/* * AltiVec acceleration for colorspace conversion * * copyright (C) 2004 Marc Hoffman <marc.hoffman@analog.com> * * This file is part of FFmpeg. * * FFmpeg is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 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 General Public License for more details. * * You should have received a copy of the GNU 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 *//*Convert I420 YV12 to RGB in various formats,  it rejects images that are not in 420 formats,  it rejects images that don't have widths of multiples of 16,  it rejects images that don't have heights of multiples of 2.Reject defers to C simulation code.Lots of optimizations to be done here.1. Need to fix saturation code. I just couldn't get it to fly with packs   and adds, so we currently use max/min to clip.2. The inefficient use of chroma loading needs a bit of brushing up.3. Analysis of pipeline stalls needs to be done. Use shark to identify   pipeline stalls.MODIFIED to calculate coeffs from currently selected color space.MODIFIED core to be a macro where you specify the output format.ADDED UYVY conversion which is never called due to some thing in swscale.CORRECTED algorithim selection to be strict on input formats.ADDED runtime detection of AltiVec.ADDED altivec_yuv2packedX vertical scl + RGB converterMarch 27,2004PERFORMANCE ANALYSISThe C version uses 25% of the processor or ~250Mips for D1 video rawvideoused as test.The AltiVec version uses 10% of the processor or ~100Mips for D1 videosame sequence.720 * 480 * 30  ~10MPSso we have roughly 10 clocks per pixel. This is too high, something hasto be wrong.OPTIMIZED clip codes to utilize vec_max and vec_packs removing theneed for vec_min.OPTIMIZED DST OUTPUT cache/DMA controls. We are pretty much guaranteed to havethe input video frame, it was just decompressed so it probably resides in L1caches. However, we are creating the output video stream. This needs to use theDSTST instruction to optimize for the cache. We couple this with the fact thatwe are not going to be visiting the input buffer again so we mark it LeastRecently Used. This shaves 25% of the processor cycles off.Now memcpy is the largest mips consumer in the system, probably dueto the inefficient X11 stuff.GL libraries seem to be very slow on this machine 1.33Ghz PB runningJaguar, this is not the case for my 1Ghz PB.  I thought it might bea versioning issue, however I have libGL.1.2.dylib for bothmachines. (We need to figure this out now.)GL2 libraries work now with patch for RGB32.NOTE: quartz vo driver ARGB32_to_RGB24 consumes 30% of the processor.Integrated luma prescaling adjustment for saturation/contrast/brightnessadjustment.*/#include <stdio.h>#include <stdlib.h>#include <string.h>#include <inttypes.h>#include <assert.h>#include "config.h"#ifdef HAVE_MALLOC_H#include <malloc.h>#endif#include "rgb2rgb.h"#include "swscale.h"#include "swscale_internal.h"#undef PROFILE_THE_BEAST#undef INC_SCALINGtypedef unsigned char ubyte;typedef signed char   sbyte;/* RGB interleaver, 16 planar pels 8-bit samples per channel in   homogeneous vector registers x0,x1,x2 are interleaved with the   following technique:      o0 = vec_mergeh (x0,x1);      o1 = vec_perm (o0, x2, perm_rgb_0);      o2 = vec_perm (o0, x2, perm_rgb_1);      o3 = vec_mergel (x0,x1);      o4 = vec_perm (o3,o2,perm_rgb_2);      o5 = vec_perm (o3,o2,perm_rgb_3);  perm_rgb_0:   o0(RG).h v1(B) --> o1*              0   1  2   3   4             rgbr|gbrg|brgb|rgbr             0010 0100 1001 0010             0102 3145 2673 894A  perm_rgb_1:   o0(RG).h v1(B) --> o2              0   1  2   3   4             gbrg|brgb|bbbb|bbbb             0100 1001 1111 1111             B5CD 6EF7 89AB CDEF  perm_rgb_2:   o3(RG).l o2(rgbB.l) --> o4*              0   1  2   3   4             gbrg|brgb|rgbr|gbrg             1111 1111 0010 0100             89AB CDEF 0182 3945  perm_rgb_2:   o3(RG).l o2(rgbB.l) ---> o5*              0   1  2   3   4             brgb|rgbr|gbrg|brgb             1001 0010 0100 1001             a67b 89cA BdCD eEFf*/staticconst vector unsigned char  perm_rgb_0 = (const vector unsigned char)AVV(0x00,0x01,0x10,0x02,0x03,0x11,0x04,0x05,                                               