URL
https://opencores.org/ocsvn/oc-h264-encoder/oc-h264-encoder/trunk
Subversion Repositories oc-h264-encoder
[/] [oc-h264-encoder/] [trunk/] [x264/] [patches/] [x264-e381f6d-or32-or1ksim-with-fp-1.2.patch] - Rev 51
Compare with Previous | Blame | View Log
diff --exclude=.git --exclude=/gitignore -Naur x264/common/common.c x264-or/common/common.c --- x264/common/common.c 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/common/common.c 2009-10-28 15:05:29.000000000 +0100 @@ -764,8 +764,10 @@ align_buf = malloc( i_size ); #elif defined( HAVE_MALLOC_H ) align_buf = memalign( 16, i_size ); + //fprintf(stderr, "memalign, result addr: 0x%.8x\n", (unsigned int) align_buf); #else uint8_t *buf = malloc( i_size + 15 + sizeof(void **) + sizeof(int) ); + //fprintf(stderr, "malloc, result addr: 0x%.8x\n", (unsigned int) buf); if( buf ) { align_buf = buf + 15 + sizeof(void **) + sizeof(int); diff --exclude=.git --exclude=/gitignore -Naur x264/common/common.h x264-or/common/common.h --- x264/common/common.h 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/common/common.h 2009-11-10 22:42:24.000000000 +0100 @@ -51,7 +51,8 @@ #define X264_BFRAME_MAX 16 #define X264_THREAD_MAX 128 #define X264_PCM_COST (386*8) -#define X264_LOOKAHEAD_MAX 250 +//#define X264_LOOKAHEAD_MAX 250 +#define X264_LOOKAHEAD_MAX 140 // arbitrary, but low because SATD scores are 1/4 normal #define X264_LOOKAHEAD_QP 12 @@ -260,7 +261,7 @@ */ #define X264_SCAN8_SIZE (6*8) #define X264_SCAN8_0 (4+1*8) - +/* this array indicates the position in the */ static const int x264_scan8[16+2*4+3] = { /* Luma */ @@ -518,7 +519,9 @@ /* space for p_fenc and p_fdec */ #define FENC_STRIDE 16 #define FDEC_STRIDE 32 - ALIGNED_16( uint8_t fenc_buf[24*FENC_STRIDE] ); + /* comment regarding FDEC_STRIDE: the catch with fdec is that the chroma blocks must be aligned to 8 and the luma blocks must be aligned to 16 but we need to have pixels on the left side for prediction so we have a lot of padding on the left side. so for luma: X X X X X X X X X X X X X X X L _ _ _ _ _ ... where X are junk, L is the left pixel, _ are normal pixels, so 32 wide total */ + + ALIGNED_16( uint8_t fenc_buf[24*FENC_STRIDE] ); /* FENC stores both Y, U, and V, 16x16 + 8x8 + 8x8 = 24 * 16*/ ALIGNED_16( uint8_t fdec_buf[27*FDEC_STRIDE] ); /* i4x4 and i8x8 backup data, for skipping the encode stage when possible */ @@ -551,9 +554,29 @@ /* pointer over mb of the references */ int i_fref[2]; - uint8_t *p_fref[2][32][4+2]; /* last: lN, lH, lV, lHV, cU, cV */ - uint16_t *p_integral[2][16]; - + uint8_t *p_fref[2][32][4+2]; /* last: 4hpel pointers, 2 chroma pointers: lN (normal), lH (horizontally interpolated), lV (vertically interpolated), lHV (center position), cU (chroma U), cV (chroma V) + Important!: This is our post-motion compensation frame data, stored in half-pixel (hpel) resolution. + During motion compensation, hpel data is calculated via 6-tap FIR filter (very similar to Lanczos). It is an even filter (even function) so every input pixel corresponds to 4 output pixels + one of those 4 is equal to the input, so the data contains one fullpel pixel, one H, one V, one C (aka HV) + From this data, qpel is calculated via interpolation, which is easy and quick to do. + F 0 H 1 F + 2 3 4 5 6 + V 7 C 8 V + 9 A B C D + F _ H _ F + Here 0 is the average of F and H + 2 is the average of F and V + 3 is the average of H and V + 4 is the average of H and C + etc. + So each qpel position is the result of averaging two known hpel positions.*/ + uint16_t *p_integral[2][16]; /* pointers to "integral": for each position X,Y, it's the sum of values in an 8x8 block with that point X,Y as the upper left of that block + Discussion on this value: + < pengvado> Dark_Shikari: "sum of values in an 8x8 block" is what I would call "dc". the variable name comes because it used to be an indefinite integral, i.e. the sum of everything above and to the left of that point. + < pengvado> but if you're only interested in 16x16, 16x8, 8x16, and 8x8, then caching 8x8 dcs is strictly faster + */ + + /* fref stride */ int i_stride[3]; } pic; diff --exclude=.git --exclude=/gitignore -Naur x264/common/frame.c x264-or/common/frame.c --- x264/common/frame.c 2009-10-25 17:41:10.000000000 +0100 +++ x264-or/common/frame.c 2009-10-28 15:05:29.000000000 +0100 @@ -31,6 +31,8 @@ x264_frame_t *frame; int i, j; + //V(fprintf(stderr, "x264_frame_new\n")); + int i_mb_count = h->mb.i_mb_count; int i_stride, i_width, i_lines; int i_padv = PADV << h->param.b_interlaced; @@ -981,11 +983,14 @@ x264_frame_t *x264_frame_pop_unused( x264_t *h, int b_fdec ) { + //V(fprintf(stderr, "x264_frame_pop_unused\n")); x264_frame_t *frame; if( h->frames.unused[b_fdec][0] ) frame = x264_frame_pop( h->frames.unused[b_fdec] ); else frame = x264_frame_new( h, b_fdec ); + + //V(fprintf(stderr, "x264_frame_pop_unused: frame ptr = 0x%.8x\n", (unsigned long) frame)); if( !frame ) return NULL; frame->b_last_minigop_bframe = 0; diff --exclude=.git --exclude=/gitignore -Naur x264/common/macroblock.c x264-or/common/macroblock.c --- x264/common/macroblock.c 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/common/macroblock.c 2009-11-09 13:02:31.000000000 +0100 @@ -118,7 +118,7 @@ int i_count = 0; - if( i_refc == -2 ) + if( i_refc == -2 )/* -2 = unavailable */ { i_refc = h->mb.cache.ref[i_list][X264_SCAN8_0 - 8 - 1]; mv_c = h->mb.cache.mv[i_list][X264_SCAN8_0 - 8 - 1]; diff --exclude=.git --exclude=/gitignore -Naur x264/common/mdate.c x264-or/common/mdate.c --- x264/common/mdate.c 2009-10-25 17:41:10.000000000 +0100 +++ x264-or/common/mdate.c 2009-10-28 15:05:29.000000000 +0100 @@ -32,9 +32,12 @@ int64_t x264_mdate( void ) { #ifndef __MINGW32__ + /* --jb struct timeval tv_date; gettimeofday( &tv_date, NULL ); return( (int64_t) tv_date.tv_sec * 1000000 + (int64_t) tv_date.tv_usec ); + */ + return 0; #else struct _timeb tb; _ftime(&tb); diff --exclude=.git --exclude=/gitignore -Naur x264/common/or32/or32.h x264-or/common/or32/or32.h --- x264/common/or32/or32.h 1970-01-01 01:00:00.000000000 +0100 +++ x264-or/common/or32/or32.h 2009-11-17 10:20:27.000000000 +0100 @@ -0,0 +1,22 @@ +/* Struct for SAD/SSD hardware module */ +typedef struct { + uint32_t control; + uint32_t result; + uint32_t pix_ptr1; + uint32_t pix_ptr2; + uint32_t stride1; + uint32_t stride2; + uint32_t x; + uint32_t y; +} x264_or32_sadssdmod_regs_t; + +#define OR32_SADSSDMOD_REG_BASE 0x26400000 + +#define OR32_SADSSDMOD_CONTROL_REG_BASE (OR32_SADSSDMOD_REG_BASE+0) +#define OR32_SADSSDMOD_CONTROL_REG_BUSY 0x1 +#define OR32_SADSSDMOD_CONTROL_REG_MODE_MASK 0x2 +#define OR32_SADSSDMOD_CONTROL_REG_MODE_SAD (OR32_SADSSDMOD_CONTROL_REG_MODE_MASK & 0x0) +#define OR32_SADSSDMOD_CONTROL_REG_MODE_SSD (OR32_SADSSDMOD_CONTROL_REG_MODE_MASK & 0xff) + + +#define OR32_SADSSDMOD 1 diff --exclude=.git --exclude=/gitignore -Naur x264/common/pixel.c x264-or/common/pixel.c --- x264/common/pixel.c 2009-10-25 17:41:10.000000000 +0100 +++ x264-or/common/pixel.c 2009-11-17 10:26:39.000000000 +0100 @@ -22,6 +22,7 @@ *****************************************************************************/ #include "common.h" +#include "x264.h" #ifdef HAVE_MMX # include "x86/pixel.h" @@ -35,29 +36,51 @@ #ifdef ARCH_UltraSparc # include "sparc/pixel.h" #endif +#ifdef ARCH_OR32 +# include "or32/or32.h" +#endif /**************************************************************************** * pixel_sad_WxH ****************************************************************************/ -#define PIXEL_SAD_C( name, lx, ly ) \ -static int name( uint8_t *pix1, int i_stride_pix1, \ - uint8_t *pix2, int i_stride_pix2 ) \ -{ \ - int i_sum = 0; \ - int x, y; \ - for( y = 0; y < ly; y++ ) \ - { \ - for( x = 0; x < lx; x++ ) \ - { \ - i_sum += abs( pix1[x] - pix2[x] ); \ - } \ - pix1 += i_stride_pix1; \ - pix2 += i_stride_pix2; \ - } \ - return i_sum; \ -} - +#if OR32_SADSSDMOD +#define PIXEL_SAD_C( name, lx, ly ) \ + static int name( uint8_t *pix1, int i_stride_pix1, \ + uint8_t *pix2, int i_stride_pix2 ) \ + { /* Configure and set the hardware SAD/SSD module running */ \ + volatile x264_or32_sadssdmod_regs_t* sadssdmod_regs; \ + sadssdmod_regs = (x264_or32_sadssdmod_regs_t*) OR32_SADSSDMOD_REG_BASE; \ + sadssdmod_regs->pix_ptr1 = (uint32_t)pix1; \ + sadssdmod_regs->pix_ptr2 = (uint32_t)pix2; \ + sadssdmod_regs->stride1 = (uint32_t)i_stride_pix1; \ + sadssdmod_regs->stride2 = (uint32_t)i_stride_pix2; \ + sadssdmod_regs->x = (uint32_t)lx; \ + sadssdmod_regs->y = (uint32_t)ly; \ + sadssdmod_regs->control = (uint32_t)(OR32_SADSSDMOD_CONTROL_REG_MODE_SAD | \ + OR32_SADSSDMOD_CONTROL_REG_BUSY) ; \ + while (sadssdmod_regs->control & OR32_SADSSDMOD_CONTROL_REG_BUSY); \ + return (int) sadssdmod_regs->result; \ + } +#else +#define PIXEL_SAD_C( name, lx, ly ) \ + static int name( uint8_t *pix1, int i_stride_pix1, \ + uint8_t *pix2, int i_stride_pix2 ) \ + { \ +int i_sum = 0; \ +int x, y; \ +for( y = 0; y < ly; y++ ) \ + { \ + for( x = 0; x < lx; x++ ) \ + { \ + i_sum += abs( pix1[x] - pix2[x] ); \ + } \ + pix1 += i_stride_pix1; \ + pix2 += i_stride_pix2; \ + } \ +return i_sum; \ +} +#endif PIXEL_SAD_C( x264_pixel_sad_16x16, 16, 16 ) PIXEL_SAD_C( x264_pixel_sad_16x8, 16, 8 ) @@ -71,10 +94,29 @@ /**************************************************************************** * pixel_ssd_WxH ****************************************************************************/ +#if OR32_SADSSDMOD +#define PIXEL_SSD_C( name, lx, ly ) \ +static int name( uint8_t *pix1, int i_stride_pix1, \ + uint8_t *pix2, int i_stride_pix2 ) \ +{ /* Configure and set the hardware SAD/SSD module running */ \ + volatile x264_or32_sadssdmod_regs_t* sadssdmod_regs; \ + sadssdmod_regs = (x264_or32_sadssdmod_regs_t*) OR32_SADSSDMOD_REG_BASE; \ + sadssdmod_regs->pix_ptr1 = (uint32_t)pix1; \ + sadssdmod_regs->pix_ptr2 = (uint32_t)pix2; \ + sadssdmod_regs->stride1 = (uint32_t)i_stride_pix1; \ + sadssdmod_regs->stride2 = (uint32_t)i_stride_pix2; \ + sadssdmod_regs->x = (uint32_t)lx; \ + sadssdmod_regs->y = (uint32_t)ly; \ + sadssdmod_regs->control = (uint32_t)(OR32_SADSSDMOD_CONTROL_REG_MODE_SSD | \ + OR32_SADSSDMOD_CONTROL_REG_BUSY) ; \ + while (sadssdmod_regs->control & OR32_SADSSDMOD_CONTROL_REG_BUSY); \ + return (int) sadssdmod_regs->result; \ +} +#else #define PIXEL_SSD_C( name, lx, ly ) \ static int name( uint8_t *pix1, int i_stride_pix1, \ uint8_t *pix2, int i_stride_pix2 ) \ -{ \ +{ \ int i_sum = 0; \ int x, y; \ for( y = 0; y < ly; y++ ) \ @@ -89,7 +131,7 @@ } \ return i_sum; \ } - +#endif PIXEL_SSD_C( x264_pixel_ssd_16x16, 16, 16 ) PIXEL_SSD_C( x264_pixel_ssd_16x8, 16, 8 ) PIXEL_SSD_C( x264_pixel_ssd_8x16, 8, 16 ) @@ -387,6 +429,7 @@ /**************************************************************************** * pixel_sad_x4 + * Debug printf : V(fprintf(stderr, "x264_pixel_sadx4 size : *fenc: 0x%.8x *pix0: 0x%.8x *pix1: 0x%.8x *pix2: 0x%.8x *pix3: 0x%.8x i_stride: %d *scores: 0x%.8x\n", (unsigned long) fenc, (unsigned long) pix0, (unsigned long) pix1, (unsigned long)pix2, (unsigned long)pix3, i_stride, (unsigned long) scores)); \ ****************************************************************************/ #define SAD_X( size ) \ static void x264_pixel_sad_x3_##size( uint8_t *fenc, uint8_t *pix0, uint8_t *pix1, uint8_t *pix2, int i_stride, int scores[3] )\ diff --exclude=.git --exclude=/gitignore -Naur x264/configure x264-or/configure --- x264/configure 2009-10-25 17:41:10.000000000 +0100 +++ x264-or/configure 2009-10-28 15:05:29.000000000 +0100 @@ -29,7 +29,7 @@ rm -f conftest.c [ -n "$1" ] && echo "#include <$1>" > conftest.c echo "int main () { $3 return 0; }" >> conftest.c - $CC conftest.c $CFLAGS $LDFLAGS $2 -o conftest 2>$DEVNULL + $CC conftest.c $CFLAGS $LDFLAGS $2 -o conftest #2>$DEVNULL } as_check() { diff --exclude=.git --exclude=/gitignore -Naur x264/encoder/analyse.c x264-or/encoder/analyse.c --- x264/encoder/analyse.c 2009-10-25 17:41:10.000000000 +0100 +++ x264-or/encoder/analyse.c 2009-11-09 13:11:23.000000000 +0100 @@ -27,6 +27,7 @@ #include <unistd.h> #include "common/common.h" +#include "x264.h" #include "common/cpu.h" #include "macroblock.h" #include "me.h" @@ -34,6 +35,7 @@ #include "analyse.h" #include "rdo.c" + typedef struct { /* 16x16 */ @@ -233,40 +235,75 @@ 5, 3, 3, 1 }; +extern const float x264_analyse_init_log2_array[]; //jb + static void x264_analyse_update_cache( x264_t *h, x264_mb_analysis_t *a ); static uint16_t x264_cost_ref[92][3][33]; static x264_pthread_mutex_t cost_ref_mutex = X264_PTHREAD_MUTEX_INITIALIZER; +/* +float log2f( float n ) +{ +// log(n)/log(2) is log2. +//return log10( n ) / log10( 2 ); + return (float) (log( n ) / _M_LOG2_E); +} +*/ int x264_analyse_init_costs( x264_t *h, int qp ) { int i, j; int lambda = x264_lambda_tab[qp]; + float tmp_f,lambda_f = lambda; + if( h->cost_mv[lambda] ) return 0; + + fprintf( stderr, + "x264_analyse_init_costs: lambda (= x264_lambda_tab[%d]) = %d\n", + qp, lambda); + /* factor of 4 from qpel, 2 from sign, and 2 because mv can be opposite from mvp */ CHECKED_MALLOC( h->cost_mv[lambda], (4*4*2048 + 1) * sizeof(uint16_t) ); + h->cost_mv[lambda] += 2*4*2048; + + /* Computationally expensive thing: logs, floating point math, much of -- jb*/ for( i = 0; i <= 2*4*2048; i++ ) - { - h->cost_mv[lambda][-i] = - h->cost_mv[lambda][i] = lambda * (log2f(i+1)*2 + 0.718f + !!i) + .5f; - } + { + + //h->cost_mv[lambda][i] = lambda * (log2f(i+1)*2 + 0.718f + !!i) + .5f; + //h->cost_mv[lambda][i] = lambda * (log2(i+1)*2 + 0.718f + !!i) + .5f; + + // precalulated array, doing the above few calculations for us + tmp_f = lambda_f * x264_analyse_init_log2_array[i]; + tmp_f += .5f; + h->cost_mv[lambda][-i] = + h->cost_mv[lambda][i] = tmp_f; + + + } x264_pthread_mutex_lock( &cost_ref_mutex ); + for( i = 0; i < 3; i++ ) - for( j = 0; j < 33; j++ ) - x264_cost_ref[lambda][i][j] = i ? lambda * bs_size_te( i, j ) : 0; + for( j = 0; j < 33; j++ ) + x264_cost_ref[lambda][i][j] = i ? lambda * bs_size_te( i, j ) : 0; + x264_pthread_mutex_unlock( &cost_ref_mutex ); + + /* for now, disable -- jb if( h->param.analyse.i_me_method >= X264_ME_ESA && !h->cost_mv_fpel[lambda][0] ) - { + { + printf("analyse_init_costs: h->param.analyse.i_me_method >= X264_ME_ESA && !h->cost_mv_fpel[lambda][0]\n"); for( j=0; j<4; j++ ) - { + { CHECKED_MALLOC( h->cost_mv_fpel[lambda][j], (4*2048 + 1) * sizeof(uint16_t) ); h->cost_mv_fpel[lambda][j] += 2*2048; for( i = -2*2048; i < 2*2048; i++ ) - h->cost_mv_fpel[lambda][j][i] = h->cost_mv[lambda][i*4+j]; + h->cost_mv_fpel[lambda][j][i] = h->cost_mv[lambda][i*4+j]; } } + */ return 0; fail: return -1; @@ -288,6 +325,7 @@ /* initialize an array of lambda*nbits for all possible mvs */ static void x264_mb_analyse_load_costs( x264_t *h, x264_mb_analysis_t *a ) { + //V(fprintf(stderr, "x264_mb_analyse_load_costs() called\n")); a->p_cost_mv = h->cost_mv[a->i_lambda]; a->p_cost_ref0 = x264_cost_ref[a->i_lambda][x264_clip3(h->sh.i_num_ref_idx_l0_active-1,0,2)]; a->p_cost_ref1 = x264_cost_ref[a->i_lambda][x264_clip3(h->sh.i_num_ref_idx_l1_active-1,0,2)]; @@ -1161,6 +1199,7 @@ (m)->p_fenc[1] = &(src)[1][((xoff)>>1)+((yoff)>>1)*FENC_STRIDE]; \ (m)->p_fenc[2] = &(src)[2][((xoff)>>1)+((yoff)>>1)*FENC_STRIDE]; + #define LOAD_HPELS(m, src, list, ref, xoff, yoff) \ (m)->p_fref[0] = &(src)[0][(xoff)+(yoff)*(m)->i_stride[0]]; \ (m)->p_fref[1] = &(src)[1][(xoff)+(yoff)*(m)->i_stride[0]]; \ @@ -1183,20 +1222,35 @@ /* 16x16 Search on all ref frame */ m.i_pixel = PIXEL_16x16; - m.p_cost_mv = a->p_cost_mv; - LOAD_FENC( &m, h->mb.pic.p_fenc, 0, 0 ); + m.p_cost_mv = a->p_cost_mv; /* where a->p_cost_mv = h->cost_mv[a->i_lambda] from x264_mb_analyse_init(). a->i_lambda related to qp of MB, and h->cost_mv[] calculated at beginning in x264_mb_analyse_local_costs() */ + /* LOAD_FENC(m, src, xoff, yoff) : + h->mb.pic.p_fenc: pointer over mb of the frame to be compressed, i think array of 3 pointers, one each to Y, Cb, Cr data. + x264_me_t m is loaded with appropriate stride variables, and pointers to the appropriate pixels + Essentially loads the me struct with stride and offsets, this is done in a macro because it's convenient + because the many partitions can use the same macro to setup x264_me_t struct + */ + LOAD_FENC( &m, h->mb.pic.p_fenc, 0, 0 ); + //printf("x264_mb_analyse_inter_p16x16: m.i_stride[0]: %d\n",m.i_stride[0]); + //printf("x264_mb_analyse_inter_p16x16: m.i_stride[1]: %d\n",m.i_stride[1]); + //usually m.i_stride[0] is 416 and m.i_stride[1] is 208 -- jb a->l0.me16x16.cost = INT_MAX; + + //printf("x264_mb_analyse_inter_p16x16: h->mb.pic.i_fref[0] = %d\n",h->mb.pic.i_fref[0]); + + // Usually just the single reference frame here for( i_ref = 0; i_ref < h->mb.pic.i_fref[0]; i_ref++ ) { - const int i_ref_cost = REF_COST( 0, i_ref ); - i_halfpel_thresh -= i_ref_cost; + const int i_ref_cost = REF_COST( 0, i_ref ); /* looks in a->p_cost_ref0/1[], which is set in x264_mb_analyse_load_costs() */ + i_halfpel_thresh -= i_ref_cost; m.i_ref_cost = i_ref_cost; m.i_ref = i_ref; /* search with ref */ - LOAD_HPELS( &m, h->mb.pic.p_fref[0][i_ref], 0, i_ref, 0, 0 ); - x264_mb_predict_mv_16x16( h, 0, i_ref, m.mvp ); + /* m, src, list, ref, xoff, yoff */ + LOAD_HPELS( &m, h->mb.pic.p_fref[0][i_ref], 0, i_ref, 0 , 0 ); + /* h,i_list, i_ref, int16_t mvp[2] */ + x264_mb_predict_mv_16x16( h, 0, i_ref, m.mvp ); x264_mb_predict_mv_ref16x16( h, 0, i_ref, mvc, &i_mvc ); x264_me_search_ref( h, &m, mvc, i_mvc, p_halfpel_thresh ); diff --exclude=.git --exclude=/gitignore -Naur x264/encoder/analyse_gen_init_array.sh x264-or/encoder/analyse_gen_init_array.sh --- x264/encoder/analyse_gen_init_array.sh 1970-01-01 01:00:00.000000000 +0100 +++ x264-or/encoder/analyse_gen_init_array.sh 2009-10-28 15:05:29.000000000 +0100 @@ -0,0 +1,32 @@ +#!/bin/bash +# Pre-generate an array of numbers according to the algorithm from analyse_init_costs() +# to speedup calculation. Here we calculate the stuff in brackets: +# lambda * (log2f(i+1)*2 + 0.718f + !!i) + .5f; +outfile=analyse_init_log2.c + +echo; echo "Generating analyse_init_log2_array[] to $outfile" +echo "This may take a minute or two..." + +#Size of the array: 4*2*2048 +numvals=16384 + +rm -f $outfile + +echo "const float x264_analyse_init_log2_array[] = {" >> $outfile + +n=0 + + for num in `seq 1 $numvals`; + do +# Use bc to do the calculation, scale down to the output to one decimal place + newnum=`echo "a=((l($num)/l(2))*2)+1.718; scale=1; a=(a*2)/2; a" | bc -l` + echo -n $newnum >> $outfile + if [ $num -lt $numvals ]; then echo -n ", " >> $outfile; fi; + # Newline every 16 values + let "n = $num % 16" + if [ $n -eq 0 ]; then echo >> $outfile; fi; + done +echo "};" >> $outfile + +echo "finished" +echo diff --exclude=.git --exclude=/gitignore -Naur x264/encoder/encoder.c x264-or/encoder/encoder.c --- x264/encoder/encoder.c 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/encoder/encoder.c 2009-10-28 15:05:29.000000000 +0100 @@ -62,6 +62,9 @@ static void x264_frame_dump( x264_t *h ) { + //printf("frame_dump()\n"); + return; + FILE *f = fopen( h->param.psz_dump_yuv, "r+b" ); int i, y; if( !f ) @@ -301,6 +304,7 @@ if( ( h->param.b_cabac && (h->cabac.p_end - h->cabac.p < 2500) ) || ( h->out.bs.p_end - h->out.bs.p < 2500 ) ) { + //V(fprintf(stderr, "x264_bitstream_check_buffer: h->out.bs.p = 0x%.8x, h->out.bs.p_end = 0x%.8x\n",h->out.bs.p_end, h->out.bs.p )); intptr_t delta; int i; @@ -695,6 +699,8 @@ static void mbcmp_init( x264_t *h ) { int satd = !h->mb.b_lossless && h->param.analyse.i_subpel_refine > 1; + //V(if (satd)fprintf(stderr, "macroblock compare functions. satd enabled\n")); + //V(if (!satd)fprintf(stderr, "macroblock compare functions. satd not enabled\n")); memcpy( h->pixf.mbcmp, satd ? h->pixf.satd : h->pixf.sad_aligned, sizeof(h->pixf.mbcmp) ); memcpy( h->pixf.mbcmp_unaligned, satd ? h->pixf.satd : h->pixf.sad, sizeof(h->pixf.mbcmp_unaligned) ); h->pixf.intra_mbcmp_x3_16x16 = satd ? h->pixf.intra_satd_x3_16x16 : h->pixf.intra_sad_x3_16x16; @@ -750,15 +756,14 @@ x264_t *h; char buf[1000], *p; int i, qp, i_slicetype_length; - + CHECKED_MALLOCZERO( h, sizeof(x264_t) ); - + /* Create a copy of param */ memcpy( &h->param, param, sizeof(x264_param_t) ); - if( param->param_free ) param->param_free( param ); - + if( x264_validate_parameters( h ) < 0 ) goto fail; @@ -772,7 +777,7 @@ h->param.rc.psz_stat_in = strdup( h->param.rc.psz_stat_in ); x264_set_aspect_ratio( h, param, 1 ); - + x264_reduce_fraction( &h->param.i_fps_num, &h->param.i_fps_den ); /* Init x264_t */ @@ -831,8 +836,8 @@ h->i_ref0 = 0; h->i_ref1 = 0; - x264_rdo_init(); - + x264_rdo_init(); + /* init CPU functions */ x264_predict_16x16_init( h->param.cpu, h->predict_16x16 ); x264_predict_8x8c_init( h->param.cpu, h->predict_8x8c ); @@ -870,17 +875,25 @@ p += sprintf( p, " none!" ); x264_log( h, X264_LOG_INFO, "%s\n", buf ); + /* // should skip during cycle accurate -- jb */ + // Check the different quantisation points and their costs for( qp = h->param.rc.i_qp_min; qp <= h->param.rc.i_qp_max; qp++ ) + { + //printf("%d\n", qp); if( x264_analyse_init_costs( h, qp ) ) - goto fail; + goto fail; + } if( x264_analyse_init_costs( h, X264_LOOKAHEAD_QP ) ) goto fail; if( h->cost_mv[1][2013] != 24 ) - { + { + //fprintf(stderr, "h->cost_mv[1][2013] = %d\n",h->cost_mv[1][2013]); x264_log( h, X264_LOG_ERROR, "MV cost test failed: x264 has been miscompiled!