RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

rename new lls code to lls2 to avoid conflict with the old which has a different ABI 

also remove failed attempt at a compatibility layer, the code simply cannot work

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2013-11-16 16:42:57 +01:00
parent bbe66ef912
commit c3814ab654
6 changed files with 28 additions and 54 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
libavcodec/lpc.c
View File

@ -208,7 +208,7 @@ int ff_lpc_calc_coefs(LPCContext *s,
}
if (lpc_type == FF_LPC_TYPE_CHOLESKY) {
LLSModel m[2];
LLSModel2 m[2];
LOCAL_ALIGNED(32, double, var, [FFALIGN(MAX_LPC_ORDER+1,4)]);
double av_uninit(weight);
memset(var, 0, FFALIGN(MAX_LPC_ORDER+1,4)*sizeof(*var));
@ -217,7 +217,7 @@ int ff_lpc_calc_coefs(LPCContext *s,
m[0].coeff[max_order-1][j] = -lpc[max_order-1][j];
for(; pass<lpc_passes; pass++){
avpriv_init_lls(&m[pass&1], max_order);
avpriv_init_lls2(&m[pass&1], max_order);
weight=0;
for(i=max_order; i<blocksize; i++){
@ -238,7 +238,7 @@ int ff_lpc_calc_coefs(LPCContext *s,
m[pass&1].update_lls(&m[pass&1], var);
}
avpriv_solve_lls(&m[pass&1], 0.001, 0);
avpriv_solve_lls2(&m[pass&1], 0.001, 0);
}
for(i=0; i<max_order; i++){

2
libavcodec/version.h
View File

@ -30,7 +30,7 @@
#define LIBAVCODEC_VERSION_MAJOR 55
#define LIBAVCODEC_VERSION_MINOR 43
#define LIBAVCODEC_VERSION_MICRO 100
#define LIBAVCODEC_VERSION_MICRO 101
#define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
LIBAVCODEC_VERSION_MINOR, \

35
libavutil/lls2.c
View File

@ -32,7 +32,7 @@
#include "version.h"
#include "lls2.h"
static void update_lls(LLSModel *m, double *var)
static void update_lls(LLSModel2 *m, double *var)
{
int i, j;
@ -43,7 +43,7 @@ static void update_lls(LLSModel *m, double *var)
}
}
void avpriv_solve_lls(LLSModel *m, double threshold, unsigned short min_order)
void avpriv_solve_lls2(LLSModel2 *m, double threshold, unsigned short min_order)
{
int i, j, k;
double (*factor)[MAX_VARS_ALIGN] = (void *) &m->covariance[1][0];
@ -100,7 +100,7 @@ void avpriv_solve_lls(LLSModel *m, double threshold, unsigned short min_order)
}
}
static double evaluate_lls(LLSModel *m, double *param, int order)
static double evaluate_lls(LLSModel2 *m, double *param, int order)
{
int i;
double out = 0;
@ -111,9 +111,9 @@ static double evaluate_lls(LLSModel *m, double *param, int order)
return out;
}
av_cold void avpriv_init_lls(LLSModel *m, int indep_count)
av_cold void avpriv_init_lls2(LLSModel2 *m, int indep_count)
{
memset(m, 0, sizeof(LLSModel));
memset(m, 0, sizeof(LLSModel2));
m->indep_count = indep_count;
m->update_lls = update_lls;
m->evaluate_lls = evaluate_lls;
@ -121,25 +121,6 @@ av_cold void avpriv_init_lls(LLSModel *m, int indep_count)
ff_init_lls_x86(m);
}
#if FF_API_LLS_PRIVATE
av_cold void av_init_lls(LLSModel *m, int indep_count)
{
avpriv_init_lls(m, indep_count);
}
void av_update_lls(LLSModel *m, double *param, double decay)
{
m->update_lls(m, param);
}
void av_solve_lls(LLSModel *m, double threshold, int min_order)
{
avpriv_solve_lls(m, threshold, min_order);
}
double av_evaluate_lls(LLSModel *m, double *param, int order)
{
return m->evaluate_lls(m, param, order);
}
#endif /* FF_API_LLS_PRIVATE */
#ifdef TEST
#include <stdio.h>
@ -148,12 +129,12 @@ double av_evaluate_lls(LLSModel *m, double *param, int order)
int main(void)
{
LLSModel m;
LLSModel2 m;
int i, order;
AVLFG lfg;
av_lfg_init(&lfg, 1);
avpriv_init_lls(&m, 3);
avpriv_init_lls2(&m, 3);
for (i = 0; i < 100; i++) {
LOCAL_ALIGNED(32, double, var, [4]);
@ -164,7 +145,7 @@ int main(void)
var[2] = var[1] + av_lfg_get(&lfg) / (double) UINT_MAX - 0.5;
var[3] = var[2] + av_lfg_get(&lfg) / (double) UINT_MAX - 0.5;
m.update_lls(&m, var);
avpriv_solve_lls(&m, 0.001, 0);
avpriv_solve_lls2(&m, 0.001, 0);
for (order = 0; order < 3; order++) {
eval = m.evaluate_lls(&m, var + 1, order);
printf("real:%9f order:%d pred:%9f var:%f coeffs:%f %9f %9f\n",

