On Thu, Oct 06, 2022 at 02:54:30AM +0200, Andreas Rheinhardt wrote: > We currently mostly do not empty the MMX state in our MMX > DSP functions; instead we only do so before code that might > be using x87 code. This is a violation of the System V i386 ABI > (and maybe of other ABIs, too): > "The CPU shall be in x87 mode upon entry to a function. Therefore, > every function that uses the MMX registers is required to issue an > emms or femms instruction after using MMX registers, before returning > or calling another function." (See 2.2.1 in [1]) > This patch does not intend to change all these functions to abide > by the ABI; it only does so for ff_simple_idct_mmx(), as this > function can by called by external users, because it is exported > via AVDCT (i.e. via avcodec_dct_init()). Without this, the following > fragment will assert (in i386): > av_force_cpu_flags(AV_CPU_FLAG_MMX | AV_CPU_FLAG_MMXEXT); > int16_t *blk = av_malloc(64 * sizeof(*blk)); > AVDCT *avdct = avcodec_dct_alloc(); > avcodec_dct_init(avdct); > avdct->idct(blk); > av_assert0_fpu(); > > [1]: https://raw.githubusercontent.com/wiki/hjl-tools/x86-psABI/intel386-psABI-1.1.pdf > > Signed-off-by: Andreas Rheinhardt > --- > libavcodec/x86/simple_idct.asm | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/libavcodec/x86/simple_idct.asm b/libavcodec/x86/simple_idct.asm > index dcf0da6df1..e3a29efc33 100644 > --- a/libavcodec/x86/simple_idct.asm > +++ b/libavcodec/x86/simple_idct.asm > @@ -845,6 +845,7 @@ INIT_MMX mmx > > cglobal simple_idct, 1, 2, 8, 128, block, t0 > IDCT > + emms > RET as comments where requested What speed impact does this have on old CPUs ? can this be made compile time optional so someone building for an old machiene can avoid the speedhit while ignoring the ABI (which worked fine) (this case would then require a 2nd function with the emms for AVDCT) thx [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB I am the wisest man alive, for I know one thing, and that is that I know nothing. -- Socrates