From: Michael Niedermayer <michaelni@gmx.at>
Date: Sun, 30 Jun 2013 09:30:07 +0000 (+0200)
Subject: Merge commit '502ab21af0ca68f76d6112722c46d2f35c004053'
X-Git-Tag: android-x86-4.4-r1~2655
X-Git-Url: http://git.osdn.net/view?a=commitdiff_plain;h=78b547963362cc9827d92dc0808534b757ab05d4;p=android-x86%2Fexternal-ffmpeg.git

Merge commit '502ab21af0ca68f76d6112722c46d2f35c004053'

* commit '502ab21af0ca68f76d6112722c46d2f35c004053':
  x86: lpc: simd av_update_lls

The versions are bumped due to changes in lls.h which is used across
libraries affecting intra library ABI
(This version bump also covers changes to lls.h in the immedeatly previous
 commits)

Merged-by: Michael Niedermayer <michaelni@gmx.at>
---

78b547963362cc9827d92dc0808534b757ab05d4
diff --cc libavcodec/lpc.c
index e2c0aacc43,144bbede44..559a87d161
--- a/libavcodec/lpc.c
+++ b/libavcodec/lpc.c
@@@ -201,11 -200,10 +201,13 @@@ int ff_lpc_calc_coefs(LPCContext *s
              ref[i] = fabs(lpc[i][i]);
      } else if (lpc_type == FF_LPC_TYPE_CHOLESKY) {
          LLSModel m[2];
-         double var[MAX_LPC_ORDER+1], av_uninit(weight);
+         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));
  
 +        if(lpc_passes <= 0)
 +            lpc_passes = 2;
 +
          for(pass=0; pass<lpc_passes; pass++){
              avpriv_init_lls(&m[pass&1], max_order);
  
diff --cc libavcodec/version.h
index d0af912e26,7c82519e06..b9a2bc4b0d
--- a/libavcodec/version.h
+++ b/libavcodec/version.h
@@@ -26,11 -26,9 +26,11 @@@
   * Libavcodec version macros.
   */
  
 +#include "libavutil/avutil.h"
 +
  #define LIBAVCODEC_VERSION_MAJOR 55
- #define LIBAVCODEC_VERSION_MINOR  17
- #define LIBAVCODEC_VERSION_MICRO 101
 -#define LIBAVCODEC_VERSION_MINOR 10
 -#define LIBAVCODEC_VERSION_MICRO  0
++#define LIBAVCODEC_VERSION_MINOR  18
++#define LIBAVCODEC_VERSION_MICRO 100
  
  #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
                                                 LIBAVCODEC_VERSION_MINOR, \
diff --cc libavutil/version.h
index 72c5b117d4,fa79e12a68..b03aa4a871
--- a/libavutil/version.h
+++ b/libavutil/version.h
@@@ -74,9 -36,9 +74,9 @@@
   * @{
   */
  
 -#define LIBAVUTIL_VERSION_MAJOR 52
 -#define LIBAVUTIL_VERSION_MINOR 13
 -#define LIBAVUTIL_VERSION_MICRO  0
 +#define LIBAVUTIL_VERSION_MAJOR  52
- #define LIBAVUTIL_VERSION_MINOR  37
- #define LIBAVUTIL_VERSION_MICRO 102
++#define LIBAVUTIL_VERSION_MINOR  38
++#define LIBAVUTIL_VERSION_MICRO 100
  