23
libavutil/lls2.h
View File

@ -30,12 +30,12 @@
#define MAX_VARS 32
#define MAX_VARS_ALIGN FFALIGN(MAX_VARS+1,4)
//FIXME avoid direct access to LLSModel from outside
//FIXME avoid direct access to LLSModel2 from outside
/**
* Linear least squares model.
*/
typedef struct LLSModel {
typedef struct LLSModel2 {
DECLARE_ALIGNED(32, double, covariance[MAX_VARS_ALIGN][MAX_VARS_ALIGN]);
DECLARE_ALIGNED(32, double, coeff[MAX_VARS][MAX_VARS]);
double variance[MAX_VARS];
@ -47,25 +47,18 @@ typedef struct LLSModel {
* 32-byte aligned, and any padding elements must be initialized
* (i.e not denormal/nan).
*/
void (*update_lls)(struct LLSModel *m, double *var);
void (*update_lls)(struct LLSModel2 *m, double *var);
/**
* Inner product of var[] and the LPC coefs.
* @param m this context
* @param var training samples, excluding the value to be predicted. unaligned.
* @param order lpc order
*/
double (*evaluate_lls)(struct LLSModel *m, double *var, int order);
} LLSModel;
double (*evaluate_lls)(struct LLSModel2 *m, double *var, int order);
} LLSModel2;
void avpriv_init_lls(LLSModel *m, int indep_count);
void ff_init_lls_x86(LLSModel *m);
void avpriv_solve_lls(LLSModel *m, double threshold, unsigned short min_order);
#if FF_API_LLS_PRIVATE
void av_init_lls(LLSModel *m, int indep_count);
void av_update_lls(LLSModel *m, double *param, double decay);
void av_solve_lls(LLSModel *m, double threshold, int min_order);
double av_evaluate_lls(LLSModel *m, double *param, int order);
#endif /* FF_API_LLS_PRIVATE */
void avpriv_init_lls2(LLSModel2 *m, int indep_count);
void ff_init_lls_x86(LLSModel2 *m);
void avpriv_solve_lls2(LLSModel2 *m, double threshold, unsigned short min_order);
#endif /* AVUTIL_LLS_H */

8
libavutil/x86/lls.asm
View File

@ -29,7 +29,7 @@ SECTION .text
%define COVAR_STRIDE MAX_VARS_ALIGN*8
%define COVAR(x,y) [covarq + (x)*8 + (y)*COVAR_STRIDE]
struc LLSModel
struc LLSModel2
.covariance: resq MAX_VARS_ALIGN*MAX_VARS_ALIGN
.coeff: resq MAX_VARS*MAX_VARS
.variance: resq MAX_VARS
@ -49,7 +49,7 @@ INIT_XMM sse2
%define movdqa movaps
cglobal update_lls, 2,5,8, ctx, var, i, j, covar2
%define covarq ctxq
mov id, [ctxq + LLSModel.indep_count]
mov id, [ctxq + LLSModel2.indep_count]
lea varq, [varq + iq*8]
neg iq
mov covar2q, covarq
@ -129,7 +129,7 @@ cglobal update_lls, 2,5,8, ctx, var, i, j, covar2
INIT_YMM avx
cglobal update_lls, 3,6,8, ctx, var, count, i, j, count2
%define covarq ctxq
mov countd, [ctxq + LLSModel.indep_count]
mov countd, [ctxq + LLSModel2.indep_count]
lea count2d, [countq-2]
xor id, id
.loopi:
@ -206,7 +206,7 @@ cglobal evaluate_lls, 3,4,2, ctx, var, order, i
%define coefsq ctxq
mov id, orderd
imul orderd, MAX_VARS
lea coefsq, [ctxq + LLSModel.coeff + orderq*8]
lea coefsq, [ctxq + LLSModel2.coeff + orderq*8]
movsd m0, [varq]
movhpd m0, [varq + 8]
mulpd m0, [coefsq]

8
libavutil/x86/lls_init.c
View File

@ -23,11 +23,11 @@
#include "libavutil/lls2.h"
#include "libavutil/x86/cpu.h"
void ff_update_lls_sse2(LLSModel *m, double *var);
void ff_update_lls_avx(LLSModel *m, double *var);
double ff_evaluate_lls_sse2(LLSModel *m, double *var, int order);
void ff_update_lls_sse2(LLSModel2 *m, double *var);
void ff_update_lls_avx(LLSModel2 *m, double *var);
double ff_evaluate_lls_sse2(LLSModel2 *m, double *var, int order);
av_cold void ff_init_lls_x86(LLSModel *m)
av_cold void ff_init_lls_x86(LLSModel2 *m)
{
int cpu_flags = av_get_cpu_flags();
if (EXTERNAL_SSE2(cpu_flags)) {