/intmath.h b/libavutil/x86/intmath.h index 40743fd13e..1520c25ec9 100644 --- a/libavutil/x86/intmath.h +++ b/libavutil/x86/intmath.h @@ -110,8 +110,8 @@ static av_always_inline av_const double av_clipd_sse2(double a, double amin, dou #if defined(ASSERT_LEVEL) && ASSERT_LEVEL >= 2 if (amin > amax) abort(); #endif - __asm__ ("minsd %2, %0 \n\t" - "maxsd %1, %0 \n\t" + __asm__ ("maxsd %1, %0 \n\t" + "minsd %2, %0 \n\t" : "+&x"(a) : "xm"(amin), "xm"(amax)); return a; } @@ -126,8 +126,8 @@ static av_always_inline av_const float av_clipf_sse(float a, float amin, float a #if defined(ASSERT_LEVEL) && ASSERT_LEVEL >= 2 if (amin > amax) abort(); #endif - __asm__ ("minss %2, %0 \n\t" - "maxss %1, %0 \n\t" + __asm__ ("maxss %1, %0 \n\t" + "minss %2, %0 \n\t" : "+&x"(a) : "xm"(amin), "xm"(amax)); return a; }
From: Mark Reid <mindmark@gmail.com> Changes av_clipf to return amin if a is nan. Before if a is nan av_clipf_c returned nan and av_clipf_sse would return amax. Now the both should behave the same. This works because nan > amin is false. The max(nan, amin) will be amin. --- libavutil/common.h | 12 ++++++------ libavutil/x86/intmath.h | 8 ++++---- 2 files changed, 10 insertions(+), 10 deletions(-) -- 2.31.1.windows.1