From: James Almer <jamrial@gmail.com>
To: ffmpeg-devel@ffmpeg.org
Subject: Re: [FFmpeg-devel] [PATCH 2/3] checkasm/blockdsp: use smallest allowed aligned buffers for fill_block_tab tests
Date: Tue, 7 May 2024 12:26:53 -0300
Message-ID: <4c270bfb-7f64-4e25-8914-38350ea94fbe@gmail.com> (raw)
In-Reply-To: <GV1P250MB0737C8E01E27C2A2075930578FE42@GV1P250MB0737.EURP250.PROD.OUTLOOK.COM>
On 5/7/2024 12:14 PM, Andreas Rheinhardt wrote:
> James Almer:
>> The requirement is either 8 or 16 bytes alignment, not 32.
>> This should help finding bugs in asm implementations.
>>
>> Signed-off-by: James Almer <jamrial@gmail.com>
>> ---
>> tests/checkasm/blockdsp.c | 23 +++++++++--------------
>> 1 file changed, 9 insertions(+), 14 deletions(-)
>>
>> diff --git a/tests/checkasm/blockdsp.c b/tests/checkasm/blockdsp.c
>> index ab87fc8fa4..f67a38d302 100644
>> --- a/tests/checkasm/blockdsp.c
>> +++ b/tests/checkasm/blockdsp.c
>> @@ -29,11 +29,6 @@
>> #include "libavutil/intreadwrite.h"
>> #include "libavutil/mem_internal.h"
>>
>> -typedef struct {
>> - const char *name;
>> - int size;
>> -} test;
>> -
>> #define randomize_buffers(size) \
>> do { \
>> int i; \
>> @@ -58,18 +53,18 @@ do { \
>> } while (0)
>>
>> static void check_fill(BlockDSPContext *h){
>> - const test tests[] = {
>> - {"fill_block_tab[0]", 16},
>> - {"fill_block_tab[1]", 8},
>> - };
>> - LOCAL_ALIGNED_32(uint8_t, buf0, [16 * 16]);
>> - LOCAL_ALIGNED_32(uint8_t, buf1, [16 * 16]);
>> + LOCAL_ALIGNED_16(uint8_t, buf0_16, [16 * 16]);
>> + LOCAL_ALIGNED_16(uint8_t, buf1_16, [16 * 16]);
>> + LOCAL_ALIGNED_8(uint8_t, buf0_8, [8 * 8]);
>> + LOCAL_ALIGNED_8(uint8_t, buf1_8, [8 * 8]);
>>
>> - for (size_t t = 0; t < FF_ARRAY_ELEMS(tests); ++t) {
>> - int n = tests[t].size;
>> + for (int t = 0; t < 2; ++t) {
>> + uint8_t *buf0 = t ? buf0_8 : buf0_16;
>> + uint8_t *buf1 = t ? buf1_8 : buf1_16;
>> + int n = 16 - 8 * t;
>> declare_func(void, uint8_t *block, uint8_t value,
>> ptrdiff_t line_size, int h);
>> - if (check_func(h->fill_block_tab[t], "blockdsp.%s", tests[t].name)) {
>> + if (check_func(h->fill_block_tab[t], "blockdsp.fill_block_tab[%d]", t)) {
>> uint8_t value = rnd();
>> memset(buf0, 0, sizeof(*buf0) * n * n);
>> memset(buf1, 0, sizeof(*buf0) * n * n);
>
> 1. I wouldn't be surprised if the *_8 buffers were still 16 byte
> aligned. You should probably force 8 byte alignment by using a 16
> byte-aligned buffer with an offset of eight.
Amended the following locally:
> diff --git a/tests/checkasm/blockdsp.c b/tests/checkasm/blockdsp.c
> index 8c1f8281d2..5f4d46b8fa 100644
> --- a/tests/checkasm/blockdsp.c
> +++ b/tests/checkasm/blockdsp.c
> @@ -55,12 +55,10 @@ do { \
> static void check_fill(BlockDSPContext *h){
> LOCAL_ALIGNED_16(uint8_t, buf0_16, [16 * 16]);
> LOCAL_ALIGNED_16(uint8_t, buf1_16, [16 * 16]);
> - LOCAL_ALIGNED_8(uint8_t, buf0_8, [8 * 8]);
> - LOCAL_ALIGNED_8(uint8_t, buf1_8, [8 * 8]);
>
> for (int t = 0; t < 2; ++t) {
> - uint8_t *buf0 = t ? buf0_8 : buf0_16;
> - uint8_t *buf1 = t ? buf1_8 : buf1_16;
> + uint8_t *buf0 = buf0_16 + t * /* force 8 byte alignment */ 8;
> + uint8_t *buf1 = buf1_16 + t * /* force 8 byte alignment */ 8;
> int n = 16 - 8 * t;
> declare_func(void, uint8_t *block, uint8_t value,
> ptrdiff_t line_size, int h);
> 2. Can you also extend this test to actually test the case of stride !=
> width? (And negative strides.)
Maybe later.
_______________________________________________
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".
prev parent reply other threads:[~2024-05-07 15:27 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-07 0:27 [FFmpeg-devel] [PATCH] checkasm/blockdsp: don't randomize the buffers for fill_block_tab James Almer
2024-05-07 5:44 ` Martin Storsjö
2024-05-07 10:49 ` Andreas Rheinhardt
2024-05-07 10:52 ` Martin Storsjö
2024-05-07 11:11 ` James Almer
2024-05-07 15:02 ` [FFmpeg-devel] [PATCH 2/3] checkasm/blockdsp: use smallest allowed aligned buffers for fill_block_tab tests James Almer
2024-05-07 15:02 ` [FFmpeg-devel] [PATCH 3/3] x86/blockdsp: add sse2 and avx2 versions of fill_block_tab James Almer
2024-05-07 15:10 ` Andreas Rheinhardt
2024-05-07 15:15 ` James Almer
2024-05-07 15:14 ` [FFmpeg-devel] [PATCH 2/3] checkasm/blockdsp: use smallest allowed aligned buffers for fill_block_tab tests Andreas Rheinhardt
2024-05-07 15:26 ` James Almer [this message]
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=4c270bfb-7f64-4e25-8914-38350ea94fbe@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