From: Andreas Rheinhardt <andreas.rheinhardt@outlook.com> To: ffmpeg-devel@ffmpeg.org Subject: Re: [FFmpeg-devel] [PATCHv5 1/4] lavc/h263dsp: add DCT dequantisation functions Date: Thu, 13 Jun 2024 04:23:12 +0200 Message-ID: <AS8P250MB0744E473342FF2BD5623A3B18FC12@AS8P250MB0744.EURP250.PROD.OUTLOOK.COM> (raw) In-Reply-To: <20240612044723.175502-1-remi@remlab.net> Rémi Denis-Courmont: > Note that optimised implementations of these functions will be taken > into actual use only if MpegEncContext.dct_unquantize_h263_{inter,intra} > are *not* overloaded by existing optimisations. > > --- > This adds the plus ones back, saving two branch instructions in C and > one in assembler (at the cost of two unconditional adds). > I don't see how this saves branch instructions: len - 1 is allowed to be 0 in h263_dct_unquantize_intra and therefore the initial check can't be avoided (presuming the compiler produces an initial check+do-loop). It seems to me that your intra assembly simply ignores this. If you were to check for the case of nCoeffs == 0 in dct_unquantize_h263_intra_c (before the call!), you could write assembly with this restriction in mind. It would also allow the compiler to avoid the branch in case it is known that nCoeffs == 63. > --- > libavcodec/h263dsp.c | 26 ++++++++++++++++++++++++++ > libavcodec/h263dsp.h | 4 ++++ > 2 files changed, 30 insertions(+) > > diff --git a/libavcodec/h263dsp.c b/libavcodec/h263dsp.c > index 6a13353499..f4523a68c1 100644 > --- a/libavcodec/h263dsp.c > +++ b/libavcodec/h263dsp.c > @@ -19,10 +19,34 @@ > #include <stdint.h> > > #include "libavutil/attributes.h" > +#include "libavutil/avassert.h" > #include "libavutil/common.h" > #include "config.h" > #include "h263dsp.h" > > +static void h263_dct_unquantize_inter_c(int16_t *block, size_t len, > + int qmul, int qadd) > +{ > + for (size_t i = 0; i < len; i++) { > + int level = block[i]; > + > + if (level) { > + if (level < 0) > + level = level * qmul - qadd; > + else > + level = level * qmul + qadd; > + block[i] = level; > + } > + } > +} > + > +static void h263_dct_unquantize_intra_c(int16_t *block, size_t len, > + int qmul, int qadd) > +{ > + av_assert1(len >= 1); > + h263_dct_unquantize_inter_c(block + 1, len - 1, qmul, qadd); > +} > + > const uint8_t ff_h263_loop_filter_strength[32] = { > 0, 1, 1, 2, 2, 3, 3, 4, 4, 4, 5, 5, 6, 6, 7, 7, > 7, 8, 8, 8, 9, 9, 9, 10, 10, 10, 11, 11, 11, 12, 12, 12 > @@ -116,6 +140,8 @@ static void h263_v_loop_filter_c(uint8_t *src, int stride, int qscale) > > av_cold void ff_h263dsp_init(H263DSPContext *ctx) > { > + ctx->h263_dct_unquantize_intra = h263_dct_unquantize_intra_c; > + ctx->h263_dct_unquantize_inter = h263_dct_unquantize_inter_c; > ctx->h263_h_loop_filter = h263_h_loop_filter_c; > ctx->h263_v_loop_filter = h263_v_loop_filter_c; > > diff --git a/libavcodec/h263dsp.h b/libavcodec/h263dsp.h > index 2dccd23392..0ecbe83314 100644 > --- a/libavcodec/h263dsp.h > +++ b/libavcodec/h263dsp.h > @@ -24,6 +24,10 @@ > extern const uint8_t ff_h263_loop_filter_strength[32]; > > typedef struct H263DSPContext { > + void (*h263_dct_unquantize_intra)(int16_t *block /* align 16 */, > + size_t len, int mul, int add); > + void (*h263_dct_unquantize_inter)(int16_t *block /* align 16 */, > + size_t len, int mul, int add); > void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale); > void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale); > } H263DSPContext; _______________________________________________ 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:[~2024-06-13 2:23 UTC|newest] Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top 2024-06-12 4:47 Rémi Denis-Courmont 2024-06-12 4:47 ` [FFmpeg-devel] [PATCH 2/4] lavc/mpegvideo: use H263DSP dequant function Rémi Denis-Courmont 2024-06-14 14:33 ` James Almer 2024-06-14 14:45 ` Rémi Denis-Courmont 2024-06-14 15:11 ` Rémi Denis-Courmont 2024-06-14 15:41 ` James Almer 2024-06-14 19:03 ` Rémi Denis-Courmont 2024-06-12 4:47 ` [FFmpeg-devel] [PATCH 3/4] checkasm/h263dsp: test dct_unquantize_{intra, inter} Rémi Denis-Courmont 2024-06-12 18:39 ` James Almer 2024-06-12 19:15 ` Rémi Denis-Courmont 2024-06-12 19:39 ` James Almer 2024-06-12 19:52 ` Rémi Denis-Courmont 2024-06-13 19:41 ` [FFmpeg-devel] [PATCHv6 " Rémi Denis-Courmont 2024-06-12 4:47 ` [FFmpeg-devel] [PATCH 4/4] lavc/h263dsp: R-V V " Rémi Denis-Courmont 2024-06-13 19:41 ` [FFmpeg-devel] [PATCHv6 " Rémi Denis-Courmont 2024-06-12 17:40 ` [FFmpeg-devel] [PATCHv5 1/4] lavc/h263dsp: add DCT dequantisation functions James Almer 2024-06-12 18:10 ` Rémi Denis-Courmont 2024-06-13 2:23 ` Andreas Rheinhardt [this message] 2024-06-13 19:22 ` Rémi Denis-Courmont
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=AS8P250MB0744E473342FF2BD5623A3B18FC12@AS8P250MB0744.EURP250.PROD.OUTLOOK.COM \ --to=andreas.rheinhardt@outlook.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