\n" ); goto fail; } + //exit(0); // jb + h->out.i_nal = 0; h->out.i_bitstream = X264_MAX( 1000000, h->param.i_width * h->param.i_height * 4 * ( h->param.rc.i_rc_method == X264_RC_ABR ? pow( 0.95, h->param.rc.i_qp_min ) @@ -908,16 +921,17 @@ if( x264_macroblock_cache_init( h->thread[i] ) < 0 ) goto fail; } - + if( x264_lookahead_init( h, i_slicetype_length ) ) goto fail; if( x264_ratecontrol_new( h ) < 0 ) goto fail; - + + /* // skip this for now, it's a fopen() call we won't really do, plus it's an option we'll never need --jb if( h->param.psz_dump_yuv ) { - /* create or truncate the reconstructed video file */ + //create or truncate the reconstructed video file FILE *f = fopen( h->param.psz_dump_yuv, "w" ); if( f ) fclose( f ); @@ -927,6 +941,8 @@ goto fail; } } + */ + x264_log( h, X264_LOG_INFO, "profile %s, level %d.%d\n", h->sps->i_profile_idc == PROFILE_BASELINE ? "Baseline" : @@ -1307,6 +1323,8 @@ int slice_max_size = h->param.i_slice_max_size > 0 ? (h->param.i_slice_max_size-3-NALU_OVERHEAD)*8 : INT_MAX; int starting_bits = bs_pos(&h->out.bs); + //V(fprintf(stderr, "x264_slice_write\n")); + /* Slice */ x264_nal_start( h, h->i_nal_type, h->i_nal_ref_idc ); @@ -1320,6 +1338,7 @@ /* init cabac */ x264_cabac_context_init( &h->cabac, h->sh.i_type, h->sh.i_qp, h->sh.i_cabac_init_idc ); x264_cabac_encode_init ( &h->cabac, h->out.bs.p, h->out.bs.p_end ); + //V(fprintf(stderr, "\tcabac inited\n")); } h->mb.i_last_qp = h->sh.i_qp; h->mb.i_last_dqp = 0; @@ -1362,7 +1381,10 @@ x264_macroblock_encode( h ); if( x264_bitstream_check_buffer( h ) ) + { + //V(fprintf(stderr, "\tx264_bitstream_check_buffer failed\n")); return -1; + } if( h->param.b_cabac ) { @@ -1522,6 +1544,8 @@ x264_fdec_filter_row( h, h->sps->i_mb_height ); } + //V(fprintf(stderr, "x264_slice_write ok\n")); + return 0; } @@ -1625,8 +1649,8 @@ int i_nal_type, i; int i_nal_ref_idc; - int i_global_qp; + int i_global_qp; if( h->param.i_threads > 1) { int i = ++h->i_thread_phase; @@ -1653,6 +1677,7 @@ /* no data out */ *pi_nal = 0; *pp_nal = NULL; + /* ------------------- Setup new frame from picture -------------------- */ if( pic_in != NULL ) @@ -1663,7 +1688,10 @@ return -1; if( x264_frame_copy_picture( h, fenc, pic_in ) < 0 ) + { + //V(fprintf(stderr, "x264_frame_copy_picture returned >0\n")); return -1; + } if( h->param.i_width != 16 * h->sps->i_mb_width || h->param.i_height != 16 * h->sps->i_mb_height ) @@ -1684,13 +1712,15 @@ /* 2: Place the frame into the queue for its slice type decision */ x264_lookahead_put_frame( h, fenc ); - + + //V(fprintf(stderr, "x264_encoder_encode setup new frame from picture\n")); + if( h->frames.i_input <= h->frames.i_delay + 1 - h->param.i_threads ) { /* Nothing yet to encode, waiting for filling of buffers */ pic_out->i_type = X264_TYPE_AUTO; return 0; - } + } } else { @@ -1873,7 +1903,10 @@ } else if( (intptr_t)x264_slices_write( h ) ) + { + //V(fprintf(stderr, "x264_encoder_encode(): x264_slices_write() failed\n")); return -1; + } return x264_encoder_frame_end( thread_oldest, thread_current, pp_nal, pi_nal, pic_out ); } diff --exclude=.git --exclude=/gitignore -Naur x264/encoder/me.c x264-or/encoder/me.c --- x264/encoder/me.c 2009-10-25 17:41:10.000000000 +0100 +++ x264-or/encoder/me.c 2009-11-10 13:36:34.000000000 +0100 @@ -33,17 +33,17 @@ * the subme=8,9 values are much higher because any amount of satd search makes * up its time by reducing the number of qpel-rd iterations. */ static const int subpel_iterations[][4] = - {{0,0,0,0}, - {1,1,0,0}, - {0,1,1,0}, - {0,2,1,0}, - {0,2,1,1}, - {0,2,1,2}, - {0,0,2,2}, - {0,0,2,2}, - {0,0,4,10}, - {0,0,4,10}, - {0,0,4,10}}; + {{0,0,0,0}, + {1,1,0,0}, + {0,1,1,0}, + {0,2,1,0}, + {0,2,1,1}, + {0,2,1,2}, + {0,0,2,2}, + {0,0,2,2}, + {0,0,4,10}, + {0,0,4,10}, + {0,0,4,10}}; /* (x-1)%6 */ static const int mod6m1[8] = {5,0,1,2,3,4,5,0}; @@ -53,1108 +53,1114 @@ static void refine_subpel( x264_t *h, x264_me_t *m, int hpel_iters, int qpel_iters, int *p_halfpel_thresh, int b_refine_qpel ); -#define BITS_MVD( mx, my )\ - (p_cost_mvx[(mx)<<2] + p_cost_mvy[(my)<<2]) +#define BITS_MVD( mx, my ) \ + (p_cost_mvx[(mx)<<2] + p_cost_mvy[(my)<<2]) -#define COST_MV( mx, my )\ -{\ - int cost = h->pixf.fpelcmp[i_pixel]( p_fenc, FENC_STRIDE,\ - &p_fref[(my)*stride+(mx)], stride )\ - + BITS_MVD(mx,my);\ - COPY3_IF_LT( bcost, cost, bmx, mx, bmy, my );\ -} - -#define COST_MV_HPEL( mx, my ) \ -{ \ - int stride2 = 16; \ +#define COST_MV( mx, my ) \ + { \ + int cost = h->pixf.fpelcmp[i_pixel]( p_fenc, FENC_STRIDE, \ + &p_fref[(my)*stride+(mx)], stride ) \ + + BITS_MVD(mx,my); \ + COPY3_IF_LT( bcost, cost, bmx, mx, bmy, my ); \ + } + +#define COST_MV_HPEL( mx, my ) \ + { \ + int stride2 = 16; \ uint8_t *src = h->mc.get_ref( pix, &stride2, m->p_fref, stride, mx, my, bw, bh ); \ int cost = h->pixf.fpelcmp[i_pixel]( p_fenc, FENC_STRIDE, src, stride2 ) \ - + p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \ - COPY3_IF_LT( bpred_cost, cost, bpred_mx, mx, bpred_my, my ); \ -} - -#define COST_MV_X3_DIR( m0x, m0y, m1x, m1y, m2x, m2y, costs )\ -{\ - uint8_t *pix_base = p_fref + bmx + bmy*stride;\ - h->pixf.fpelcmp_x3[i_pixel]( p_fenc,\ - pix_base + (m0x) + (m0y)*stride,\ - pix_base + (m1x) + (m1y)*stride,\ - pix_base + (m2x) + (m2y)*stride,\ - stride, costs );\ - (costs)[0] += BITS_MVD( bmx+(m0x), bmy+(m0y) );\ - (costs)[1] += BITS_MVD( bmx+(m1x), bmy+(m1y) );\ - (costs)[2] += BITS_MVD( bmx+(m2x), bmy+(m2y) );\ -} - -#define COST_MV_X4_DIR( m0x, m0y, m1x, m1y, m2x, m2y, m3x, m3y, costs )\ -{\ - uint8_t *pix_base = p_fref + bmx + bmy*stride;\ - h->pixf.fpelcmp_x4[i_pixel]( p_fenc,\ - pix_base + (m0x) + (m0y)*stride,\ - pix_base + (m1x) + (m1y)*stride,\ - pix_base + (m2x) + (m2y)*stride,\ - pix_base + (m3x) + (m3y)*stride,\ - stride, costs );\ - (costs)[0] += BITS_MVD( bmx+(m0x), bmy+(m0y) );\ - (costs)[1] += BITS_MVD( bmx+(m1x), bmy+(m1y) );\ - (costs)[2] += BITS_MVD( bmx+(m2x), bmy+(m2y) );\ - (costs)[3] += BITS_MVD( bmx+(m3x), bmy+(m3y) );\ -} - -#define COST_MV_X4( m0x, m0y, m1x, m1y, m2x, m2y, m3x, m3y )\ -{\ - uint8_t *pix_base = p_fref + omx + omy*stride;\ - h->pixf.fpelcmp_x4[i_pixel]( p_fenc,\ - pix_base + (m0x) + (m0y)*stride,\ - pix_base + (m1x) + (m1y)*stride,\ - pix_base + (m2x) + (m2y)*stride,\ - pix_base + (m3x) + (m3y)*stride,\ - stride, costs );\ - costs[0] += BITS_MVD( omx+(m0x), omy+(m0y) );\ - costs[1] += BITS_MVD( omx+(m1x), omy+(m1y) );\ - costs[2] += BITS_MVD( omx+(m2x), omy+(m2y) );\ - costs[3] += BITS_MVD( omx+(m3x), omy+(m3y) );\ - COPY3_IF_LT( bcost, costs[0], bmx, omx+(m0x), bmy, omy+(m0y) );\ - COPY3_IF_LT( bcost, costs[1], bmx, omx+(m1x), bmy, omy+(m1y) );\ - COPY3_IF_LT( bcost, costs[2], bmx, omx+(m2x), bmy, omy+(m2y) );\ - COPY3_IF_LT( bcost, costs[3], bmx, omx+(m3x), bmy, omy+(m3y) );\ -} - -#define COST_MV_X3_ABS( m0x, m0y, m1x, m1y, m2x, m2y )\ -{\ - h->pixf.fpelcmp_x3[i_pixel]( p_fenc,\ - p_fref + (m0x) + (m0y)*stride,\ - p_fref + (m1x) + (m1y)*stride,\ - p_fref + (m2x) + (m2y)*stride,\ - stride, costs );\ - costs[0] += p_cost_mvx[(m0x)<<2]; /* no cost_mvy */\ - costs[1] += p_cost_mvx[(m1x)<<2];\ - costs[2] += p_cost_mvx[(m2x)<<2];\ - COPY3_IF_LT( bcost, costs[0], bmx, m0x, bmy, m0y );\ - COPY3_IF_LT( bcost, costs[1], bmx, m1x, bmy, m1y );\ - COPY3_IF_LT( bcost, costs[2], bmx, m2x, bmy, m2y );\ -} + + p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \ + COPY3_IF_LT( bpred_cost, cost, bpred_mx, mx, bpred_my, my ); \ + } + +#define COST_MV_X3_DIR( m0x, m0y, m1x, m1y, m2x, m2y, costs ) \ + { \ + uint8_t *pix_base = p_fref + bmx + bmy*stride; \ + h->pixf.fpelcmp_x3[i_pixel]( p_fenc, \ + pix_base + (m0x) + (m0y)*stride, \ + pix_base + (m1x) + (m1y)*stride, \ + pix_base + (m2x) + (m2y)*stride, \ + stride, costs ); \ + (costs)[0] += BITS_MVD( bmx+(m0x), bmy+(m0y) ); \ + (costs)[1] += BITS_MVD( bmx+(m1x), bmy+(m1y) ); \ + (costs)[2] += BITS_MVD( bmx+(m2x), bmy+(m2y) ); \ + } + +#define COST_MV_X4_DIR( m0x, m0y, m1x, m1y, m2x, m2y, m3x, m3y, costs ) \ + { \ + uint8_t *pix_base = p_fref + bmx + bmy*stride; \ + h->pixf.fpelcmp_x4[i_pixel]( p_fenc, \ + pix_base + (m0x) + (m0y)*stride, \ + pix_base + (m1x) + (m1y)*stride, \ + pix_base + (m2x) + (m2y)*stride, \ + pix_base + (m3x) + (m3y)*stride, \ + stride, costs ); \ + (costs)[0] += BITS_MVD( bmx+(m0x), bmy+(m0y) ); \ + (costs)[1] += BITS_MVD( bmx+(m1x), bmy+(m1y) ); \ + (costs)[2] += BITS_MVD( bmx+(m2x), bmy+(m2y) ); \ + (costs)[3] += BITS_MVD( bmx+(m3x), bmy+(m3y) ); \ + } + +#define COST_MV_X4( m0x, m0y, m1x, m1y, m2x, m2y, m3x, m3y ) \ + { \ + uint8_t *pix_base = p_fref + omx + omy*stride; \ + h->pixf.fpelcmp_x4[i_pixel]( p_fenc, \ + pix_base + (m0x) + (m0y)*stride, \ + pix_base + (m1x) + (m1y)*stride, \ + pix_base + (m2x) + (m2y)*stride, \ + pix_base + (m3x) + (m3y)*stride, \ + stride, costs ); \ + costs[0] += BITS_MVD( omx+(m0x), omy+(m0y) ); \ + costs[1] += BITS_MVD( omx+(m1x), omy+(m1y) ); \ + costs[2] += BITS_MVD( omx+(m2x), omy+(m2y) ); \ + costs[3] += BITS_MVD( omx+(m3x), omy+(m3y) ); \ + COPY3_IF_LT( bcost, costs[0], bmx, omx+(m0x), bmy, omy+(m0y) ); \ + COPY3_IF_LT( bcost, costs[1], bmx, omx+(m1x), bmy, omy+(m1y) ); \ + COPY3_IF_LT( bcost, costs[2], bmx, omx+(m2x), bmy, omy+(m2y) ); \ + COPY3_IF_LT( bcost, costs[3], bmx, omx+(m3x), bmy, omy+(m3y) ); \ + } + +#define COST_MV_X3_ABS( m0x, m0y, m1x, m1y, m2x, m2y ) \ + { \ + h->pixf.fpelcmp_x3[i_pixel]( p_fenc, \ + p_fref + (m0x) + (m0y)*stride, \ + p_fref + (m1x) + (m1y)*stride, \ + p_fref + (m2x) + (m2y)*stride, \ + stride, costs ); \ + costs[0] += p_cost_mvx[(m0x)<<2]; /* no cost_mvy */ \ + costs[1] += p_cost_mvx[(m1x)<<2]; \ + costs[2] += p_cost_mvx[(m2x)<<2]; \ + COPY3_IF_LT( bcost, costs[0], bmx, m0x, bmy, m0y ); \ + COPY3_IF_LT( bcost, costs[1], bmx, m1x, bmy, m1y ); \ + COPY3_IF_LT( bcost, costs[2], bmx, m2x, bmy, m2y ); \ + } /* 1 */ /* 101 */ /* 1 */ -#define DIA1_ITER( mx, my )\ -{\ - omx = mx; omy = my;\ - COST_MV_X4( 0,-1, 0,1, -1,0, 1,0 );\ -} - -#define CROSS( start, x_max, y_max )\ -{\ - i = start;\ - if( x_max <= X264_MIN(mv_x_max-omx, omx-mv_x_min) )\ - for( ; i < x_max-2; i+=4 )\ - COST_MV_X4( i,0, -i,0, i+2,0, -i-2,0 );\ - for( ; i < x_max; i+=2 )\ - {\ - if( omx+i <= mv_x_max )\ - COST_MV( omx+i, omy );\ - if( omx-i >= mv_x_min )\ - COST_MV( omx-i, omy );\ - }\ - i = start;\ - if( y_max <= X264_MIN(mv_y_max-omy, omy-mv_y_min) )\ - for( ; i < y_max-2; i+=4 )\ - COST_MV_X4( 0,i, 0,-i, 0,i+2, 0,-i-2 );\ - for( ; i < y_max; i+=2 )\ - {\ - if( omy+i <= mv_y_max )\ - COST_MV( omx, omy+i );\ - if( omy-i >= mv_y_min )\ - COST_MV( omx, omy-i );\ - }\ -} +#define DIA1_ITER( mx, my ) \ + { \ + omx = mx; omy = my; \ + COST_MV_X4( 0,-1, 0,1, -1,0, 1,0 ); \ + } + +#define CROSS( start, x_max, y_max ) \ + { \ + i = start; \ + if( x_max <= X264_MIN(mv_x_max-omx, omx-mv_x_min) ) \ + for( ; i < x_max-2; i+=4 ) \ + COST_MV_X4( i,0, -i,0, i+2,0, -i-2,0 ); \ + for( ; i < x_max; i+=2 ) \ + { \ + if( omx+i <= mv_x_max ) \ + COST_MV( omx+i, omy ); \ + if( omx-i >= mv_x_min ) \ + COST_MV( omx-i, omy ); \ + } \ + i = start; \ + if( y_max <= X264_MIN(mv_y_max-omy, omy-mv_y_min) ) \ + for( ; i < y_max-2; i+=4 ) \ + COST_MV_X4( 0,i, 0,-i, 0,i+2, 0,-i-2 ); \ + for( ; i < y_max; i+=2 ) \ + { \ + if( omy+i <= mv_y_max ) \ + COST_MV( omx, omy+i ); \ + if( omy-i >= mv_y_min ) \ + COST_MV( omx, omy-i ); \ + } \ + } void x264_me_search_ref( x264_t *h, x264_me_t *m, int16_t (*mvc)[2], int i_mvc, int *p_halfpel_thresh ) { - const int bw = x264_pixel_size[m->i_pixel].w; - const int bh = x264_pixel_size[m->i_pixel].h; - const int i_pixel = m->i_pixel; - const int stride = m->i_stride[0]; - int i_me_range = h->param.analyse.i_me_range; - int bmx, bmy, bcost; - int bpred_mx = 0, bpred_my = 0, bpred_cost = COST_MAX; - int omx, omy, pmx, pmy; - uint8_t *p_fenc = m->p_fenc[0]; - uint8_t *p_fref = m->p_fref[0]; - ALIGNED_ARRAY_16( uint8_t, pix,[16*16] ); - - int i, j; - int dir; - int costs[16]; - - int mv_x_min = h->mb.mv_min_fpel[0]; - int mv_y_min = h->mb.mv_min_fpel[1]; - int mv_x_max = h->mb.mv_max_fpel[0]; - int mv_y_max = h->mb.mv_max_fpel[1]; + const int bw = x264_pixel_size[m->i_pixel].w; // x264_pixel_size[] in common/pixel.h + const int bh = x264_pixel_size[m->i_pixel].h; + const int i_pixel = m->i_pixel; // macroblock pixel size + const int stride = m->i_stride[0]; // array stride (usually used only in Y direction) + int i_me_range = h->param.analyse.i_me_range; /* integer pixel motion estimation search range (from predicted mv) */ + int bmx, bmy, bcost; + int bpred_mx = 0, bpred_my = 0, bpred_cost = COST_MAX; + int omx, omy, pmx, pmy; + uint8_t *p_fenc = m->p_fenc[0]; + uint8_t *p_fref = m->p_fref[0]; + //V(fprintf(stderr, "s264_me_search_ref *p_fenc = 0x%.8x *p_fref = 0x%.8x\n", (unsigned long) p_fenc, (unsigned long) p_fref)); + ALIGNED_ARRAY_16( uint8_t, pix,[16*16] ); + + int i, j; + int dir; + int costs[16]; + + //fprintf(stderr, "x264_me_search_ref: x264_t *h = 0x%.8x\n", (unsigned int) h); + /* Ranges for MVs - perhaps we're on/near the frame boarder and can't go beyond it, these ranges indicate that. They're set in x264_mb_analyse_init() */ + int mv_x_min = h->mb.mv_min_fpel[0]; + int mv_y_min = h->mb.mv_min_fpel[1]; + int mv_x_max = h->mb.mv_max_fpel[0]; + int mv_y_max = h->mb.mv_max_fpel[1]; #define CHECK_MVRANGE(mx,my) ( mx >= mv_x_min && mx <= mv_x_max && my >= mv_y_min && my <= mv_y_max ) - const uint16_t *p_cost_mvx = m->p_cost_mv - m->mvp[0]; - const uint16_t *p_cost_mvy = m->p_cost_mv - m->mvp[1]; - - bmx = x264_clip3( m->mvp[0], mv_x_min*4, mv_x_max*4 ); - bmy = x264_clip3( m->mvp[1], mv_y_min*4, mv_y_max*4 ); - pmx = ( bmx + 2 ) >> 2; - pmy = ( bmy + 2 ) >> 2; - bcost = COST_MAX; + const uint16_t *p_cost_mvx = m->p_cost_mv - m->mvp[0]; + const uint16_t *p_cost_mvy = m->p_cost_mv - m->mvp[1]; + //V(fprintf(stderr, "p_cost_mvx: 0x%.8x, ( m->p_cost_mv: 0x%.8x m->mvp[0]: 0x%.8x )p_cost_mvy: 0x%.8x (m->p_cost_mv: 0x%.8x m->mvp[1]: 0x%.8x)\n", (unsigned long) p_cost_mvx, (unsigned long) m->p_cost_mv, (unsigned long) m->mvp[0], (unsigned long) p_cost_mvy, (unsigned long) m->p_cost_mv, (unsigned long) m->mvp[1])); + + /* m->mvp has the media MV (x in [0] and y in [1]) */ + bmx = x264_clip3( m->mvp[0], mv_x_min*4, mv_x_max*4 ); + bmy = x264_clip3( m->mvp[1], mv_y_min*4, mv_y_max*4 ); + pmx = ( bmx + 2 ) >> 2; + pmy = ( bmy + 2 ) >> 2; + bcost = COST_MAX; - /* try extra predictors if provided */ - if( h->mb.i_subpel_refine >= 3 ) + /* try extra predictors if provided */ + if( h->mb.i_subpel_refine >= 3 ) { - uint32_t bmv = pack16to32_mask(bmx,bmy); - COST_MV_HPEL( bmx, bmy ); - for( i = 0; i < i_mvc; i++ ) + uint32_t bmv = pack16to32_mask(bmx,bmy); + COST_MV_HPEL( bmx, bmy ); + for( i = 0; i < i_mvc; i++ ) { - if( *(uint32_t*)mvc[i] && (bmv - *(uint32_t*)mvc[i]) ) + if( *(uint32_t*)mvc[i] && (bmv - *(uint32_t*)mvc[i]) ) { - int mx = x264_clip3( mvc[i][0], mv_x_min*4, mv_x_max*4 ); - int my = x264_clip3( mvc[i][1], mv_y_min*4, mv_y_max*4 ); - COST_MV_HPEL( mx, my ); + int mx = x264_clip3( mvc[i][0], mv_x_min*4, mv_x_max*4 ); + int my = x264_clip3( mvc[i][1], mv_y_min*4, mv_y_max*4 ); + COST_MV_HPEL( mx, my ); } } - bmx = ( bpred_mx + 2 ) >> 2; - bmy = ( bpred_my + 2 ) >> 2; - COST_MV( bmx, bmy ); + bmx = ( bpred_mx + 2 ) >> 2; + bmy = ( bpred_my + 2 ) >> 2; + COST_MV( bmx, bmy ); } - else + else { - /* check the MVP */ - COST_MV( pmx, pmy ); - /* Because we are rounding the predicted motion vector to fullpel, there will be - * an extra MV cost in 15 out of 16 cases. However, when the predicted MV is - * chosen as the best predictor, it is often the case that the subpel search will - * result in a vector at or next to the predicted motion vector. Therefore, it is - * sensible to remove the cost of the MV from the rounded MVP to avoid unfairly - * biasing against use of the predicted motion vector. */ - bcost -= BITS_MVD( pmx, pmy ); - for( i = 0; i < i_mvc; i++ ) + /* check the MVP */ + COST_MV( pmx, pmy ); + /* Because we are rounding the predicted motion vector to fullpel, there will be + * an extra MV cost in 15 out of 16 cases. However, when the predicted MV is + * chosen as the best predictor, it is often the case that the subpel search will + * result in a vector at or next to the predicted motion vector. Therefore, it is + * sensible to remove the cost of the MV from the rounded MVP to avoid unfairly + * biasing against use of the predicted motion vector. */ + bcost -= BITS_MVD( pmx, pmy ); + for( i = 0; i < i_mvc; i++ ) { - int mx = (mvc[i][0] + 2) >> 2; - int my = (mvc[i][1] + 2) >> 2; - if( (mx | my) && ((mx-bmx) | (my-bmy)) ) + int mx = (mvc[i][0] + 2) >> 2; + int my = (mvc[i][1] + 2) >> 2; + if( (mx | my) && ((mx-bmx) | (my-bmy)) ) { - mx = x264_clip3( mx, mv_x_min, mv_x_max ); - my = x264_clip3( my, mv_y_min, mv_y_max ); - COST_MV( mx, my ); + mx = x264_clip3( mx, mv_x_min, mv_x_max ); + my = x264_clip3( my, mv_y_min, mv_y_max ); + COST_MV( mx, my ); } } } - COST_MV( 0, 0 ); + COST_MV( 0, 0 ); - switch( h->mb.i_me_method ) + switch( h->mb.i_me_method ) { case X264_ME_DIA: - /* diamond search, radius 1 */ - i = 0; - bcost <<= 4; - do + /* diamond search, radius 1 */ + //V(fprintf(stderr, "x264_me_search_ref; diamond search. *p_fenc = 0x%.8x *p_fref = 0x%.8x\n", (unsigned long) p_fenc, (unsigned long) p_fref)); + i = 0; + bcost <<= 4; + do { - COST_MV_X4_DIR( 0,-1, 0,1, -1,0, 1,0, costs ); - COPY1_IF_LT( bcost, (costs[0]<<4)+1 ); - COPY1_IF_LT( bcost, (costs[1]<<4)+3 ); - COPY1_IF_LT( bcost, (costs[2]<<4)+4 ); - COPY1_IF_LT( bcost, (costs[3]<<4)+12 ); - if( !(bcost&15) ) - break; - bmx -= (bcost<<28)>>30; - bmy -= (bcost<<30)>>30; - bcost &= ~15; - if( !CHECK_MVRANGE(bmx, bmy) ) - break; + COST_MV_X4_DIR( 0,-1, 0,1, -1,0, 1,0, costs ); + COPY1_IF_LT( bcost, (costs[0]<<4)+1 ); + COPY1_IF_LT( bcost, (costs[1]<<4)+3 ); + COPY1_IF_LT( bcost, (costs[2]<<4)+4 ); + COPY1_IF_LT( bcost, (costs[3]<<4)+12 ); + if( !(bcost&15) ) + break; + bmx -= (bcost<<28)>>30; + bmy -= (bcost<<30)>>30; + bcost &= ~15; + if( !CHECK_MVRANGE(bmx, bmy) ) + break; } while( ++i < i_me_range ); - bcost >>= 4; - break; + bcost >>= 4; + break; case X264_ME_HEX: -me_hex2: - /* hexagon search, radius 2 */ + me_hex2: + /* hexagon search, radius 2 */ #if 0 - for( i = 0; i < i_me_range/2; i++ ) + for( i = 0; i < i_me_range/2; i++ ) { - omx = bmx; omy = bmy; - COST_MV( omx-2, omy ); - COST_MV( omx-1, omy+2 ); - COST_MV( omx+1, omy+2 ); - COST_MV( omx+2, omy ); - COST_MV( omx+1, omy-2 ); - COST_MV( omx-1, omy-2 ); - if( bmx == omx && bmy == omy ) - break; - if( !CHECK_MVRANGE(bmx, bmy) ) - break; + omx = bmx; omy = bmy; + COST_MV( omx-2, omy ); + COST_MV( omx-1, omy+2 ); + COST_MV( omx+1, omy+2 ); + COST_MV( omx+2, omy ); + COST_MV( omx+1, omy-2 ); + COST_MV( omx-1, omy-2 ); + if( bmx == omx && bmy == omy ) + break; + if( !CHECK_MVRANGE(bmx, bmy) ) + break; } #else - /* equivalent to the above, but eliminates duplicate candidates */ + /* equivalent to the above, but eliminates duplicate candidates */ - /* hexagon */ - COST_MV_X3_DIR( -2,0, -1, 2, 1, 2, costs ); - COST_MV_X3_DIR( 2,0, 1,-2, -1,-2, costs+3 ); - bcost <<= 3; - COPY1_IF_LT( bcost, (costs[0]<<3)+2 ); - COPY1_IF_LT( bcost, (costs[1]<<3)+3 ); - COPY1_IF_LT( bcost, (costs[2]<<3)+4 ); - COPY1_IF_LT( bcost, (costs[3]<<3)+5 ); - COPY1_IF_LT( bcost, (costs[4]<<3)+6 ); - COPY1_IF_LT( bcost, (costs[5]<<3)+7 ); + /* hexagon */ + COST_MV_X3_DIR( -2,0, -1, 2, 1, 2, costs ); + COST_MV_X3_DIR( 2,0, 1,-2, -1,-2, costs+3 ); + bcost <<= 3; + COPY1_IF_LT( bcost, (costs[0]<<3)+2 ); + COPY1_IF_LT( bcost, (costs[1]<<3)+3 ); + COPY1_IF_LT( bcost, (costs[2]<<3)+4 ); + COPY1_IF_LT( bcost, (costs[3]<<3)+5 ); + COPY1_IF_LT( bcost, (costs[4]<<3)+6 ); + COPY1_IF_LT( bcost, (costs[5]<<3)+7 ); - if( bcost&7 ) + if( bcost&7 ) { - dir = (bcost&7)-2; - bmx += hex2[dir+1][0]; - bmy += hex2[dir+1][1]; - /* half hexagon, not overlapping the previous iteration */ - for( i = 1; i < i_me_range/2 && CHECK_MVRANGE(bmx, bmy); i++ ) + dir = (bcost&7)-2; + bmx += hex2[dir+1][0]; + bmy += hex2[dir+1][1]; + /* half hexagon, not overlapping the previous iteration */ + for( i = 1; i < i_me_range/2 && CHECK_MVRANGE(bmx, bmy); i++ ) { - COST_MV_X3_DIR( hex2[dir+0][0], hex2[dir+0][1], - hex2[dir+1][0], hex2[dir+1][1], - hex2[dir+2][0], hex2[dir+2][1], - costs ); - bcost &= ~7; - COPY1_IF_LT( bcost, (costs[0]<<3)+1 ); - COPY1_IF_LT( bcost, (costs[1]<<3)+2 ); - COPY1_IF_LT( bcost, (costs[2]<<3)+3 ); - if( !(bcost&7) ) - break; - dir += (bcost&7)-2; - dir = mod6m1[dir+1]; - bmx += hex2[dir+1][0]; - bmy += hex2[dir+1][1]; + COST_MV_X3_DIR( hex2[dir+0][0], hex2[dir+0][1], + hex2[dir+1][0], hex2[dir+1][1], + hex2[dir+2][0], hex2[dir+2][1], + costs ); + bcost &= ~7; + COPY1_IF_LT( bcost, (costs[0]<<3)+1 ); + COPY1_IF_LT( bcost, (costs[1]<<3)+2 ); + COPY1_IF_LT( bcost, (costs[2]<<3)+3 ); + if( !