From: Nuo Mi <nuomi2021@gmail.com> To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org> Subject: Re: [FFmpeg-devel] [PATCH 3/3] checkasm: add tests for vvc dmvr Date: Fri, 26 Jul 2024 15:42:32 +0800 Message-ID: <CAFXK13cE7TN_MQ39_z2nydhv3UqPgdaNB31Zyxo=P4mnZMHi=w@mail.gmail.com> (raw) In-Reply-To: <751c9fed-4aa3-4f0c-8de4-f51c06451434@gmail.com> On Fri, Jul 26, 2024 at 9:36 AM James Almer <jamrial@gmail.com> wrote: > On 7/25/2024 10:35 AM, Nuo Mi wrote: > > dmvr_8_12x20_c: 186.2 > > dmvr_8_12x20_avx2: 25.7 > > dmvr_8_20x12_c: 181.7 > > dmvr_8_20x12_avx2: 25.2 > > dmvr_8_20x20_c: 283.2 > > dmvr_8_20x20_avx2: 32.0 > > dmvr_10_12x20_c: 90.0 > > dmvr_10_12x20_avx2: 15.7 > > dmvr_10_20x12_c: 41.0 > > dmvr_10_20x12_avx2: 14.7 > > dmvr_10_20x20_c: 81.5 > > dmvr_10_20x20_avx2: 26.7 > > dmvr_12_12x20_c: 190.7 > > dmvr_12_12x20_avx2: 20.2 > > dmvr_12_20x12_c: 187.2 > > dmvr_12_20x12_avx2: 20.2 > > dmvr_12_20x20_c: 292.7 > > dmvr_12_20x20_avx2: 27.2 > > dmvr_h_8_12x20_c: 317.0 > > dmvr_h_8_12x20_avx2: 37.0 > > dmvr_h_8_20x12_c: 340.0 > > dmvr_h_8_20x12_avx2: 41.0 > > dmvr_h_8_20x20_c: 540.7 > > dmvr_h_8_20x20_avx2: 64.0 > > dmvr_h_10_12x20_c: 322.7 > > dmvr_h_10_12x20_avx2: 30.7 > > dmvr_h_10_20x12_c: 344.2 > > dmvr_h_10_20x12_avx2: 34.0 > > dmvr_h_10_20x20_c: 529.0 > > dmvr_h_10_20x20_avx2: 51.5 > > dmvr_h_12_12x20_c: 326.7 > > dmvr_h_12_12x20_avx2: 33.5 > > dmvr_h_12_20x12_c: 331.7 > > dmvr_h_12_20x12_avx2: 51.2 > > dmvr_h_12_20x20_c: 534.0 > > dmvr_h_12_20x20_avx2: 62.7 > > dmvr_hv_8_12x20_c: 650.0 > > dmvr_hv_8_12x20_avx2: 57.2 > > dmvr_hv_8_20x12_c: 676.2 > > dmvr_hv_8_20x12_avx2: 70.0 > > dmvr_hv_8_20x20_c: 1068.5 > > dmvr_hv_8_20x20_avx2: 103.2 > > dmvr_hv_10_12x20_c: 649.0 > > dmvr_hv_10_12x20_avx2: 48.2 > > dmvr_hv_10_20x12_c: 677.7 > > dmvr_hv_10_20x12_avx2: 59.7 > > dmvr_hv_10_20x20_c: 1093.5 > > dmvr_hv_10_20x20_avx2: 91.7 > > dmvr_hv_12_12x20_c: 660.0 > > dmvr_hv_12_12x20_avx2: 58.7 > > dmvr_hv_12_20x12_c: 682.7 > > dmvr_hv_12_20x12_avx2: 72.0 > > dmvr_hv_12_20x20_c: 1094.0 > > dmvr_hv_12_20x20_avx2: 113.2 > > dmvr_v_8_12x20_c: 325.7 > > dmvr_v_8_12x20_avx2: 31.2 > > dmvr_v_8_20x12_c: 326.2 > > dmvr_v_8_20x12_avx2: 38.5 > > dmvr_v_8_20x20_c: 538.5 > > dmvr_v_8_20x20_avx2: 54.2 > > dmvr_v_10_12x20_c: 318.5 > > dmvr_v_10_12x20_avx2: 23.7 > > dmvr_v_10_20x12_c: 330.7 > > dmvr_v_10_20x12_avx2: 40.5 > > dmvr_v_10_20x20_c: 567.5 > > dmvr_v_10_20x20_avx2: 48.0 > > dmvr_v_12_12x20_c: 335.2 > > dmvr_v_12_12x20_avx2: 30.0 > > dmvr_v_12_20x12_c: 330.2 > > dmvr_v_12_20x12_avx2: 39.5 > > dmvr_v_12_20x20_c: 535.2 > > dmvr_v_12_20x20_avx2: 60.0 > > --- > > tests/checkasm/vvc_mc.c | 59 +++++++++++++++++++++++++++++++++++++++++ > > 1 file changed, 59 insertions(+) > > > > diff --git a/tests/checkasm/vvc_mc.c b/tests/checkasm/vvc_mc.c > > index bc6b580f42..62fa6aa7d0 100644 > > --- a/tests/checkasm/vvc_mc.c > > +++ b/tests/checkasm/vvc_mc.c > > @@ -324,6 +324,64 @@ static void check_avg(void) > > report("avg"); > > } > > > > +#define SR_RANGE 2 > > +static void check_dmvr(void) > > +{ > > + LOCAL_ALIGNED_32(uint16_t, dst0, [DST_BUF_SIZE]); > > + LOCAL_ALIGNED_32(uint16_t, dst1, [DST_BUF_SIZE]); > > + LOCAL_ALIGNED_32(uint8_t, src0, [SRC_BUF_SIZE]); > > + LOCAL_ALIGNED_32(uint8_t, src1, [SRC_BUF_SIZE]); > > + const int dst_stride = MAX_PB_SIZE * sizeof(int16_t); > > + > > + VVCDSPContext c; > > + declare_func(void, int16_t *dst, const uint8_t *src, ptrdiff_t > src_stride, int height, > > + intptr_t mx, intptr_t my, int width); > > + > > + for (int bit_depth = 8; bit_depth <= 12; bit_depth += 2) { > > + ff_vvc_dsp_init(&c, bit_depth); > > + randomize_pixels(src0, src1, SRC_BUF_SIZE); > > + for (int i = 0; i < 2; i++) { > > + for (int j = 0; j < 2; j++) { > > + for (int h = 8; h <= 16; h *= 2) { > > + for (int w = 8; w <= 16; w *= 2) { > > + const int pred_w = w + 2 * SR_RANGE; > > + const int pred_h = h + 2 * SR_RANGE; > > + const int mx = rnd() % > VVC_INTER_LUMA_DMVR_FACTS; > > + const int my = rnd() % > VVC_INTER_LUMA_DMVR_FACTS; > > + const char *type; > > + > > + if (w * h < 128) > > + continue; > > So h == 8 && w == 8 is not tested? > Hi James, thank you for the review. Yes, DMVR operates on subblocks with a maximum size of 16x16, and it also requires that the width multiplied by the height be at least 128. Therefore, only block sizes of 8x16, 16x8, and 16x16 are valid. see: 8.5.1 General decoding process for coding units coded in inter prediction mode and https://vicuesoft.com/blog/titles/DMVR_in_VVC/ > > > + > > + switch ((j << 1) | i) { > > + case 0: type = "dmvr"; break; // 0 0 > > + case 1: type = "dmvr_h"; break; // 0 1 > > + case 2: type = "dmvr_v"; break; // 1 0 > > + case 3: type = "dmvr_hv"; break; // 1 1 > > + } > > + > > + if (check_func(c.inter.dmvr[j][i], > "%s_%d_%dx%d", type, bit_depth, pred_w, pred_h)) { > > + memset(dst0, 0, DST_BUF_SIZE); > > + memset(dst1, 0, DST_BUF_SIZE); > > + call_ref(dst0, src0 + SRC_OFFSET, > PIXEL_STRIDE, pred_h, mx, my, pred_w); > > + call_new(dst1, src1 + SRC_OFFSET, > PIXEL_STRIDE, pred_h, mx, my, pred_w); > > + for (int k = 0; k < pred_h; k++) { > > + if (memcmp(dst0 + k * dst_stride, dst1 > + k * dst_stride, pred_w * sizeof(int16_t))) { > > + fail(); > > + break; > > + } > > + } > > + > > + bench_new(dst1, src1 + SRC_OFFSET, > PIXEL_STRIDE, pred_h, mx, my, pred_w); > > + } > > + } > > + } > > + } > > + } > > + } > > + report("dmvr"); > > +} > > + > > static void check_vvc_sad(void) > > { > > const int bit_depth = 10; > > @@ -363,6 +421,7 @@ static void check_vvc_sad(void) > > > > void checkasm_check_vvc_mc(void) > > { > > + check_dmvr(); > > check_vvc_sad(); > > check_put_vvc_luma(); > > check_put_vvc_luma_uni(); > _______________________________________________ > 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". > _______________________________________________ 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:[~2024-07-26 7:42 UTC|newest] Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top [not found] <20240725133546.19125-1-nuomi2021@gmail.com> 2024-07-25 13:35 ` [FFmpeg-devel] [PATCH 2/3] x86/vvcdec: add dmvr avx2 code Nuo Mi 2024-07-25 13:35 ` [FFmpeg-devel] [PATCH 3/3] checkasm: add tests for vvc dmvr Nuo Mi 2024-07-26 1:20 ` James Almer 2024-07-26 7:42 ` Nuo Mi [this message] 2024-08-11 14:00 ` Nuo Mi 2024-08-15 12:45 ` Nuo Mi
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='CAFXK13cE7TN_MQ39_z2nydhv3UqPgdaNB31Zyxo=P4mnZMHi=w@mail.gmail.com' \ --to=nuomi2021@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