  #define LIBAVUTIL_VERSION_INT   AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \
                                                 LIBAVUTIL_VERSION_MINOR, \
diff --cc libavutil/x86/lls.asm
index 0000000000,92c00fcda1..1b56db4cb4
mode 000000,100644..100644
--- a/libavutil/x86/lls.asm
+++ b/libavutil/x86/lls.asm
@@@ -1,0 -1,196 +1,196 @@@
+ ;******************************************************************************
+ ;* linear least squares model
+ ;*
+ ;* Copyright (c) 2013 Loren Merritt
+ ;*
 -;* This file is part of Libav.
++;* This file is part of FFmpeg.
+ ;*
 -;* Libav is free software; you can redistribute it and/or
++;* 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.
+ ;*
 -;* Libav is distributed in the hope that it will be useful,
++;* 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 Libav; if not, write to the Free Software
++;* License along with FFmpeg; if not, write to the Free Software
+ ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ ;******************************************************************************
+ 
+ %include "x86util.asm"
+ 
+ SECTION .text
+ 
+ %define MAX_VARS 32
+ %define MAX_VARS_ALIGN (MAX_VARS+4)
+ %define COVAR_STRIDE MAX_VARS_ALIGN*8
+ %define COVAR(x,y) [covarq + (x)*8 + (y)*COVAR_STRIDE]
+ 
+ struc LLSModel
+     .covariance:  resq MAX_VARS_ALIGN*MAX_VARS_ALIGN
+     .coeff:       resq MAX_VARS*MAX_VARS
+     .variance:    resq MAX_VARS
+     .indep_count: resd 1
+ endstruc
+ 
+ %macro ADDPD_MEM 2
+ %if cpuflag(avx)
+     vaddpd %2, %1
+ %else
+     addpd  %2, %1
+ %endif
+     mova   %1, %2
+ %endmacro
+ 
+ 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]
+     lea   varq, [varq + iq*8]
+     neg     iq
+     mov covar2q, covarq
+ .loopi:
+     ; Compute all 3 pairwise products of a 2x2 block that lies on the diagonal
+     mova    m1, [varq + iq*8]
+     mova    m3, [varq + iq*8 + 16]
+     pshufd  m4, m1, q1010
+     pshufd  m5, m1, q3232
+     pshufd  m6, m3, q1010
+     pshufd  m7, m3, q3232
+     mulpd   m0, m1, m4
+     mulpd   m1, m1, m5
+     lea covarq, [covar2q + 16]
+     ADDPD_MEM COVAR(-2,0), m0
+     ADDPD_MEM COVAR(-2,1), m1
+     lea     jq, [iq + 2]
+     cmp     jd, -2
+     jg .skip4x4
+ .loop4x4:
+     ; Compute all 16 pairwise products of a 4x4 block
+     mulpd   m0, m4, m3
+     mulpd   m1, m5, m3
+     mulpd   m2, m6, m3
+     mulpd   m3, m3, m7
+     ADDPD_MEM COVAR(0,0), m0
+     ADDPD_MEM COVAR(0,1), m1
+     ADDPD_MEM COVAR(0,2), m2
+     ADDPD_MEM COVAR(0,3), m3
+     mova    m3, [varq + jq*8 + 16]
+     mulpd   m0, m4, m3
+     mulpd   m1, m5, m3
+     mulpd   m2, m6, m3
+     mulpd   m3, m3, m7
+     ADDPD_MEM COVAR(2,0), m0
+     ADDPD_MEM COVAR(2,1), m1
+     ADDPD_MEM COVAR(2,2), m2
+     ADDPD_MEM COVAR(2,3), m3
+     mova    m3, [varq + jq*8 + 32]
+     add covarq, 32
+     add     jq, 4
+     cmp     jd, -2
+     jle .loop4x4
+ .skip4x4:
+     test    jd, jd
+     jg .skip2x4
+     mulpd   m4, m3
+     mulpd   m5, m3
+     mulpd   m6, m3
+     mulpd   m7, m3
+     ADDPD_MEM COVAR(0,0), m4
+     ADDPD_MEM COVAR(0,1), m5
+     ADDPD_MEM COVAR(0,2), m6
+     ADDPD_MEM COVAR(0,3), m7
+ .skip2x4:
+     add     iq, 4
+     add covar2q, 4*COVAR_STRIDE+32
+     cmp     id, -2
+     jle .loopi
+     test    id, id
+     jg .ret
+     mov     jq, iq
+     %define covarq covar2q
+ .loop2x1:
+     movsd   m0, [varq + iq*8]
+     movlhps m0, m0
