On date Saturday 2024-01-13 04:51:06 +0100, Michael Niedermayer wrote:
> This is the 64bit version of Chris Doty-Humphreys SFC64
>
> Compared to the LCGs these produce much better quality numbers.
> Compared to LFGs this needs less state. (our LFG has 224 byte
> state for its 32bit version) this has 32byte state
> Also the initialization for our LFG is slower.
> This is also much faster than KISS or PCG.
>
> This commit replaces the broken LCG used before.
> (broken as it had only a period ~200M due to being put in a double)
>
> This changes the output from random() which is why libswresample.mak
> is updated, update was done using the command in libswresample.mak
>
> Signed-off-by: Michael Niedermayer <mich...@niedermayer.cc>
> ---
> libavutil/eval.c | 24 +++-
> libavutil/sfc64.h | 85 ++++++++++++++
> tests/fate/libswresample.mak | 208 +++++++++++++++++------------------
> tests/ref/fate/eval | 2 +-
> 4 files changed, 210 insertions(+), 109 deletions(-)
> create mode 100644 libavutil/sfc64.h
>
> diff --git a/libavutil/eval.c b/libavutil/eval.c
> index dc6b3697bc2..349015d4fa3 100644
> --- a/libavutil/eval.c
> +++ b/libavutil/eval.c
> @@ -35,6 +35,7 @@
> #include "internal.h"
> #include "log.h"
> #include "mathematics.h"
> +#include "sfc64.h"
> #include "time.h"
> #include "avstring.h"
> #include "timer.h"
> @@ -55,6 +56,7 @@ typedef struct Parser {
> void *log_ctx;
> #define VARS 10
> double *var;
> + FFSFC64 *prng_state;
> } Parser;
>
> static const AVClass eval_class = {
> @@ -173,6 +175,7 @@ struct AVExpr {
> } a;
> struct AVExpr *param[3];
> double *var;
> + FFSFC64 *prng_state;
> };
>
> static double etime(double v)
> @@ -231,8 +234,14 @@ static double eval_expr(Parser *p, AVExpr *e)
>
> #define COMPUTE_NEXT_RANDOM() \
> int idx = av_clip(eval_expr(p, e->param[0]), 0, VARS-1); \
> - uint64_t r = isnan(p->var[idx]) ? 0 : p->var[idx]; \
> - r = r * 1664525 + 1013904223; \
> + FFSFC64 *s = p->prng_state + idx; \
> + uint64_t r; \
> + \
> + if (!s->counter) { \
> + r = isnan(p->var[idx]) ? 0 : p->var[idx]; \
> + ff_sfc64_init(s, r, r, r, 12); \
> + } \
> + r = ff_sfc64_get(s); \
> p->var[idx] = r; \
>
> case e_random: {
> @@ -329,7 +338,11 @@ static double eval_expr(Parser *p, AVExpr *e)
> case e_div: return e->value * (d2 ? (d / d2) : d * INFINITY);
> case e_add: return e->value * (d + d2);
> case e_last:return e->value * d2;
> - case e_st : return e->value * (p->var[av_clip(d, 0,
> VARS-1)]= d2);
> + case e_st : {
> + int index = av_clip(d, 0, VARS-1);
> + p->prng_state[index].counter = 0;
> + return e->value * (p->var[index]= d2);
> + }
> case e_hypot:return e->value * hypot(d, d2);
> case e_atan2:return e->value * atan2(d, d2);
> case e_bitand: return isnan(d) || isnan(d2) ? NAN : e->value
> * ((long int)d & (long int)d2);
> @@ -349,6 +362,7 @@ void av_expr_free(AVExpr *e)
> av_expr_free(e->param[1]);
> av_expr_free(e->param[2]);
> av_freep(&e->var);
> + av_freep(&e->prng_state);
> av_freep(&e);
> }
>
> @@ -736,7 +750,8 @@ int av_expr_parse(AVExpr **expr, const char *s,
> goto end;
> }
> e->var= av_mallocz(sizeof(double) *VARS);
> - if (!e->var) {
> + e->prng_state = av_mallocz(sizeof(*e->prng_state) *VARS);
> + if (!e->var || !e->prng_state) {
> ret = AVERROR(ENOMEM);
> goto end;
> }
> @@ -778,6 +793,7 @@ double av_expr_eval(AVExpr *e, const double
> *const_values, void *opaque)
> {
> Parser p = { 0 };
> p.var= e->var;
> + p.prng_state= e->prng_state;
>
> p.const_values = const_values;
> p.opaque = opaque;
> diff --git a/libavutil/sfc64.h b/libavutil/sfc64.h
> new file mode 100644
> index 00000000000..05f1e84cc68
> --- /dev/null
> +++ b/libavutil/sfc64.h
> @@ -0,0 +1,85 @@
> +/*
> + * Copyright (c) 2024 Michael Niedermayer <michael-ffm...@niedermayer.cc>
> + *
> + * This file is part of FFmpeg.
> + *
> + * FFmpeg is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU Lesser General Public
> + * License as published by the Free Software Foundation; either
> + * version 2.1 of the License, or (at your option) any later version.
> + *
> + * FFmpeg is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + * Lesser General Public License for more details.
> + *
> + * You should have received a copy of the GNU Lesser General Public
> + * License along with FFmpeg; if not, write to the Free Software
> + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301
> USA
> + *
> + * This is a implementation of SFC64, a 64-bit PRNG by Chris Doty-Humphrey.
> + *
> + * This Generator is much faster (0m1.872s) than 64bit KISS (0m3.823s) and
> PCG-XSH-RR-64/32 (0m2.700s)
> + * And passes testu01 and practrand test suits.
> + *
nit: possibly better to put this in @file
> + */
> +
> +/**
> + * @file
> + * simple Pseudo Random Number Generator
> + *
> + */
> +
> +#ifndef AVUTIL_SFC64_H
> +#define AVUTIL_SFC64_H
> +
> +#include <inttypes.h>
> +
> +typedef struct FFSFC64 {
> + uint64_t a,b,c,counter;
> +} FFSFC64;
> +
> +static inline uint64_t ff_sfc64_get(FFSFC64 *s) {
> + uint64_t tmp = s->a + s->b + s->counter++;
> + s->a = s->b ^ (s->b >> 11);
> + s->b = s->c + (s->c << 3); // This is a multiply by 9
> + s->c = (s->c << 24 | s->c >> 40) + tmp;
> + return tmp;
> +}
> +
> +/**
> + * returns the previous random value, and steps the generator backward.
> + *
> + * Its safe to take values before the first, but such values can be highly
> + * correlated to the seeds.
Return ..., and step the generator...
It is safe to take values before the first, but such values can be highly
correlated to the seeds.
> + */
> +static inline uint64_t ff_sfc64_reverse_get(FFSFC64 *s) {
> + uint64_t prev_c = s->b * 0x8E38E38E38E38E39;
> + uint64_t tmp = s->c - (prev_c << 24 | prev_c >> 40);
> + s->b = s->a ^ (s->a >> 11);
> + s->b ^= s->b >> 22;
> + s->b ^= s->b >> 44;
> +
> + s->a = tmp - s->b - --s->counter;
> + s->c = prev_c;
> +
> + return tmp;
> +}
> +
> +/**
> + * Initialize sfc64 with up to 3 seeds.
> + *
> + * @param rounds number of rounds mixing up state during init. generally
> 8-18, larger numbers will help with bad quality seeds.
> + * 12 is a good choice if all 3 seeds are equal
uppercase Generally after the dot.
Looks good to me apart for the minor nits.
Thanks!!
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