(bcost&7) ) + break; + dir += (bcost&7)-2; + dir = mod6m1[dir+1]; + bmx += hex2[dir+1][0]; + bmy += hex2[dir+1][1]; } } - bcost >>= 3; + bcost >>= 3; #endif - /* square refine */ - dir = 0; - COST_MV_X4_DIR( 0,-1, 0,1, -1,0, 1,0, costs ); - COPY2_IF_LT( bcost, costs[0], dir, 1 ); - COPY2_IF_LT( bcost, costs[1], dir, 2 ); - COPY2_IF_LT( bcost, costs[2], dir, 3 ); - COPY2_IF_LT( bcost, costs[3], dir, 4 ); - COST_MV_X4_DIR( -1,-1, -1,1, 1,-1, 1,1, costs ); - COPY2_IF_LT( bcost, costs[0], dir, 5 ); - COPY2_IF_LT( bcost, costs[1], dir, 6 ); - COPY2_IF_LT( bcost, costs[2], dir, 7 ); - COPY2_IF_LT( bcost, costs[3], dir, 8 ); - bmx += square1[dir][0]; - bmy += square1[dir][1]; - break; + /* square refine */ + dir = 0; + COST_MV_X4_DIR( 0,-1, 0,1, -1,0, 1,0, costs ); + COPY2_IF_LT( bcost, costs[0], dir, 1 ); + COPY2_IF_LT( bcost, costs[1], dir, 2 ); + COPY2_IF_LT( bcost, costs[2], dir, 3 ); + COPY2_IF_LT( bcost, costs[3], dir, 4 ); + COST_MV_X4_DIR( -1,-1, -1,1, 1,-1, 1,1, costs ); + COPY2_IF_LT( bcost, costs[0], dir, 5 ); + COPY2_IF_LT( bcost, costs[1], dir, 6 ); + COPY2_IF_LT( bcost, costs[2], dir, 7 ); + COPY2_IF_LT( bcost, costs[3], dir, 8 ); + bmx += square1[dir][0]; + bmy += square1[dir][1]; + break; case X264_ME_UMH: - { - /* Uneven-cross Multi-Hexagon-grid Search - * as in JM, except with different early termination */ - - static const int x264_pixel_size_shift[7] = { 0, 1, 1, 2, 3, 3, 4 }; - - int ucost1, ucost2; - int cross_start = 1; + { + /* Uneven-cross Multi-Hexagon-grid Search + * as in JM, except with different early termination */ + + static const int x264_pixel_size_shift[7] = { 0, 1, 1, 2, 3, 3, 4 }; + + int ucost1, ucost2; + int cross_start = 1; + + /* refine predictors */ + ucost1 = bcost; + DIA1_ITER( pmx, pmy ); + if( pmx | pmy ) + DIA1_ITER( 0, 0 ); + + if(i_pixel == PIXEL_4x4) + goto me_hex2; + + ucost2 = bcost; + if( (bmx | bmy) && ((bmx-pmx) | (bmy-pmy)) ) + DIA1_ITER( bmx, bmy ); + if( bcost == ucost2 ) + cross_start = 3; + omx = bmx; omy = bmy; - /* refine predictors */ - ucost1 = bcost; - DIA1_ITER( pmx, pmy ); - if( pmx | pmy ) - DIA1_ITER( 0, 0 ); - - if(i_pixel == PIXEL_4x4) - goto me_hex2; - - ucost2 = bcost; - if( (bmx | bmy) && ((bmx-pmx) | (bmy-pmy)) ) - DIA1_ITER( bmx, bmy ); - if( bcost == ucost2 ) - cross_start = 3; - omx = bmx; omy = bmy; - - /* early termination */ + /* early termination */ #define SAD_THRESH(v) ( bcost < ( v >> x264_pixel_size_shift[i_pixel] ) ) - if( bcost == ucost2 && SAD_THRESH(2000) ) - { - COST_MV_X4( 0,-2, -1,-1, 1,-1, -2,0 ); - COST_MV_X4( 2, 0, -1, 1, 1, 1, 0,2 ); - if( bcost == ucost1 && SAD_THRESH(500) ) - break; - if( bcost == ucost2 ) - { - int range = (i_me_range>>1) | 1; - CROSS( 3, range, range ); - COST_MV_X4( -1,-2, 1,-2, -2,-1, 2,-1 ); - COST_MV_X4( -2, 1, 2, 1, -1, 2, 1, 2 ); - if( bcost == ucost2 ) - break; - cross_start = range + 2; - } - } - - /* adaptive search range */ - if( i_mvc ) - { - /* range multipliers based on casual inspection of some statistics of - * average distance between current predictor and final mv found by ESA. - * these have not been tuned much by actual encoding. */ - static const int range_mul[4][4] = - { - { 3, 3, 4, 4 }, - { 3, 4, 4, 4 }, - { 4, 4, 4, 5 }, - { 4, 4, 5, 6 }, - }; - int mvd; - int sad_ctx, mvd_ctx; - int denom = 1; - - if( i_mvc == 1 ) - { - if( i_pixel == PIXEL_16x16 ) - /* mvc is probably the same as mvp, so the difference isn't meaningful. - * but prediction usually isn't too bad, so just use medium range */ - mvd = 25; - else - mvd = abs( m->mvp[0] - mvc[0][0] ) - + abs( m->mvp[1] - mvc[0][1] ); - } - else - { - /* calculate the degree of agreement between predictors. */ - /* in 16x16, mvc includes all the neighbors used to make mvp, - * so don't count mvp separately. */ - denom = i_mvc - 1; - mvd = 0; - if( i_pixel != PIXEL_16x16 ) - { - mvd = abs( m->mvp[0] - mvc[0][0] ) - + abs( m->mvp[1] - mvc[0][1] ); - denom++; - } - mvd += x264_predictor_difference( mvc, i_mvc ); - } - - sad_ctx = SAD_THRESH(1000) ? 0 - : SAD_THRESH(2000) ? 1 - : SAD_THRESH(4000) ? 2 : 3; - mvd_ctx = mvd < 10*denom ? 0 - : mvd < 20*denom ? 1 - : mvd < 40*denom ? 2 : 3; - - i_me_range = i_me_range * range_mul[mvd_ctx][sad_ctx] / 4; - } - - /* FIXME if the above DIA2/OCT2/CROSS found a new mv, it has not updated omx/omy. - * we are still centered on the same place as the DIA2. is this desirable? */ - CROSS( cross_start, i_me_range, i_me_range/2 ); - - COST_MV_X4( -2,-2, -2,2, 2,-2, 2,2 ); - - /* hexagon grid */ - omx = bmx; omy = bmy; - const uint16_t *p_cost_omvx = p_cost_mvx + omx*4; - const uint16_t *p_cost_omvy = p_cost_mvy + omy*4; - i = 1; - do - { - static const int hex4[16][2] = { - { 0,-4}, { 0, 4}, {-2,-3}, { 2,-3}, - {-4,-2}, { 4,-2}, {-4,-1}, { 4,-1}, - {-4, 0}, { 4, 0}, {-4, 1}, { 4, 1}, - {-4, 2}, { 4, 2}, {-2, 3}, { 2, 3}, - }; - - if( 4*i > X264_MIN4( mv_x_max-omx, omx-mv_x_min, - mv_y_max-omy, omy-mv_y_min ) ) - { - for( j = 0; j < 16; j++ ) - { - int mx = omx + hex4[j][0]*i; - int my = omy + hex4[j][1]*i; - if( CHECK_MVRANGE(mx, my) ) - COST_MV( mx, my ); - } - } - else - { - int dir = 0; - uint8_t *pix_base = p_fref + omx + (omy-4*i)*stride; - int dy = i*stride; -#define SADS(k,x0,y0,x1,y1,x2,y2,x3,y3)\ - h->pixf.fpelcmp_x4[i_pixel]( p_fenc,\ - pix_base x0*i+(y0-2*k+4)*dy,\ - pix_base x1*i+(y1-2*k+4)*dy,\ - pix_base x2*i+(y2-2*k+4)*dy,\ - pix_base x3*i+(y3-2*k+4)*dy,\ - stride, costs+4*k );\ - pix_base += 2*dy; + if( bcost == ucost2 && SAD_THRESH(2000) ) + { + COST_MV_X4( 0,-2, -1,-1, 1,-1, -2,0 ); + COST_MV_X4( 2, 0, -1, 1, 1, 1, 0,2 ); + if( bcost == ucost1 && SAD_THRESH(500) ) + break; + if( bcost == ucost2 ) + { + int range = (i_me_range>>1) | 1; + CROSS( 3, range, range ); + COST_MV_X4( -1,-2, 1,-2, -2,-1, 2,-1 ); + COST_MV_X4( -2, 1, 2, 1, -1, 2, 1, 2 ); + if( bcost == ucost2 ) + break; + cross_start = range + 2; + } + } + + /* adaptive search range */ + if( i_mvc ) + { + /* range multipliers based on casual inspection of some statistics of + * average distance between current predictor and final mv found by ESA. + * these have not been tuned much by actual encoding. */ + static const int range_mul[4][4] = + { + { 3, 3, 4, 4 }, + { 3, 4, 4, 4 }, + { 4, 4, 4, 5 }, + { 4, 4, 5, 6 }, + }; + int mvd; + int sad_ctx, mvd_ctx; + int denom = 1; + + if( i_mvc == 1 ) + { + if( i_pixel == PIXEL_16x16 ) + /* mvc is probably the same as mvp, so the difference isn't meaningful. + * but prediction usually isn't too bad, so just use medium range */ + mvd = 25; + else + mvd = abs( m->mvp[0] - mvc[0][0] ) + + abs( m->mvp[1] - mvc[0][1] ); + } + else + { + /* calculate the degree of agreement between predictors. */ + /* in 16x16, mvc includes all the neighbors used to make mvp, + * so don't count mvp separately. */ + denom = i_mvc - 1; + mvd = 0; + if( i_pixel != PIXEL_16x16 ) + { + mvd = abs( m->mvp[0] - mvc[0][0] ) + + abs( m->mvp[1] - mvc[0][1] ); + denom++; + } + mvd += x264_predictor_difference( mvc, i_mvc ); + } + + sad_ctx = SAD_THRESH(1000) ? 0 + : SAD_THRESH(2000) ? 1 + : SAD_THRESH(4000) ? 2 : 3; + mvd_ctx = mvd < 10*denom ? 0 + : mvd < 20*denom ? 1 + : mvd < 40*denom ? 2 : 3; + + i_me_range = i_me_range * range_mul[mvd_ctx][sad_ctx] / 4; + } + + /* FIXME if the above DIA2/OCT2/CROSS found a new mv, it has not updated omx/omy. + * we are still centered on the same place as the DIA2. is this desirable? */ + CROSS( cross_start, i_me_range, i_me_range/2 ); + + COST_MV_X4( -2,-2, -2,2, 2,-2, 2,2 ); + + /* hexagon grid */ + omx = bmx; omy = bmy; + const uint16_t *p_cost_omvx = p_cost_mvx + omx*4; + const uint16_t *p_cost_omvy = p_cost_mvy + omy*4; + i = 1; + do + { + static const int hex4[16][2] = { + { 0,-4}, { 0, 4}, {-2,-3}, { 2,-3}, + {-4,-2}, { 4,-2}, {-4,-1}, { 4,-1}, + {-4, 0}, { 4, 0}, {-4, 1}, { 4, 1}, + {-4, 2}, { 4, 2}, {-2, 3}, { 2, 3}, + }; + + if( 4*i > X264_MIN4( mv_x_max-omx, omx-mv_x_min, + mv_y_max-omy, omy-mv_y_min ) ) + { + for( j = 0; j < 16; j++ ) + { + int mx = omx + hex4[j][0]*i; + int my = omy + hex4[j][1]*i; + if( CHECK_MVRANGE(mx, my) ) + COST_MV( mx, my ); + } + } + else + { + int dir = 0; + uint8_t *pix_base = p_fref + omx + (omy-4*i)*stride; + int dy = i*stride; +#define SADS(k,x0,y0,x1,y1,x2,y2,x3,y3) \ + h->pixf.fpelcmp_x4[i_pixel]( p_fenc, \ + pix_base x0*i+(y0-2*k+4)*dy, \ + pix_base x1*i+(y1-2*k+4)*dy, \ + pix_base x2*i+(y2-2*k+4)*dy, \ + pix_base x3*i+(y3-2*k+4)*dy, \ + stride, costs+4*k ); \ + pix_base += 2*dy; #define ADD_MVCOST(k,x,y) costs[k] += p_cost_omvx[x*4*i] + p_cost_omvy[y*4*i] #define MIN_MV(k,x,y) COPY2_IF_LT( bcost, costs[k], dir, x*16+(y&15) ) - SADS( 0, +0,-4, +0,+4, -2,-3, +2,-3 ); - SADS( 1, -4,-2, +4,-2, -4,-1, +4,-1 ); - SADS( 2, -4,+0, +4,+0, -4,+1, +4,+1 ); - SADS( 3, -4,+2, +4,+2, -2,+3, +2,+3 ); - ADD_MVCOST( 0, 0,-4 ); - ADD_MVCOST( 1, 0, 4 ); - ADD_MVCOST( 2,-2,-3 ); - ADD_MVCOST( 3, 2,-3 ); - ADD_MVCOST( 4,-4,-2 ); - ADD_MVCOST( 5, 4,-2 ); - ADD_MVCOST( 6,-4,-1 ); - ADD_MVCOST( 7, 4,-1 ); - ADD_MVCOST( 8,-4, 0 ); - ADD_MVCOST( 9, 4, 0 ); - ADD_MVCOST( 10,-4, 1 ); - ADD_MVCOST( 11, 4, 1 ); - ADD_MVCOST( 12,-4, 2 ); - ADD_MVCOST( 13, 4, 2 ); - ADD_MVCOST( 14,-2, 3 ); - ADD_MVCOST( 15, 2, 3 ); - MIN_MV( 0, 0,-4 ); - MIN_MV( 1, 0, 4 ); - MIN_MV( 2,-2,-3 ); - MIN_MV( 3, 2,-3 ); - MIN_MV( 4,-4,-2 ); - MIN_MV( 5, 4,-2 ); - MIN_MV( 6,-4,-1 ); - MIN_MV( 7, 4,-1 ); - MIN_MV( 8,-4, 0 ); - MIN_MV( 9, 4, 0 ); - MIN_MV( 10,-4, 1 ); - MIN_MV( 11, 4, 1 ); - MIN_MV( 12,-4, 2 ); - MIN_MV( 13, 4, 2 ); - MIN_MV( 14,-2, 3 ); - MIN_MV( 15, 2, 3 ); + SADS( 0, +0,-4, +0,+4, -2,-3, +2,-3 ); + SADS( 1, -4,-2, +4,-2, -4,-1, +4,-1 ); + SADS( 2, -4,+0, +4,+0, -4,+1, +4,+1 ); + SADS( 3, -4,+2, +4,+2, -2,+3, +2,+3 ); + ADD_MVCOST( 0, 0,-4 ); + ADD_MVCOST( 1, 0, 4 ); + ADD_MVCOST( 2,-2,-3 ); + ADD_MVCOST( 3, 2,-3 ); + ADD_MVCOST( 4,-4,-2 ); + ADD_MVCOST( 5, 4,-2 ); + ADD_MVCOST( 6,-4,-1 ); + ADD_MVCOST( 7, 4,-1 ); + ADD_MVCOST( 8,-4, 0 ); + ADD_MVCOST( 9, 4, 0 ); + ADD_MVCOST( 10,-4, 1 ); + ADD_MVCOST( 11, 4, 1 ); + ADD_MVCOST( 12,-4, 2 ); + ADD_MVCOST( 13, 4, 2 ); + ADD_MVCOST( 14,-2, 3 ); + ADD_MVCOST( 15, 2, 3 ); + MIN_MV( 0, 0,-4 ); + MIN_MV( 1, 0, 4 ); + MIN_MV( 2,-2,-3 ); + MIN_MV( 3, 2,-3 ); + MIN_MV( 4,-4,-2 ); + MIN_MV( 5, 4,-2 ); + MIN_MV( 6,-4,-1 ); + MIN_MV( 7, 4,-1 ); + MIN_MV( 8,-4, 0 ); + MIN_MV( 9, 4, 0 ); + MIN_MV( 10,-4, 1 ); + MIN_MV( 11, 4, 1 ); + MIN_MV( 12,-4, 2 ); + MIN_MV( 13, 4, 2 ); + MIN_MV( 14,-2, 3 ); + MIN_MV( 15, 2, 3 ); #undef SADS #undef ADD_MVCOST #undef MIN_MV - if(dir) - { - bmx = omx + i*(dir>>4); - bmy = omy + i*((dir<<28)>>28); - } - } - } while( ++i <= i_me_range/4 ); - if( bmy <= mv_y_max && bmy >= mv_y_min ) - goto me_hex2; - break; - } + if(dir) + { + bmx = omx + i*(dir>>4); + bmy = omy + i*((dir<<28)>>28); + } + } + } while( ++i <= i_me_range/4 ); + if( bmy <= mv_y_max && bmy >= mv_y_min ) + goto me_hex2; + break; + } case X264_ME_ESA: case X264_ME_TESA: - { - const int min_x = X264_MAX( bmx - i_me_range, mv_x_min ); - const int min_y = X264_MAX( bmy - i_me_range, mv_y_min ); - const int max_x = X264_MIN( bmx + i_me_range, mv_x_max ); - const int max_y = X264_MIN( bmy + i_me_range, mv_y_max ); - /* SEA is fastest in multiples of 4 */ - const int width = (max_x - min_x + 3) & ~3; - int my; + { + const int min_x = X264_MAX( bmx - i_me_range, mv_x_min ); + const int min_y = X264_MAX( bmy - i_me_range, mv_y_min ); + const int max_x = X264_MIN( bmx + i_me_range, mv_x_max ); + const int max_y = X264_MIN( bmy + i_me_range, mv_y_max ); + /* SEA is fastest in multiples of 4 */ + const int width = (max_x - min_x + 3) & ~3; + int my; #if 0 - /* plain old exhaustive search */ - int mx; - for( my = min_y; my <= max_y; my++ ) - for( mx = min_x; mx <= max_x; mx++ ) - COST_MV( mx, my ); + /* plain old exhaustive search */ + int mx; + for( my = min_y; my <= max_y; my++ ) + for( mx = min_x; mx <= max_x; mx++ ) + COST_MV( mx, my ); #else - /* successive elimination by comparing DC before a full SAD, - * because sum(abs(diff)) >= abs(diff(sum)). */ - uint16_t *sums_base = m->integral; - /* due to a GCC bug on some platforms (win32?), zero[] may not actually be aligned. - * this is not a problem because it is not used for any SSE instructions. */ - ALIGNED_16( static uint8_t zero[8*FENC_STRIDE] ); - ALIGNED_ARRAY_16( int, enc_dc,[4] ); - int sad_size = i_pixel <= PIXEL_8x8 ? PIXEL_8x8 : PIXEL_4x4; - int delta = x264_pixel_size[sad_size].w; - int16_t *xs = h->scratch_buffer; - int xn; - uint16_t *cost_fpel_mvx = h->cost_mv_fpel[x264_lambda_tab[h->mb.i_qp]][-m->mvp[0]&3] + (-m->mvp[0]>>2); - - h->pixf.sad_x4[sad_size]( zero, p_fenc, p_fenc+delta, - p_fenc+delta*FENC_STRIDE, p_fenc+delta+delta*FENC_STRIDE, - FENC_STRIDE, enc_dc ); - if( delta == 4 ) - sums_base += stride * (h->fenc->i_lines[0] + PADV*2); - if( i_pixel == PIXEL_16x16 || i_pixel == PIXEL_8x16 || i_pixel == PIXEL_4x8 ) - delta *= stride; - if( i_pixel == PIXEL_8x16 || i_pixel == PIXEL_4x8 ) - enc_dc[1] = enc_dc[2]; - - if( h->mb.i_me_method == X264_ME_TESA ) - { - // ADS threshold, then SAD threshold, then keep the best few SADs, then SATD - mvsad_t *mvsads = (mvsad_t *)(xs + ((width+15)&~15)); - int nmvsad = 0, limit; - int sad_thresh = i_me_range <= 16 ? 10 : i_me_range <= 24 ? 11 : 12; - int bsad = h->pixf.sad[i_pixel]( p_fenc, FENC_STRIDE, p_fref+bmy*stride+bmx, stride ) - + BITS_MVD( bmx, bmy ); - for( my = min_y; my <= max_y; my++ ) - { - int ycost = p_cost_mvy[my<<2]; - if( bsad <= ycost ) - continue; - bsad -= ycost; - xn = h->pixf.ads[i_pixel]( enc_dc, sums_base + min_x + my * stride, delta, - cost_fpel_mvx+min_x, xs, width, bsad*17/16 ); - for( i=0; i<xn-2; i+=3 ) - { - uint8_t *ref = p_fref+min_x+my*stride; - int sads[3]; - h->pixf.sad_x3[i_pixel]( p_fenc, ref+xs[i], ref+xs[i+1], ref+xs[i+2], stride, sads ); - for( j=0; j<3; j++ ) - { - int sad = sads[j] + cost_fpel_mvx[xs[i+j]]; - if( sad < bsad*sad_thresh>>3 ) - { - COPY1_IF_LT( bsad, sad ); - mvsads[nmvsad].sad = sad + ycost; - mvsads[nmvsad].mx = min_x+xs[i+j]; - mvsads[nmvsad].my = my; - nmvsad++; - } - } - } - for( ; i<xn; i++ ) - { - int mx = min_x+xs[i]; - int sad = h->pixf.sad[i_pixel]( p_fenc, FENC_STRIDE, p_fref+mx+my*stride, stride ) - + cost_fpel_mvx[xs[i]]; - if( sad < bsad*sad_thresh>>3 ) - { - COPY1_IF_LT( bsad, sad ); - mvsads[nmvsad].sad = sad + ycost; - mvsads[nmvsad].mx = mx; - mvsads[nmvsad].my = my; - nmvsad++; - } - } - bsad += ycost; - } - - limit = i_me_range / 2; - sad_thresh = bsad*sad_thresh>>3; - while( nmvsad > limit*2 && sad_thresh > bsad ) - { - // halve the range if the domain is too large... eh, close enough - sad_thresh = (sad_thresh + bsad) >> 1; - for( i=0; i<nmvsad && mvsads[i].sad <= sad_thresh; i++ ); - for( j=i; j<nmvsad; j++ ) - { - /* mvsad_t is not guaranteed to be 8 bytes on all archs, so check before using explicit write-combining */ - if( sizeof( mvsad_t ) == sizeof( uint64_t ) ) - *(uint64_t*)&mvsads[i] = *(uint64_t*)&mvsads[j]; - else - mvsads[i] = mvsads[j]; - i += mvsads[j].sad <= sad_thresh; - } - nmvsad = i; - } - while( nmvsad > limit ) - { - int bsad = mvsads[0].sad; - int bi = 0; - for( i=1; i<nmvsad; i++ ) - COPY2_IF_GT( bsad, mvsads[i].sad, bi, i ); - nmvsad--; - mvsads[bi] = mvsads[nmvsad]; - if( sizeof( mvsad_t ) == sizeof( uint64_t ) ) - *(uint64_t*)&mvsads[bi] = *(uint64_t*)&mvsads[nmvsad]; - else - mvsads[bi] = mvsads[nmvsad]; - } - for( i=0; i<nmvsad; i++ ) - COST_MV( mvsads[i].mx, mvsads[i].my ); - } - else - { - // just ADS and SAD - for( my = min_y; my <= max_y; my++ ) - { - int ycost = p_cost_mvy[my<<2]; - if( bcost <= ycost ) - continue; - bcost -= ycost; - xn = h->pixf.ads[i_pixel]( enc_dc, sums_base + min_x + my * stride, delta, - cost_fpel_mvx+min_x, xs, width, bcost ); - for( i=0; i<xn-2; i+=3 ) - COST_MV_X3_ABS( min_x+xs[i],my, min_x+xs[i+1],my, min_x+xs[i+2],my ); - bcost += ycost; - for( ; i<xn; i++ ) - COST_MV( min_x+xs[i], my ); - } - } + /* successive elimination by comparing DC before a full SAD, + * because sum(abs(diff)) >= abs(diff(sum)). */ + uint16_t *sums_base = m->integral; + /* due to a GCC bug on some platforms (win32?), zero[] may not actually be aligned. + * this is not a problem because it is not used for any SSE instructions. */ + ALIGNED_16( static uint8_t zero[8*FENC_STRIDE] ); + ALIGNED_ARRAY_16( int, enc_dc,[4] ); + int sad_size = i_pixel <= PIXEL_8x8 ? PIXEL_8x8 : PIXEL_4x4; + int delta = x264_pixel_size[sad_size].w; + int16_t *xs = h->scratch_buffer; + int xn; + uint16_t *cost_fpel_mvx = h->cost_mv_fpel[x264_lambda_tab[h->mb.i_qp]][-m->mvp[0]&3] + (-m->mvp[0]>>2); + + h->pixf.sad_x4[sad_size]( zero, p_fenc, p_fenc+delta, + p_fenc+delta*FENC_STRIDE, p_fenc+delta+delta*FENC_STRIDE, + FENC_STRIDE, enc_dc ); + if( delta == 4 ) + sums_base += stride * (h->fenc->i_lines[0] + PADV*2); + if( i_pixel == PIXEL_16x16 || i_pixel == PIXEL_8x16 || i_pixel == PIXEL_4x8 ) + delta *= stride; + if( i_pixel == PIXEL_8x16 || i_pixel == PIXEL_4x8 ) + enc_dc[1] = enc_dc[2]; + + if( h->mb.i_me_method == X264_ME_TESA ) + { + // ADS threshold, then SAD threshold, then keep the best few SADs, then SATD + mvsad_t *mvsads = (mvsad_t *)(xs + ((width+15)&~15)); + int nmvsad = 0, limit; + int sad_thresh = i_me_range <= 16 ? 10 : i_me_range <= 24 ? 11 : 12; + int bsad = h->pixf.sad[i_pixel]( p_fenc, FENC_STRIDE, p_fref+bmy*stride+bmx, stride ) + + BITS_MVD( bmx, bmy ); + for( my = min_y; my <= max_y; my++ ) + { + int ycost = p_cost_mvy[my<<2]; + if( bsad <= ycost ) + continue; + bsad -= ycost; + xn = h->pixf.ads[i_pixel]( enc_dc, sums_base + min_x + my * stride, delta, + cost_fpel_mvx+min_x, xs, width, bsad*17/16 ); + for( i=0; i<xn-2; i+=3 ) + { + uint8_t *ref = p_fref+min_x+my*stride; + int sads[3]; + h->pixf.sad_x3[i_pixel]( p_fenc, ref+xs[i], ref+xs[i+1], ref+xs[i+2], stride, sads ); + for( j=0; j<3; j++ ) + { + int sad = sads[j] + cost_fpel_mvx[xs[i+j]]; + if( sad < bsad*sad_thresh>>3 ) + { + COPY1_IF_LT( bsad, sad ); + mvsads[nmvsad].sad = sad + ycost; + mvsads[nmvsad].mx = min_x+xs[i+j]; + mvsads[nmvsad].my = my; + nmvsad++; + } + } + } + for( ; i<xn; i++ ) + { + int mx = min_x+xs[i]; + int sad = h->pixf.sad[i_pixel]( p_fenc, FENC_STRIDE, p_fref+mx+my*stride, stride ) + + cost_fpel_mvx[xs[i]]; + if( sad < bsad*sad_thresh>>3 ) + { + COPY1_IF_LT( bsad, sad ); + mvsads[nmvsad].sad = sad + ycost; + mvsads[nmvsad].mx = mx; + mvsads[nmvsad].my = my; + nmvsad++; + } + } + bsad += ycost; + } + + limit = i_me_range / 2; + sad_thresh = bsad*sad_thresh>>3; + while( nmvsad > limit*2 && sad_thresh > bsad ) + { + // halve the range if the domain is too large... eh, close enough + sad_thresh = (sad_thresh + bsad) >> 1; + for( i=0; i<nmvsad && mvsads[i].sad <= sad_thresh; i++ ); + for( j=i; j<nmvsad; j++ ) + { + /* mvsad_t is not guaranteed to be 8 bytes on all archs, so check before using explicit write-combining */ + if( sizeof( mvsad_t ) == sizeof( uint64_t ) ) + *(uint64_t*)&mvsads[i] = *(uint64_t*)&mvsads[j]; + else + mvsads[i] = mvsads[j]; + i += mvsads[j].sad <= sad_thresh; + } + nmvsad = i; + } + while( nmvsad > limit ) + { + int bsad = mvsads[0].sad; + int bi = 0; + for( i=1; i<nmvsad; i++ ) + COPY2_IF_GT( bsad, mvsads[i].sad, bi, i ); + nmvsad--; + mvsads[bi] = mvsads[nmvsad]; + if( sizeof( mvsad_t ) == sizeof( uint64_t ) ) + *(uint64_t*)&mvsads[bi] = *(uint64_t*)&mvsads[nmvsad]; + else + mvsads[bi] = mvsads[nmvsad]; + } + for( i=0; i<nmvsad; i++ ) + COST_MV( mvsads[i].mx, mvsads[i].my ); + } + else + { + // just ADS and SAD + for( my = min_y; my <= max_y; my++ ) + { + int ycost = p_cost_mvy[my<<2]; + if( bcost <= ycost ) + continue; + bcost -= ycost; + xn = h->pixf.ads[i_pixel]( enc_dc, sums_base + min_x + my * stride, delta, + cost_fpel_mvx+min_x, xs, width, bcost ); + for( i=0; i<xn-2; i+=3 ) + COST_MV_X3_ABS( min_x+xs[i],my, min_x+xs[i+1],my, min_x+xs[i+2],my ); + bcost += ycost; + for( ; i<xn; i++ ) + COST_MV( min_x+xs[i], my ); + } + } #endif - } - break; + } + break; } - /* -> qpel mv */ - if( bpred_cost < bcost ) + /* -> qpel mv */ + if( bpred_cost < bcost ) { - m->mv[0] = bpred_mx; - m->mv[1] = bpred_my; - m->cost = bpred_cost; + m->mv[0] = bpred_mx; + m->mv[1] = bpred_my; + m->cost = bpred_cost; } - else + else { - m->mv[0] = bmx << 2; - m->mv[1] = bmy << 2; - m->cost = bcost; + m->mv[0] = bmx << 2; + m->mv[1] = bmy << 2; + m->cost = bcost; } - /* compute the real cost */ - m->cost_mv = p_cost_mvx[ m->mv[0] ] + p_cost_mvy[ m->mv[1] ]; - if( bmx == pmx && bmy == pmy && h->mb.i_subpel_refine < 3 ) - m->cost += m->cost_mv; + /* compute the real cost */ + m->cost_mv = p_cost_mvx[ m->mv[0] ] + p_cost_mvy[ m->mv[1] ]; + if( bmx == pmx && bmy == pmy && h->mb.i_subpel_refine < 3 ) + m->cost += m->cost_mv; - /* subpel refine */ - if( h->mb.i_subpel_refine >= 2 ) + /* subpel refine */ + if( h->mb.i_subpel_refine >= 2 ) { - int hpel = subpel_iterations[h->mb.i_subpel_refine][2]; - int qpel = subpel_iterations[h->mb.i_subpel_refine][3]; - refine_subpel( h, m, hpel, qpel, p_halfpel_thresh, 0 ); + int hpel = subpel_iterations[h->mb.i_subpel_refine][2]; + int qpel = subpel_iterations[h->mb.i_subpel_refine][3]; + refine_subpel( h, m, hpel, qpel, p_halfpel_thresh, 0 ); } } #undef COST_MV void x264_me_refine_qpel( x264_t *h, x264_me_t *m ) { - int hpel = subpel_iterations[h->mb.i_subpel_refine][0]; - int qpel = subpel_iterations[h->mb.i_subpel_refine][1]; + int hpel = subpel_iterations[h->mb.i_subpel_refine][0]; + int qpel = subpel_iterations[h->mb.i_subpel_refine][1]; - if( m->i_pixel <= PIXEL_8x8 && h->sh.i_type == SLICE_TYPE_P ) - m->cost -= m->i_ref_cost; + if( m->i_pixel <= PIXEL_8x8 && h->sh.i_type == SLICE_TYPE_P ) + m->cost -= m->i_ref_cost; - refine_subpel( h, m, hpel, qpel, NULL, 1 ); + refine_subpel( h, m, hpel, qpel, NULL, 1 ); } -#define COST_MV_SAD( mx, my ) \ -{ \ - int stride = 16; \ +#define COST_MV_SAD( mx, my ) \ + { \ + int stride = 16; \ uint8_t *src = h->mc.get_ref( pix[0], &stride, m->p_fref, m->i_stride[0], mx, my, bw, bh ); \ int cost = h->pixf.fpelcmp[i_pixel]( m->p_fenc[0], FENC_STRIDE, src, stride ) \ - + p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \ - COPY3_IF_LT( bcost, cost, bmx, mx, bmy, my ); \ -} - -#define COST_MV_SATD( mx, my, dir ) \ -if( b_refine_qpel || (dir^1) != odir ) \ -{ \ - int stride = 16; \ - uint8_t *src = h->mc.get_ref( pix[0], &stride, m->p_fref, m->i_stride[0], mx, my, bw, bh ); \ - int cost = h->pixf.mbcmp_unaligned[i_pixel]( m->p_fenc[0], FENC_STRIDE, src, stride ) \ - + p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \ - if( b_chroma_me && cost < bcost ) \ - { \ - h->mc.mc_chroma( pix[0], 8, m->p_fref[4], m->i_stride[1], mx, my, bw/2, bh/2 ); \ - cost += h->pixf.mbcmp[i_pixel+3]( m->p_fenc[1], FENC_STRIDE, pix[0], 8 ); \ - if( cost < bcost ) \ - { \ - h->mc.mc_chroma( pix[0], 8, m->p_fref[5], m->i_stride[1], mx, my, bw/2, bh/2 ); \ - cost += h->pixf.mbcmp[i_pixel+3]( m->p_fenc[2], FENC_STRIDE, pix[0], 8 ); \ - } \ - } \ - if( cost < bcost ) \ - { \ - bcost = cost; \ - bmx = mx; \ - bmy = my; \ - bdir = dir; \ - } \ -} + + p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \ + COPY3_IF_LT( bcost, cost, bmx, mx, bmy, my ); \ + } + +#define COST_MV_SATD( mx, my, dir ) \ + if( b_refine_qpel || (dir^1) != odir ) \ + { \ + int stride = 16; \ + uint8_t *src = h->mc.get_ref( pix[0], &stride, m->p_fref, m->i_stride[0], mx, my, bw, bh ); \ + int cost = h->pixf.mbcmp_unaligned[i_pixel]( m->p_fenc[0], FENC_STRIDE, src, stride ) \ + + p_cost_mvx[ mx ] + p_cost_mvy[ my ]; \ + if( b_chroma_me && cost < bcost ) \ + { \ + h->mc.mc_chroma( pix[0], 8, m->p_fref[4], m->i_stride[1], mx, my, bw/2, bh/2 ); \ + cost += h->pixf.mbcmp[i_pixel+3]( m->p_fenc[1], FENC_STRIDE, pix[0], 8 ); \ + if( cost < bcost ) \ + { \ + h->mc.mc_chroma( pix[0], 8, m->p_fref[5], m->i_stride[1], mx, my, bw/2, bh/2 ); \ + cost += h->pixf.mbcmp[i_pixel+3]( m->p_fenc[2], FENC_STRIDE, pix[0], 8 ); \ + } \ + } \ + if( cost < bcost ) \ + { \ + bcost = cost; \ + bmx = mx; \ + bmy = my; \ + bdir = dir; \ + } \ + } static void refine_subpel( x264_t *h, x264_me_t *m, int hpel_iters, int qpel_iters, int *p_halfpel_thresh, int b_refine_qpel ) { - const int bw = x264_pixel_size[m->i_pixel].w; - const int bh = x264_pixel_size[m->i_pixel].h; - const uint16_t *p_cost_mvx = m->p_cost_mv - m->mvp[0]; - const uint16_t *p_cost_mvy = m->p_cost_mv - m->mvp[1]; - const int i_pixel = m->i_pixel; - const int b_chroma_me = h->mb.b_chroma_me && i_pixel <= PIXEL_8x8; - - ALIGNED_ARRAY_16( uint8_t, pix,[2],[32*18] ); // really 17x17, but round up for alignment - int omx, omy; - int i; - - int bmx = m->mv[0]; - int bmy = m->mv[1]; - int bcost = m->cost; - int odir = -1, bdir; + const int bw = x264_pixel_size[m->i_pixel].w; + const int bh = x264_pixel_size[m->i_pixel].h; + const uint16_t *p_cost_mvx = m->p_cost_mv - m->mvp[0]; + const uint16_t *p_cost_mvy = m->p_cost_mv - m->mvp[1]; + const int i_pixel = m->i_pixel; + const int b_chroma_me = h->mb.b_chroma_me && i_pixel <= PIXEL_8x8; + + ALIGNED_ARRAY_16( uint8_t, pix,[2],[32*18] ); // really 17x17, but round up for alignment + int omx, omy; + int i; + + int bmx = m->mv[0]; + int bmy = m->mv[1]; + int bcost = m->cost; + int odir = -1, bdir; - /* try the subpel component of the predicted mv */ - if( hpel_iters && h->mb.i_subpel_refine < 3 ) + /* try the subpel component of the predicted mv */ + if( hpel_iters && h->mb.i_subpel_refine < 3 ) { - int mx = x264_clip3( m->mvp[0], h->mb.mv_min_spel[0]+2, h->mb.mv_max_spel[0]-2 ); - int my = x264_clip3( m->mvp[1], h->mb.mv_min_spel[1]+2, h->mb.mv_max_spel[1]-2 ); - if( (mx-bmx)|(my-bmy) ) - COST_MV_SAD( mx, my ); + int mx = x264_clip3( m->mvp[0], h->mb.mv_min_spel[0]+2, h->mb.mv_max_spel[0]-2 ); + int my = x264_clip3( m->mvp[1], h->mb.mv_min_spel[1]+2, h->mb.mv_max_spel[1]-2 ); + if( (mx-bmx)|(my-bmy) ) + COST_MV_SAD( mx, my ); } - /* halfpel diamond search */ - for( i = hpel_iters; i > 0; i-- ) + /* halfpel diamond search */ + for( i = hpel_iters; i > 0; i-- ) { - int omx = bmx, omy = bmy; - int costs[4]; - int stride = 32; // candidates are either all hpel or all qpel, so one stride is enough - uint8_t *src0, *src1, *src2, *src3; - src0 = h->mc.get_ref( pix[0], &stride, m->p_fref, m->i_stride[0], omx, omy-2, bw, bh+1 ); - src2 = h->mc.get_ref( pix[1], &stride, m->p_fref, m->i_stride[0], omx-2, omy, bw+4, bh ); - src1 = src0 + stride; - src3 = src2 + 1; - h->pixf.fpelcmp_x4[i_pixel]( m->p_fenc[0], src0, src1, src2, src3, stride, costs ); - COPY2_IF_LT( bcost, costs[0] + p_cost_mvx[omx ] + p_cost_mvy[omy-2], bmy, omy-2 ); - COPY2_IF_LT( bcost, costs[1] + p_cost_mvx[omx ] + p_cost_mvy[omy+2], bmy, omy+2 ); - COPY3_IF_LT( bcost, costs[2] + p_cost_mvx[omx-2] + p_cost_mvy[omy ], bmx, omx-2, bmy, omy ); - COPY3_IF_LT( bcost, costs[3] + p_cost_mvx[omx+2] + p_cost_mvy[omy ], bmx, omx+2, bmy, omy ); - if( (bmx == omx) & (bmy == omy) ) - break; + int omx = bmx, omy = bmy; + int costs[4]; + int stride = 32; // candidates are either all hpel or all qpel, so one stride is enough + uint8_t *src0, *src1, *src2, *src3; + src0 = h->mc.get_ref( pix[0], &stride, m->p_fref, m->i_stride[0], omx, omy-2, bw, bh+1 ); + src2 = h->mc.get_ref( pix[1], &stride, m->p_fref, m->i_stride[0], omx-2, omy, bw+4, bh ); + src1 = src0 + stride; + src3 = src2 + 1; + h->pixf.fpelcmp_x4[i_pixel]( m->p_fenc[0], src0, src1, src2, src3, stride, costs ); + COPY2_IF_LT( bcost, costs[0] + p_cost_mvx[omx ] + p_cost_mvy[omy-2], bmy, omy-2 ); + COPY2_IF_LT( bcost, costs[1] + p_cost_mvx[omx ] + p_cost_mvy[omy+2], bmy, omy+2 ); + COPY3_IF_LT( bcost, costs[2] + p_cost_mvx[omx-2] + p_cost_mvy[omy ], bmx, omx-2, bmy, omy ); + COPY3_IF_LT( bcost, costs[3] + p_cost_mvx[omx+2] + p_cost_mvy[omy ], bmx, omx+2, bmy, omy ); + if( (bmx == omx) & (bmy == omy) ) + break; } - if( !b_refine_qpel ) + if( !b_refine_qpel ) { - bcost = COST_MAX; - COST_MV_SATD( bmx, bmy, -1 ); + bcost = COST_MAX; + COST_MV_SATD( bmx, bmy, -1 ); } - /* early termination when examining multiple reference frames */ - if( p_halfpel_thresh ) + /* early termination when examining multiple reference frames */ + if( p_halfpel_thresh ) { - if( (bcost*7)>>3 > *p_halfpel_thresh ) + if( (bcost*7)>>3 > *p_halfpel_thresh ) { - m->cost = bcost; - m->mv[0] = bmx; - m->mv[1] = bmy; - // don't need cost_mv - return; + m->cost = bcost; + m->mv[0] = bmx; + m->mv[1] = bmy; + // don't need cost_mv + return; } - else if( bcost < *p_halfpel_thresh ) - *p_halfpel_thresh = bcost; + else if( bcost < *p_halfpel_thresh ) + *p_halfpel_thresh = bcost; } - /* quarterpel diamond search */ - bdir = -1; - for( i = qpel_iters; i > 0; i-- ) + /* quarterpel diamond search */ + bdir = -1; + for( i = qpel_iters; i > 0; i-- ) { - if( bmy <= h->mb.mv_min_spel[1] || bmy >= h->mb.mv_max_spel[1] ) - break; - odir = bdir; - omx = bmx; - omy = bmy; - COST_MV_SATD( omx, omy - 1, 0 ); - COST_MV_SATD( omx, omy + 1, 1 ); - COST_MV_SATD( omx - 1, omy, 2 ); - COST_MV_SATD( omx + 1, omy, 3 ); - if( bmx == omx && bmy == omy ) - break; + if( bmy <= h->mb.mv_min_spel[1] || bmy >= h->mb.mv_max_spel[1] ) + break; + odir = bdir; + omx = bmx; + omy = bmy; + COST_MV_SATD( omx, omy - 1, 0 ); + COST_MV_SATD( omx, omy + 1, 1 ); + COST_MV_SATD( omx - 1, omy, 2 ); + COST_MV_SATD( omx + 1, omy, 3 ); + if( bmx == omx && bmy == omy ) + break; } - m->cost = bcost; - m->mv[0] = bmx; - m->mv[1] = bmy; - m->cost_mv = p_cost_mvx[ bmx ] + p_cost_mvy[ bmy ]; + m->cost = bcost; + m->mv[0] = bmx; + m->mv[1] = bmy; + m->cost_mv = p_cost_mvx[ bmx ] + p_cost_mvy[ bmy ]; } -#define BIME_CACHE( dx, dy, list ) \ -{ \ - x264_me_t *m = m##list;\ - int i = 4 + 3*dx + dy; \ - int mvx = om##list##x+dx;\ - int mvy = om##list##y+dy;\ - stride##list[i] = bw;\ +#define BIME_CACHE( dx, dy, list ) \ + { \ + x264_me_t *m = m##list; \ + int i = 4 + 3*dx + dy; \ + int mvx = om##list##x+dx; \ + int mvy = om##list##y+dy; \ + stride##list[i] = bw; \ src##list[i] = h->mc.get_ref( pixy_buf[list][i], &stride##list[i], m->p_fref, m->i_stride[0], mvx, mvy, bw, bh ); \ - if( rd )\ - {\ - if( h->mb.b_interlaced & ref##list )\ - mvy += (h->mb.i_mb_y & 1)*4 - 2;\ - h->mc.mc_chroma( pixu_buf[list][i], 8, m->p_fref[4], m->i_stride[1], mvx, mvy, bw>>1, bh>>1 );\ - h->mc.mc_chroma( pixv_buf[list][i], 8, m->p_fref[5], m->i_stride[1], mvx, mvy, bw>>1, bh>>1 );\ - }\ -} - -#define BIME_CACHE2(a,b,list) \ - BIME_CACHE(a,b,list) \ - BIME_CACHE(-(a),-(b),list) - -#define BIME_CACHE8(list) \ -{\ - BIME_CACHE2( 1, 0, list );\ - BIME_CACHE2( 0, 1, list );\ - BIME_CACHE2( 1, 1, list );\ - BIME_CACHE2( 1,-1, list );\ -} + if( rd ) \ + { \ + if( h->mb.b_interlaced & ref##list ) \ + mvy += (h->mb.i_mb_y & 1)*4 - 2; \ + h->mc.mc_chroma( pixu_buf[list][i], 8, m->p_fref[4], m->i_stride[1], mvx, mvy, bw>>1, bh>>1 ); \ + h->mc.mc_chroma( pixv_buf[list][i], 8, m->p_fref[5], m->i_stride[1], mvx, mvy, bw>>1, bh>>1 ); \ + } \ + } + +#define BIME_CACHE2(a,b,list) \ + BIME_CACHE(a,b,list) \ + BIME_CACHE(-(a),-(b),list) + +#define BIME_CACHE8(list) \ + { \ + BIME_CACHE2( 1, 0, list ); \ + BIME_CACHE2( 0, 1, list ); \ + BIME_CACHE2( 1, 1, list ); \ + BIME_CACHE2( 1,-1, list ); \ + } #define SATD_THRESH 17/16 -#define COST_BIMV_SATD( m0x, m0y, m1x, m1y ) \ -if( pass == 0 || !