Re: [FFmpeg-devel] [PATCH] avutil: runtime cpu detection for mips
On Wed, Sep 30, 2015 at 02:21:18PM +, Shivraj Patil wrote: > > On Wed, Sep 30, 2015 at 07:03:46PM +0530, shivraj.pa...@imgtec.com wrote: > > From: Shivraj Patil> [...] > > > +static int get_cpuinfo(uint32_t *hwcap) { > > +FILE *f = fopen("/proc/cpuinfo", "r"); > > under qemu i get this: > > cpu_flags(raw) = 0x > cpu_flags_str(raw) = > cpu_flags(effective) = 0x > cpu_flags_str(effective) = > threads = 1 (cpu_count = 12) > > IIUC this disables all cpu extensions > is that intended ? > (we currently only have loongson hardware to test so this would mean > an end to testing imgtec specific mips optimizations) > > Shivraj:- cpu detection will not work for qemu, however we have tested it on > mips hardware at our end. iam not speaking about just this patch but we have just qemu to test on imgtec - mips i can apply this patch but it means that in the future FFmpeg will then only be tested on loongson because you effectively disable our only way to test code on imgtec mips its very strange that you want this Shivraj:- I have got the concern with this patch and so request you to please discard it for now. It will make more sense to resubmit (modified) at appropriate time in near future when MSA enabled devices are setup for testing. Till the time requesting to keep the mips testing under QEMU as it is. [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB You can kill me, but you cannot change the truth. ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avutil: runtime cpu detection for mips
On Wed, Sep 30, 2015 at 02:21:18PM +, Shivraj Patil wrote: > > On Wed, Sep 30, 2015 at 07:03:46PM +0530, shivraj.pa...@imgtec.com wrote: > > From: Shivraj Patil> [...] > > > +static int get_cpuinfo(uint32_t *hwcap) { > > +FILE *f = fopen("/proc/cpuinfo", "r"); > > under qemu i get this: > > cpu_flags(raw) = 0x > cpu_flags_str(raw) = > cpu_flags(effective) = 0x > cpu_flags_str(effective) = > threads = 1 (cpu_count = 12) > > IIUC this disables all cpu extensions > is that intended ? > (we currently only have loongson hardware to test so this would mean an end > to testing imgtec specific mips optimizations) > > Shivraj:- cpu detection will not work for qemu, however we have tested it on > mips hardware at our end. iam not speaking about just this patch but we have just qemu to test on imgtec - mips i can apply this patch but it means that in the future FFmpeg will then only be tested on loongson because you effectively disable our only way to test code on imgtec mips its very strange that you want this [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB You can kill me, but you cannot change the truth. signature.asc Description: Digital signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avutil: runtime cpu detection for mips
On Wed, Sep 30, 2015 at 02:23:43PM +, Shivraj Patil wrote: > > imgtec.com> writes: > > > +FILE *f = fopen("/proc/cpuinfo", "r"); > > Is this what every other software for mips does? > How does the kernel (or whatever sets cpuinfo) know? > > Shivraj:- We have used generic cpuinfo as it is unrestricted to access from > user space compared to the kernel setting hwcaps (via low level mips control > regs accessible) in kernel space for now. > I might be asking something stupid, but afaik /proc is linux exclusive; does this mean it's going to disable all optimisations on every other system? [...] -- Clément B. signature.asc Description: PGP signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avutil: runtime cpu detection for mips
imgtec.com> writes: > +FILE *f = fopen("/proc/cpuinfo", "r"); Is this what every other software for mips does? How does the kernel (or whatever sets cpuinfo) know? Carl Eugen ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avutil: runtime cpu detection for mips
On Wed, Sep 30, 2015 at 07:03:46PM +0530, shivraj.pa...@imgtec.com wrote: > From: Shivraj Patil[...] > +static int get_cpuinfo(uint32_t *hwcap) { > +FILE *f = fopen("/proc/cpuinfo", "r"); under qemu i get this: cpu_flags(raw) = 0x cpu_flags_str(raw) = cpu_flags(effective) = 0x cpu_flags_str(effective) = threads = 1 (cpu_count = 12) IIUC this disables all cpu extensions is that intended ? (we currently only have loongson hardware to test so this would mean an end to testing imgtec specific mips optimizations) Shivraj:- cpu detection will not work for qemu, however we have tested it on mips hardware at our end. [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Avoid a single point of failure, be that a person or equipment. ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avutil: runtime cpu detection for mips
On Wed, Sep 30, 2015 at 07:03:46PM +0530, shivraj.pa...@imgtec.com wrote: > From: Shivraj Patil[...] > +static int get_cpuinfo(uint32_t *hwcap) > +{ > +FILE *f = fopen("/proc/cpuinfo", "r"); under qemu i get this: cpu_flags(raw) = 0x cpu_flags_str(raw) = cpu_flags(effective) = 0x cpu_flags_str(effective) = threads = 1 (cpu_count = 12) IIUC this disables all cpu extensions is that intended ? (we currently only have loongson hardware to test so this would mean an end to testing imgtec specific mips optimizations) [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Avoid a single point of failure, be that a person or equipment. signature.asc Description: Digital signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avutil: runtime cpu detection for mips
imgtec.com> writes: > +FILE *f = fopen("/proc/cpuinfo", "r"); Is this what every other software for mips does? How does the kernel (or whatever sets cpuinfo) know? Shivraj:- We have used generic cpuinfo as it is unrestricted to access from user space compared to the kernel setting hwcaps (via low level mips control regs accessible) in kernel space for now. Thanks, Carl Eugen ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel