From: James Almer <jamrial@gmail.com>
To: ffmpeg-devel@ffmpeg.org
Subject: Re: [FFmpeg-devel] [PATCH 1/2] libavutil/cpu: Add AV_CPU_FLAG_SLOW_GATHER.
Date: Mon, 20 Dec 2021 11:21:28 -0300
Message-ID: <8424d6a1-df63-954e-6823-740bf1fcb891@gmail.com> (raw)
In-Reply-To: <20211220135627.615097-1-alankelly@google.com>
On 12/20/2021 10:56 AM, Alan Kelly wrote:
> This flag is set on Haswell and earlier and all AMD cpus.
> ---
> As discussed on IRC last week.
> libavutil/cpu.h | 57 +++++++++++++++++++++++----------------------
> libavutil/x86/cpu.c | 13 ++++++++++-
> 2 files changed, 41 insertions(+), 29 deletions(-)
>
> diff --git a/libavutil/cpu.h b/libavutil/cpu.h
> index ae443eccad..4272d11d73 100644
> --- a/libavutil/cpu.h
> +++ b/libavutil/cpu.h
> @@ -26,34 +26,35 @@
> #define AV_CPU_FLAG_FORCE 0x80000000 /* force usage of selected flags (OR) */
>
> /* lower 16 bits - CPU features */
> -#define AV_CPU_FLAG_MMX 0x0001 ///< standard MMX
> -#define AV_CPU_FLAG_MMXEXT 0x0002 ///< SSE integer functions or AMD MMX ext
> -#define AV_CPU_FLAG_MMX2 0x0002 ///< SSE integer functions or AMD MMX ext
> -#define AV_CPU_FLAG_3DNOW 0x0004 ///< AMD 3DNOW
> -#define AV_CPU_FLAG_SSE 0x0008 ///< SSE functions
> -#define AV_CPU_FLAG_SSE2 0x0010 ///< PIV SSE2 functions
> -#define AV_CPU_FLAG_SSE2SLOW 0x40000000 ///< SSE2 supported, but usually not faster
> - ///< than regular MMX/SSE (e.g. Core1)
> -#define AV_CPU_FLAG_3DNOWEXT 0x0020 ///< AMD 3DNowExt
> -#define AV_CPU_FLAG_SSE3 0x0040 ///< Prescott SSE3 functions
> -#define AV_CPU_FLAG_SSE3SLOW 0x20000000 ///< SSE3 supported, but usually not faster
> - ///< than regular MMX/SSE (e.g. Core1)
> -#define AV_CPU_FLAG_SSSE3 0x0080 ///< Conroe SSSE3 functions
> -#define AV_CPU_FLAG_SSSE3SLOW 0x4000000 ///< SSSE3 supported, but usually not faster
> -#define AV_CPU_FLAG_ATOM 0x10000000 ///< Atom processor, some SSSE3 instructions are slower
> -#define AV_CPU_FLAG_SSE4 0x0100 ///< Penryn SSE4.1 functions
> -#define AV_CPU_FLAG_SSE42 0x0200 ///< Nehalem SSE4.2 functions
> -#define AV_CPU_FLAG_AESNI 0x80000 ///< Advanced Encryption Standard functions
> -#define AV_CPU_FLAG_AVX 0x4000 ///< AVX functions: requires OS support even if YMM registers aren't used
> -#define AV_CPU_FLAG_AVXSLOW 0x8000000 ///< AVX supported, but slow when using YMM registers (e.g. Bulldozer)
> -#define AV_CPU_FLAG_XOP 0x0400 ///< Bulldozer XOP functions
> -#define AV_CPU_FLAG_FMA4 0x0800 ///< Bulldozer FMA4 functions
> -#define AV_CPU_FLAG_CMOV 0x1000 ///< supports cmov instruction
> -#define AV_CPU_FLAG_AVX2 0x8000 ///< AVX2 functions: requires OS support even if YMM registers aren't used
> -#define AV_CPU_FLAG_FMA3 0x10000 ///< Haswell FMA3 functions
> -#define AV_CPU_FLAG_BMI1 0x20000 ///< Bit Manipulation Instruction Set 1
> -#define AV_CPU_FLAG_BMI2 0x40000 ///< Bit Manipulation Instruction Set 2
> -#define AV_CPU_FLAG_AVX512 0x100000 ///< AVX-512 functions: requires OS support even if YMM/ZMM registers aren't used
> +#define AV_CPU_FLAG_MMX 0x0001 ///< standard MMX
> +#define AV_CPU_FLAG_MMXEXT 0x0002 ///< SSE integer functions or AMD MMX ext
> +#define AV_CPU_FLAG_MMX2 0x0002 ///< SSE integer functions or AMD MMX ext
> +#define AV_CPU_FLAG_3DNOW 0x0004 ///< AMD 3DNOW
> +#define AV_CPU_FLAG_SSE 0x0008 ///< SSE functions
> +#define AV_CPU_FLAG_SSE2 0x0010 ///< PIV SSE2 functions
> +#define AV_CPU_FLAG_SSE2SLOW 0x40000000 ///< SSE2 supported, but usually not faster
> + ///< than regular MMX/SSE (e.g. Core1)
> +#define AV_CPU_FLAG_3DNOWEXT 0x0020 ///< AMD 3DNowExt
> +#define AV_CPU_FLAG_SSE3 0x0040 ///< Prescott SSE3 functions
> +#define AV_CPU_FLAG_SSE3SLOW 0x20000000 ///< SSE3 supported, but usually not faster
> + ///< than regular MMX/SSE (e.g. Core1)
> +#define AV_CPU_FLAG_SSSE3 0x0080 ///< Conroe SSSE3 functions
