From: "Martin Storsjö" <martin@martin.st>
To: ffmpeg-devel@ffmpeg.org
Cc: Michael Niedermayer <michael@niedermayer.cc>,
Jonathan Swinney <jswinney@amazon.com>
Subject: Re: [FFmpeg-devel] [PATCH 2/2] RFC: checkasm: motion: Test different h parameters
Date: Thu, 4 Aug 2022 10:47:58 +0300 (EEST)
Message-ID: <b15afc21-0eb-c0e4-62ff-70d9ebda80@martin.st> (raw)
In-Reply-To: <20220713204716.3114529-2-martin@martin.st>
On Wed, 13 Jul 2022, Martin Storsjö wrote:
> Previously, the checkasm test always passed h=8, so no other cases
> were tested.
>
> Out of the me_cmp functions, in practice, some functions are hardcoded
> to always assume a 8x8 block (ignoring the h parameter), while others
> do use the parameter. For those with hardcoded height, both the
> reference C function and the assembly implementations ignore the
> parameter similarly.
>
> The documentation for the functions indicate that heights between
> w/2 and 2*w, within the range of 4 to 16, should be supported. This
> patch just tests random heights in that range, without knowing what
> width the current function actually uses.
> ---
> I'm not sure if it's good to have checkasm exercise cases that
> don't occur in practice or not. In particular, the aarch64
> functions have a separate implementation for non-multiple-of-4
> height, which probably doesn't ever get called in practice, while
> other SIMD implementations lack that.
>
> Alternatively, we'd improve the documentation for the expectations
> for these functions and make the test match that, and remove the
> unused non-multiple-of-4 case in the aarch64 assembly.
> ---
> tests/checkasm/motion.c | 10 ++++++----
> 1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/tests/checkasm/motion.c b/tests/checkasm/motion.c
> index 0112822174..79e4358941 100644
> --- a/tests/checkasm/motion.c
> +++ b/tests/checkasm/motion.c
> @@ -45,7 +45,7 @@ static void test_motion(const char *name, me_cmp_func test_func)
> /* motion estimation can look up to 17 bytes ahead */
> static const int look_ahead = 17;
>
> - int i, x, y, d1, d2;
> + int i, x, y, h, d1, d2;
> uint8_t *ptr;
>
> LOCAL_ALIGNED_16(uint8_t, img1, [WIDTH * HEIGHT]);
> @@ -68,14 +68,16 @@ static void test_motion(const char *name, me_cmp_func test_func)
> for (i = 0; i < ITERATIONS; i++) {
> x = rnd() % (WIDTH - look_ahead);
> y = rnd() % (HEIGHT - look_ahead);
> + // Pick a random h between 4 and 16; pick an even value.
> + h = 4 + ((rnd() % (16 + 1 - 4)) & ~1);
>
> ptr = img2 + y * WIDTH + x;
> - d2 = call_ref(NULL, img1, ptr, WIDTH, 8);
> - d1 = call_new(NULL, img1, ptr, WIDTH, 8);
> + d2 = call_ref(NULL, img1, ptr, WIDTH, h);
> + d1 = call_new(NULL, img1, ptr, WIDTH, h);
>
> if (d1 != d2) {
> fail();
> - printf("func: %s, x=%d y=%d, error: asm=%d c=%d\n", name, x, y, d1, d2);
> + printf("func: %s, x=%d y=%d h=%d, error: asm=%d c=%d\n", name, x, y, h, d1, d2);
> break;
> }
> }
> --
> 2.25.1
Ping
// Martin
_______________________________________________
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:[~2022-08-04 7:48 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-07-13 20:47 [FFmpeg-devel] [PATCH 1/2] x86: Don't hardcode the height to 8 in sad8_xy2_mmx Martin Storsjö
2022-07-13 20:47 ` [FFmpeg-devel] [PATCH 2/2] RFC: checkasm: motion: Test different h parameters Martin Storsjö
2022-08-04 7:47 ` Martin Storsjö [this message]
2022-08-16 11:16 ` Martin Storsjö
2022-08-04 7:47 ` [FFmpeg-devel] [PATCH 1/2] x86: Don't hardcode the height to 8 in sad8_xy2_mmx Martin Storsjö
2022-08-04 15:31 ` Michael Niedermayer
2022-08-04 20:29 ` Martin Storsjö
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=b15afc21-0eb-c0e4-62ff-70d9ebda80@martin.st \
--to=martin@martin.st \
--cc=ffmpeg-devel@ffmpeg.org \
--cc=jswinney@amazon.com \
--cc=michael@niedermayer.cc \
/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