0x12,0x06,0x07,0x13,0x08,0x09,0x14,0x0a),  perm_rgb_1 = (const vector unsigned char)AVV(0x0b,0x15,0x0c,0x0d,0x16,0x0e,0x0f,0x17,                                               0x18,0x19,0x1a,0x1b,0x1c,0x1d,0x1e,0x1f),  perm_rgb_2 = (const vector unsigned char)AVV(0x10,0x11,0x12,0x13,0x14,0x15,0x16,0x17,                                               0x00,0x01,0x18,0x02,0x03,0x19,0x04,0x05),  perm_rgb_3 = (const vector unsigned char)AVV(0x1a,0x06,0x07,0x1b,0x08,0x09,0x1c,0x0a,                                               0x0b,0x1d,0x0c,0x0d,0x1e,0x0e,0x0f,0x1f);#define vec_merge3(x2,x1,x0,y0,y1,y2)       \do {                                        \    typeof(x0) o0,o2,o3;                    \        o0 = vec_mergeh (x0,x1);            \        y0 = vec_perm (o0, x2, perm_rgb_0); \        o2 = vec_perm (o0, x2, perm_rgb_1); \        o3 = vec_mergel (x0,x1);            \        y1 = vec_perm (o3,o2,perm_rgb_2);   \        y2 = vec_perm (o3,o2,perm_rgb_3);   \} while(0)#define vec_mstbgr24(x0,x1,x2,ptr)      \do {                                    \    typeof(x0) _0,_1,_2;                \    vec_merge3 (x0,x1,x2,_0,_1,_2);     \    vec_st (_0, 0, ptr++);              \    vec_st (_1, 0, ptr++);              \    vec_st (_2, 0, ptr++);              \}  while (0);#define vec_mstrgb24(x0,x1,x2,ptr)      \do {                                    \    typeof(x0) _0,_1,_2;                \    vec_merge3 (x2,x1,x0,_0,_1,_2);     \    vec_st (_0, 0, ptr++);              \    vec_st (_1, 0, ptr++);              \    vec_st (_2, 0, ptr++);              \}  while (0);/* pack the pixels in rgb0 format   msb R   lsb 0*/#define vec_mstrgb32(T,x0,x1,x2,x3,ptr)                                       \do {                                                                          \    T _0,_1,_2,_3;                                                            \    _0 = vec_mergeh (x0,x1);                                                  \    _1 = vec_mergeh (x2,x3);                                                  \    _2 = (T)vec_mergeh ((vector unsigned short)_0,(vector unsigned short)_1); \    _3 = (T)vec_mergel ((vector unsigned short)_0,(vector unsigned short)_1); \    vec_st (_2, 0*16, (T *)ptr);                                              \    vec_st (_3, 1*16, (T *)ptr);                                              \    _0 = vec_mergel (x0,x1);                                                  \    _1 = vec_mergel (x2,x3);                                                  \    _2 = (T)vec_mergeh ((vector unsigned short)_0,(vector unsigned short)_1); \    _3 = (T)vec_mergel ((vector unsigned short)_0,(vector unsigned short)_1); \    vec_st (_2, 2*16, (T *)ptr);                                              \    vec_st (_3, 3*16, (T *)ptr);                                              \    ptr += 4;                                                                 \}  while (0);/*  | 1     0       1.4021   | | Y |  | 1    -0.3441 -0.7142   |x| Cb|  | 1     1.7718  0        | | Cr|  Y:      [-128 127]  Cb/Cr : [-128 127]  typical yuv conversion work on Y: 0-255 this version has been optimized for jpeg decode.*/#define vec_unh(x) \    (vector signed short) \        vec_perm(x,(typeof(x))AVV(0),\                 (vector unsigned char)AVV(0x10,0x00,0x10,0x01,0x10,0x02,0x10,0x03,\                                           0x10,0x04,0x10,0x05,0x10,0x06,0x10,0x07))#define vec_unl(x) \    (vector signed short) \        vec_perm(x,(typeof(x))AVV(0),\                 (vector unsigned char)AVV(0x10,0x08,0x10,0x09,0x10,0x0A,0x10,0x0B,\                                           0x10,0x0C,0x10,0x0D,0x10,0x0E,0x10,0x0F))#define vec_clip_s16(x) \    vec_max (vec_min (x, (vector signed short)AVV(235,235,235,235,235,235,235,235)),\                         (vector signed short)AVV( 16, 16, 16, 16, 16, 16, 16, 16))#define vec_packclp(x,y) \    (vector unsigned char)vec_packs \        ((vector unsigned short)vec_max (x,(vector signed short) AVV(0)), \         (vector unsigned short)vec_max (y,(vector signed short) AVV(0)))//#define out_pixels(a,b,c,ptr) vec_mstrgb32(typeof(a),((typeof (a))AVV(0)),a,a,a,ptr)static inline void cvtyuvtoRGB (SwsContext *c,                                vector signed short Y, vector signed short U, vector signed short V,                                vector signed short *R, vector signed short *G, vector signed short *B){    vector signed   short vx,ux,uvx;    Y = vec_mradds (Y, c->CY, c->OY);    U  = vec_sub (U,(vector signed short)                    vec_splat((vector signed short)AVV(128),0));    V  = vec_sub (V,(vector signed short)                    vec_splat((vector signed short)AVV(128),0));    //   ux  = (CBU*(u<<c->CSHIFT)+0x4000)>>15;    ux = vec_sl (U, c->CSHIFT);    *B = vec_mradds (ux, c->CBU, Y);    // vx  = (CRV*(v<<c->CSHIFT)+0x4000)>>15;    vx = vec_sl (V, c->CSHIFT);    *R = vec_mradds (vx, c->CRV, Y);    // uvx = ((CGU*u) + (CGV*v))>>15;    uvx = vec_mradds (U, c->CGU, Y);    *G  = vec_mradds (V, c->CGV, uvx);}/*  ------------------------------------------------------------------------------  CS converters  ------------------------------------------------------------------------------*/#define DEFCSP420_CVT(name,out_pixels)                                  \static int altivec_##name (SwsContext *c,                               \                           unsigned char **in, int *instrides,          \                           int srcSliceY,        int srcSliceH,         \                           unsigned char **oplanes, int *outstrides)    \{                                                                       \    int w = c->srcW;                                                    \    int h = srcSliceH;                                                  \    int i,j;                                                            \    int instrides_scl[3];                                               \    vector unsigned char y0,y1;                                         \                                                                        \    vector signed char  u,v;                                            \                                                                        \    vector signed short Y0,Y1,Y2,Y3;                                    \    vector signed short U,V;                                            \    vector signed short vx,ux,uvx;                                      \    vector signed short vx0,ux0,uvx0;                                   \    vector signed short vx1,ux1,uvx1;                                   \    vector signed short R0,G0,B0;                                       \    vector signed short R1,G1,B1;                                       \    vector unsigned char R,G,B;                                         \                                                                        \    vector unsigned char *y1ivP, *y2ivP, *uivP, *vivP;                  \    vector unsigned char align_perm;                                    \                                                                        \    vector signed short                                                 \        lCY  = c->CY,                                                   \        lOY  = c->OY,                                                   \        lCRV = c->CRV,                                                  \        lCBU = c->CBU,                                                  \        lCGU = c->CGU,                                                  \        lCGV = c->CGV;                                                  \                                                                        \    vector unsigned short lCSHIFT = c->CSHIFT;                          \                                                                        \    ubyte *y1i   = in[0];                                               \    ubyte *y2i   = in[0]+instrides[0];                                  \    ubyte *ui    = in[1];                                               \    ubyte *vi    = in[2];                                               \                                                                        \    vector unsigned char *oute                                          \        = (vector unsigned char *)                                      \            (oplanes[0]+srcSliceY*outstrides[0]);                       \