> +#define AV_CPU_FLAG_SSSE3SLOW 0x4000000 ///< SSSE3 supported, but usually not faster
> +#define AV_CPU_FLAG_ATOM 0x10000000 ///< Atom processor, some SSSE3 instructions are slower
> +#define AV_CPU_FLAG_SSE4 0x0100 ///< Penryn SSE4.1 functions
> +#define AV_CPU_FLAG_SSE42 0x0200 ///< Nehalem SSE4.2 functions
> +#define AV_CPU_FLAG_AESNI 0x80000 ///< Advanced Encryption Standard functions
> +#define AV_CPU_FLAG_AVX 0x4000 ///< AVX functions: requires OS support even if YMM registers aren't used
> +#define AV_CPU_FLAG_AVXSLOW 0x8000000 ///< AVX supported, but slow when using YMM registers (e.g. Bulldozer)
> +#define AV_CPU_FLAG_XOP 0x0400 ///< Bulldozer XOP functions
> +#define AV_CPU_FLAG_FMA4 0x0800 ///< Bulldozer FMA4 functions
> +#define AV_CPU_FLAG_CMOV 0x1000 ///< supports cmov instruction
> +#define AV_CPU_FLAG_AVX2 0x8000 ///< AVX2 functions: requires OS support even if YMM registers aren't used
> +#define AV_CPU_FLAG_FMA3 0x10000 ///< Haswell FMA3 functions
> +#define AV_CPU_FLAG_BMI1 0x20000 ///< Bit Manipulation Instruction Set 1
> +#define AV_CPU_FLAG_BMI2 0x40000 ///< Bit Manipulation Instruction Set 2
> +#define AV_CPU_FLAG_AVX512 0x100000 ///< AVX-512 functions: requires OS support even if YMM/ZMM registers aren't used
> +#define AV_CPU_FLAG_SLOW_GATHER 0x2000000 ///< CPU has slow gathers.
Don't re-indent the other flags. It will affect git blame output.
>
> #define AV_CPU_FLAG_ALTIVEC 0x0001 ///< standard
> #define AV_CPU_FLAG_VSX 0x0002 ///< ISA 2.06
> diff --git a/libavutil/x86/cpu.c b/libavutil/x86/cpu.c
> index bcd41a50a2..5770ecec72 100644
> --- a/libavutil/x86/cpu.c
> +++ b/libavutil/x86/cpu.c
> @@ -146,8 +146,16 @@ int ff_get_cpu_flags_x86(void)
> if (max_std_level >= 7) {
> cpuid(7, eax, ebx, ecx, edx);
> #if HAVE_AVX2
> - if ((rval & AV_CPU_FLAG_AVX) && (ebx & 0x00000020))
> + if ((rval & AV_CPU_FLAG_AVX) && (ebx & 0x00000020)) {
> rval |= AV_CPU_FLAG_AVX2;
> + cpuid(1, eax, ebx, ecx, std_caps);
> + family = ((eax >> 8) & 0xf) + ((eax >> 20) & 0xff);
> + model = ((eax >> 4) & 0xf) + ((eax >> 12) & 0xf0);
> + /* Haswell and earlier has slow gather */
Afaik Haswell is the first CPU with gathers, so this comment is odd. Is
there another CPU this check below will trigger for?
> + if(family == 6 && model < 70)
> + rval |= AV_CPU_FLAG_SLOW_GATHER;
> + }
> +
> #if HAVE_AVX512 /* F, CD, BW, DQ, VL */
> if ((xcr0_lo & 0xe0) == 0xe0) { /* OPMASK/ZMM state */
> if ((rval & AV_CPU_FLAG_AVX2) && (ebx & 0xd0030000) == 0xd0030000)
> @@ -196,6 +204,9 @@ int ff_get_cpu_flags_x86(void)
> used unless explicitly disabled by checking AV_CPU_FLAG_AVXSLOW. */
> if ((family == 0x15 || family == 0x16) && (rval & AV_CPU_FLAG_AVX))
> rval |= AV_CPU_FLAG_AVXSLOW;
> +
> + /* AMD cpus have slow gather */
> + rval |= AV_CPU_FLAG_SLOW_GATHER;
Don't unconditionally enable this for every CPU. Do it only for those
with AVX2.
if (rval & AV_CPU_FLAG_AVX2)
rval |= AV_CPU_FLAG_SLOW_GATHER;
> }
>
> /* XOP and FMA4 use the AVX instruction coding scheme, so they can't be
_______________________________________________
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".
next prev parent reply other threads:[~2021-12-20 14:21 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-12-20 13:56 Alan Kelly
2021-12-20 14:21 ` James Almer [this message]
2021-12-20 14:43 ` Alan Kelly
2021-12-20 14:47 ` Lynne
2021-12-20 14:53 ` James Almer
2021-12-20 14:59 ` Alan Kelly
2021-12-20 15:26 ` Lynne
2021-12-21 19:56 ` Alan Kelly
2021-12-21 20:52 ` Lynne
2021-12-20 15:01 ` Alan Kelly
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=8424d6a1-df63-954e-6823-740bf1fcb891@gmail.com \
--to=jamrial@gmail.com \
--cc=ffmpeg-devel@ffmpeg.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git