+     mulpd   m0, [varq + jq*8]
+     ADDPD_MEM COVAR(0,0), m0
+     inc     iq
+     add covarq, COVAR_STRIDE
+     test    id, id
+     jle .loop2x1
+ .ret:
+     REP_RET
+ 
+ INIT_YMM avx
+ cglobal update_lls, 3,6,8, ctx, var, count, i, j, count2
+     %define covarq ctxq
+     mov  countd, [ctxq + LLSModel.indep_count]
+     lea count2d, [countq-2]
+     xor     id, id
+ .loopi:
+     ; Compute all 10 pairwise products of a 4x4 block that lies on the diagonal
+     mova    ymm1, [varq + iq*8]
+     vbroadcastsd ymm4, [varq + iq*8]
+     vbroadcastsd ymm5, [varq + iq*8 + 8]
+     vbroadcastsd ymm6, [varq + iq*8 + 16]
+     vbroadcastsd ymm7, [varq + iq*8 + 24]
+     vextractf128 xmm3, ymm1, 1
+     vmulpd  ymm0, ymm1, ymm4
+     vmulpd  ymm1, ymm1, ymm5
+     vmulpd  xmm2, xmm3, xmm6
+     vmulpd  xmm3, xmm3, xmm7
+     ADDPD_MEM COVAR(iq  ,0), ymm0
+     ADDPD_MEM COVAR(iq  ,1), ymm1
+     ADDPD_MEM COVAR(iq+2,2), xmm2
+     ADDPD_MEM COVAR(iq+2,3), xmm3
+     lea     jd, [iq + 4]
+     cmp     jd, count2d
+     jg .skip4x4
+ .loop4x4:
+     ; Compute all 16 pairwise products of a 4x4 block
+     mova    ymm3, [varq + jq*8]
+     vmulpd  ymm0, ymm3, ymm4
+     vmulpd  ymm1, ymm3, ymm5
+     vmulpd  ymm2, ymm3, ymm6
+     vmulpd  ymm3, ymm3, ymm7
+     ADDPD_MEM COVAR(jq,0), ymm0
+     ADDPD_MEM COVAR(jq,1), ymm1
+     ADDPD_MEM COVAR(jq,2), ymm2
+     ADDPD_MEM COVAR(jq,3), ymm3
+     add     jd, 4
+     cmp     jd, count2d
+     jle .loop4x4
+ .skip4x4:
+     cmp     jd, countd
+     jg .skip2x4
+     mova    xmm3, [varq + jq*8]
+     vmulpd  xmm0, xmm3, xmm4
+     vmulpd  xmm1, xmm3, xmm5
+     vmulpd  xmm2, xmm3, xmm6
+     vmulpd  xmm3, xmm3, xmm7
+     ADDPD_MEM COVAR(jq,0), xmm0
+     ADDPD_MEM COVAR(jq,1), xmm1
+     ADDPD_MEM COVAR(jq,2), xmm2
+     ADDPD_MEM COVAR(jq,3), xmm3
+ .skip2x4:
+     add     id, 4
+     add covarq, 4*COVAR_STRIDE
+     cmp     id, count2d
+     jle .loopi
+     cmp     id, countd
+     jg .ret
+     mov     jd, id
+ .loop2x1:
+     vmovddup xmm0, [varq + iq*8]
+     vmulpd   xmm0, [varq + jq*8]
+     ADDPD_MEM COVAR(jq,0), xmm0
+     inc     id
+     add covarq, COVAR_STRIDE
+     cmp     id, countd
+     jle .loop2x1
+ .ret:
+     REP_RET
diff --cc libavutil/x86/lls_init.c
index 0000000000,8a80f83002..6437be958c
mode 000000,100644..100644
--- a/libavutil/x86/lls_init.c
+++ b/libavutil/x86/lls_init.c
@@@ -1,0 -1,38 +1,38 @@@
+ /*
+  * linear least squares model
+  *
+  * Copyright (c) 2013 Loren Merritt
+  *
 - * This file is part of Libav.
++ * This file is part of FFmpeg.
+  *
 - * Libav is free software; you can redistribute it and/or
++ * 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.
+  *
 - * Libav is distributed in the hope that it will be useful,
++ * 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 Libav; if not, write to the Free Software
++ * License along with FFmpeg; if not, write to the Free Software
+  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+  */
+ 
+ #include "libavutil/lls.h"
+ #include "libavutil/x86/cpu.h"
+ 
+ void ff_update_lls_sse2(LLSModel *m, double *var);
+ void ff_update_lls_avx(LLSModel *m, double *var);
+ 
+ av_cold void ff_init_lls_x86(LLSModel *m)
+ {
+     int cpu_flags = av_get_cpu_flags();
+     if (EXTERNAL_SSE2(cpu_flags)) {
+         m->update_lls = ff_update_lls_sse2;
+     }
+     if (EXTERNAL_AVX(cpu_flags)) {
+         m->update_lls = ff_update_lls_avx;
+     }
+ }