((visited[(m0x)&7][(m0y)&7][(m1x)&7] & (1<<((m1y)&7)))) ) \ -{ \ - int cost; \ - int i0 = 4 + 3*(m0x-om0x) + (m0y-om0y); \ - int i1 = 4 + 3*(m1x-om1x) + (m1y-om1y); \ - visited[(m0x)&7][(m0y)&7][(m1x)&7] |= (1<<((m1y)&7));\ - h->mc.avg[i_pixel]( pix, FDEC_STRIDE, src0[i0], stride0[i0], src1[i1], stride1[i1], i_weight ); \ - cost = h->pixf.mbcmp[i_pixel]( m0->p_fenc[0], FENC_STRIDE, pix, FDEC_STRIDE ) \ - + p_cost_m0x[ m0x ] + p_cost_m0y[ m0y ] \ - + p_cost_m1x[ m1x ] + p_cost_m1y[ m1y ]; \ - if( rd ) \ - { \ - if( cost < bcost * SATD_THRESH ) \ - { \ - uint64_t costrd; \ - if( cost < bcost ) \ - bcost = cost; \ - *(uint32_t*)cache0_mv = *(uint32_t*)cache0_mv2 = pack16to32_mask(m0x,m0y); \ - *(uint32_t*)cache1_mv = *(uint32_t*)cache1_mv2 = pack16to32_mask(m1x,m1y); \ - h->mc.avg[i_pixel+3]( pixu, FDEC_STRIDE, pixu_buf[0][i0], 8, pixu_buf[1][i1], 8, i_weight );\ - h->mc.avg[i_pixel+3]( pixv, FDEC_STRIDE, pixv_buf[0][i0], 8, pixv_buf[1][i1], 8, i_weight );\ - costrd = x264_rd_cost_part( h, i_lambda2, i8*4, m0->i_pixel ); \ - if( costrd < bcostrd ) \ - {\ - bcostrd = costrd;\ - bm0x = m0x; \ - bm0y = m0y; \ - bm1x = m1x; \ - bm1y = m1y; \ - }\ - } \ - } \ - else if( cost < bcost ) \ - { \ - bcost = cost; \ - bm0x = m0x; \ - bm0y = m0y; \ - bm1x = m1x; \ - bm1y = m1y; \ - } \ -} +#define COST_BIMV_SATD( m0x, m0y, m1x, m1y ) \ + if( pass == 0 || !((visited[(m0x)&7][(m0y)&7][(m1x)&7] & (1<<((m1y)&7)))) ) \ + { \ + int cost; \ + int i0 = 4 + 3*(m0x-om0x) + (m0y-om0y); \ + int i1 = 4 + 3*(m1x-om1x) + (m1y-om1y); \ + visited[(m0x)&7][(m0y)&7][(m1x)&7] |= (1<<((m1y)&7)); \ + h->mc.avg[i_pixel]( pix, FDEC_STRIDE, src0[i0], stride0[i0], src1[i1], stride1[i1], i_weight ); \ + cost = h->pixf.mbcmp[i_pixel]( m0->p_fenc[0], FENC_STRIDE, pix, FDEC_STRIDE ) \ + + p_cost_m0x[ m0x ] + p_cost_m0y[ m0y ] \ + + p_cost_m1x[ m1x ] + p_cost_m1y[ m1y ]; \ + if( rd ) \ + { \ + if( cost < bcost * SATD_THRESH ) \ + { \ + uint64_t costrd; \ + if( cost < bcost ) \ + bcost = cost; \ + *(uint32_t*)cache0_mv = *(uint32_t*)cache0_mv2 = pack16to32_mask(m0x,m0y); \ + *(uint32_t*)cache1_mv = *(uint32_t*)cache1_mv2 = pack16to32_mask(m1x,m1y); \ + h->mc.avg[i_pixel+3]( pixu, FDEC_STRIDE, pixu_buf[0][i0], 8, pixu_buf[1][i1], 8, i_weight ); \ + h->mc.avg[i_pixel+3]( pixv, FDEC_STRIDE, pixv_buf[0][i0], 8, pixv_buf[1][i1], 8, i_weight ); \ + costrd = x264_rd_cost_part( h, i_lambda2, i8*4, m0->i_pixel ); \ + if( costrd < bcostrd ) \ + { \ + bcostrd = costrd; \ + bm0x = m0x; \ + bm0y = m0y; \ + bm1x = m1x; \ + bm1y = m1y; \ + } \ + } \ + } \ + else if( cost < bcost ) \ + { \ + bcost = cost; \ + bm0x = m0x; \ + bm0y = m0y; \ + bm1x = m1x; \ + bm1y = m1y; \ + } \ + } -#define CHECK_BIDIR(a,b,c,d) \ - COST_BIMV_SATD(om0x+a, om0y+b, om1x+c, om1y+d) +#define CHECK_BIDIR(a,b,c,d) \ + COST_BIMV_SATD(om0x+a, om0y+b, om1x+c, om1y+d) static void ALWAYS_INLINE x264_me_refine_bidir( x264_t *h, x264_me_t *m0, x264_me_t *m1, int i_weight, int i8, int i_lambda2, int rd ) { - static const int pixel_mv_offs[] = { 0, 4, 4*8, 0 }; - int16_t *cache0_mv = h->mb.cache.mv[0][x264_scan8[i8*4]]; - int16_t *cache0_mv2 = cache0_mv + pixel_mv_offs[m0->i_pixel]; - int16_t *cache1_mv = h->mb.cache.mv[1][x264_scan8[i8*4]]; - int16_t *cache1_mv2 = cache1_mv + pixel_mv_offs[m0->i_pixel]; - const int i_pixel = m0->i_pixel; - const int bw = x264_pixel_size[i_pixel].w; - const int bh = x264_pixel_size[i_pixel].h; - const uint16_t *p_cost_m0x = m0->p_cost_mv - m0->mvp[0]; - const uint16_t *p_cost_m0y = m0->p_cost_mv - m0->mvp[1]; - const uint16_t *p_cost_m1x = m1->p_cost_mv - m1->mvp[0]; - const uint16_t *p_cost_m1y = m1->p_cost_mv - m1->mvp[1]; - ALIGNED_ARRAY_16( uint8_t, pixy_buf,[2],[9][16*16] ); - ALIGNED_8( uint8_t pixu_buf[2][9][8*8] ); - ALIGNED_8( uint8_t pixv_buf[2][9][8*8] ); - uint8_t *src0[9]; - uint8_t *src1[9]; - uint8_t *pix = &h->mb.pic.p_fdec[0][(i8>>1)*8*FDEC_STRIDE+(i8&1)*8]; - uint8_t *pixu = &h->mb.pic.p_fdec[1][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4]; - uint8_t *pixv = &h->mb.pic.p_fdec[2][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4]; - int ref0 = h->mb.cache.ref[0][x264_scan8[i8*4]]; - int ref1 = h->mb.cache.ref[1][x264_scan8[i8*4]]; - int stride0[9]; - int stride1[9]; - int bm0x = m0->mv[0], om0x = bm0x; - int bm0y = m0->mv[1], om0y = bm0y; - int bm1x = m1->mv[0], om1x = bm1x; - int bm1y = m1->mv[1], om1y = bm1y; - int bcost = COST_MAX; - int pass = 0; - int j; - int mc_list0 = 1, mc_list1 = 1; - uint64_t bcostrd = COST_MAX64; - /* each byte of visited represents 8 possible m1y positions, so a 4D array isn't needed */ - ALIGNED_ARRAY_16( uint8_t, visited,[8],[8][8] ); - /* all permutations of an offset in up to 2 of the dimensions */ - static const int8_t dia4d[32][4] = { - {0,0,0,1}, {0,0,0,-1}, {0,0,1,0}, {0,0,-1,0}, - {0,1,0,0}, {0,-1,0,0}, {1,0,0,0}, {-1,0,0,0}, - {0,0,1,1}, {0,0,-1,-1},{0,1,1,0}, {0,-1,-1,0}, - {1,1,0,0}, {-1,-1,0,0},{1,0,0,1}, {-1,0,0,-1}, - {0,1,0,1}, {0,-1,0,-1},{1,0,1,0}, {-1,0,-1,0}, - {0,0,-1,1},{0,0,1,-1}, {0,-1,1,0},{0,1,-1,0}, - {-1,1,0,0},{1,-1,0,0}, {1,0,0,-1},{-1,0,0,1}, - {0,-1,0,1},{0,1,0,-1}, {-1,0,1,0},{1,0,-1,0}, - }; - - if( bm0y < h->mb.mv_min_spel[1] + 8 || bm1y < h->mb.mv_min_spel[1] + 8 || - bm0y > h->mb.mv_max_spel[1] - 8 || bm1y > h->mb.mv_max_spel[1] - 8 ) - return; - - h->mc.memzero_aligned( visited, sizeof(uint8_t[8][8][8]) ); - - BIME_CACHE( 0, 0, 0 ); - BIME_CACHE( 0, 0, 1 ); - CHECK_BIDIR( 0, 0, 0, 0 ); + static const int pixel_mv_offs[] = { 0, 4, 4*8, 0 }; + int16_t *cache0_mv = h->mb.cache.mv[0][x264_scan8[i8*4]]; + int16_t *cache0_mv2 = cache0_mv + pixel_mv_offs[m0->i_pixel]; + int16_t *cache1_mv = h->mb.cache.mv[1][x264_scan8[i8*4]]; + int16_t *cache1_mv2 = cache1_mv + pixel_mv_offs[m0->i_pixel]; + const int i_pixel = m0->i_pixel; + const int bw = x264_pixel_size[i_pixel].w; + const int bh = x264_pixel_size[i_pixel].h; + const uint16_t *p_cost_m0x = m0->p_cost_mv - m0->mvp[0]; + const uint16_t *p_cost_m0y = m0->p_cost_mv - m0->mvp[1]; + const uint16_t *p_cost_m1x = m1->p_cost_mv - m1->mvp[0]; + const uint16_t *p_cost_m1y = m1->p_cost_mv - m1->mvp[1]; + ALIGNED_ARRAY_16( uint8_t, pixy_buf,[2],[9][16*16] ); + ALIGNED_8( uint8_t pixu_buf[2][9][8*8] ); + ALIGNED_8( uint8_t pixv_buf[2][9][8*8] ); + uint8_t *src0[9]; + uint8_t *src1[9]; + uint8_t *pix = &h->mb.pic.p_fdec[0][(i8>>1)*8*FDEC_STRIDE+(i8&1)*8]; + uint8_t *pixu = &h->mb.pic.p_fdec[1][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4]; + uint8_t *pixv = &h->mb.pic.p_fdec[2][(i8>>1)*4*FDEC_STRIDE+(i8&1)*4]; + int ref0 = h->mb.cache.ref[0][x264_scan8[i8*4]]; + int ref1 = h->mb.cache.ref[1][x264_scan8[i8*4]]; + int stride0[9]; + int stride1[9]; + int bm0x = m0->mv[0], om0x = bm0x; + int bm0y = m0->mv[1], om0y = bm0y; + int bm1x = m1->mv[0], om1x = bm1x; + int bm1y = m1->mv[1], om1y = bm1y; + int bcost = COST_MAX; + int pass = 0; + int j; + int mc_list0 = 1, mc_list1 = 1; + uint64_t bcostrd = COST_MAX64; + /* each byte of visited represents 8 possible m1y positions, so a 4D array isn't needed */ + ALIGNED_ARRAY_16( uint8_t, visited,[8],[8][8] ); + /* all permutations of an offset in up to 2 of the dimensions */ + static const int8_t dia4d[32][4] = { + {0,0,0,1}, {0,0,0,-1}, {0,0,1,0}, {0,0,-1,0}, + {0,1,0,0}, {0,-1,0,0}, {1,0,0,0}, {-1,0,0,0}, + {0,0,1,1}, {0,0,-1,-1},{0,1,1,0}, {0,-1,-1,0}, + {1,1,0,0}, {-1,-1,0,0},{1,0,0,1}, {-1,0,0,-1}, + {0,1,0,1}, {0,-1,0,-1},{1,0,1,0}, {-1,0,-1,0}, + {0,0,-1,1},{0,0,1,-1}, {0,-1,1,0},{0,1,-1,0}, + {-1,1,0,0},{1,-1,0,0}, {1,0,0,-1},{-1,0,0,1}, + {0,-1,0,1},{0,1,0,-1}, {-1,0,1,0},{1,0,-1,0}, + }; + + if( bm0y < h->mb.mv_min_spel[1] + 8 || bm1y < h->mb.mv_min_spel[1] + 8 || + bm0y > h->mb.mv_max_spel[1] - 8 || bm1y > h->mb.mv_max_spel[1] - 8 ) + return; + + h->mc.memzero_aligned( visited, sizeof(uint8_t[8][8][8]) ); + + BIME_CACHE( 0, 0, 0 ); + BIME_CACHE( 0, 0, 1 ); + CHECK_BIDIR( 0, 0, 0, 0 ); - for( pass = 0; pass < 8; pass++ ) + for( pass = 0; pass < 8; pass++ ) { - /* check all mv pairs that differ in at most 2 components from the current mvs. */ - /* doesn't do chroma ME. this probably doesn't matter, as the gains - * from bidir ME are the same with and without chroma ME. */ - - if( mc_list0 ) - BIME_CACHE8( 0 ); - if( mc_list1 ) - BIME_CACHE8( 1 ); - - for( j=0; j<32; j++ ) - CHECK_BIDIR( dia4d[j][0], dia4d[j][1], dia4d[j][2], dia4d[j][3] ); - - mc_list0 = (om0x-bm0x)|(om0y-bm0y); - mc_list1 = (om1x-bm1x)|(om1y-bm1y); - if( !mc_list0 && !mc_list1 ) - break; - - om0x = bm0x; - om0y = bm0y; - om1x = bm1x; - om1y = bm1y; - - if( mc_list0 ) - BIME_CACHE( 0, 0, 0 ); - if( mc_list1 ) - BIME_CACHE( 0, 0, 1 ); + /* check all mv pairs that differ in at most 2 components from the current mvs. */ + /* doesn't do chroma ME. this probably doesn't matter, as the gains + * from bidir ME are the same with and without chroma ME. */ + + if( mc_list0 ) + BIME_CACHE8( 0 ); + if( mc_list1 ) + BIME_CACHE8( 1 ); + + for( j=0; j<32; j++ ) + CHECK_BIDIR( dia4d[j][0], dia4d[j][1], dia4d[j][2], dia4d[j][3] ); + + mc_list0 = (om0x-bm0x)|(om0y-bm0y); + mc_list1 = (om1x-bm1x)|(om1y-bm1y); + if( !mc_list0 && !mc_list1 ) + break; + + om0x = bm0x; + om0y = bm0y; + om1x = bm1x; + om1y = bm1y; + + if( mc_list0 ) + BIME_CACHE( 0, 0, 0 ); + if( mc_list1 ) + BIME_CACHE( 0, 0, 1 ); } - m0->mv[0] = bm0x; - m0->mv[1] = bm0y; - m1->mv[0] = bm1x; - m1->mv[1] = bm1y; + m0->mv[0] = bm0x; + m0->mv[1] = bm0y; + m1->mv[0] = bm1x; + m1->mv[1] = bm1y; } void x264_me_refine_bidir_satd( x264_t *h, x264_me_t *m0, x264_me_t *m1, int i_weight ) { - x264_me_refine_bidir( h, m0, m1, i_weight, 0, 0, 0 ); + x264_me_refine_bidir( h, m0, m1, i_weight, 0, 0, 0 ); } void x264_me_refine_bidir_rd( x264_t *h, x264_me_t *m0, x264_me_t *m1, int i_weight, int i8, int i_lambda2 ) { - /* Motion compensation is done as part of bidir_rd; don't repeat - * it in encoding. */ - h->mb.b_skip_mc = 1; - x264_me_refine_bidir( h, m0, m1, i_weight, i8, i_lambda2, 1 ); - h->mb.b_skip_mc = 0; + /* Motion compensation is done as part of bidir_rd; don't repeat + * it in encoding. */ + h->mb.b_skip_mc = 1; + x264_me_refine_bidir( h, m0, m1, i_weight, i8, i_lambda2, 1 ); + h->mb.b_skip_mc = 0; } #undef COST_MV_SATD -#define COST_MV_SATD( mx, my, dst, avoid_mvp ) \ -{ \ - if( !avoid_mvp || !(mx == pmx && my == pmy) ) \ - { \ - int stride = 16; \ +#define COST_MV_SATD( mx, my, dst, avoid_mvp ) \ + { \ + if( !avoid_mvp || !(mx == pmx && my == pmy) ) \ + { \ + int stride = 16; \ uint8_t *src = h->mc.get_ref( pix, &stride, m->p_fref, m->i_stride[0], mx, my, bw*4, bh*4 ); \ dst = h->pixf.mbcmp_unaligned[i_pixel]( m->p_fenc[0], FENC_STRIDE, src, stride ) \ - + p_cost_mvx[mx] + p_cost_mvy[my]; \ - COPY1_IF_LT( bsatd, dst ); \ - } \ - else \ - dst = COST_MAX; \ -} - -#define COST_MV_RD( mx, my, satd, do_dir, mdir ) \ -{ \ - if( satd <= bsatd * SATD_THRESH ) \ - { \ - uint64_t cost; \ + + p_cost_mvx[mx] + p_cost_mvy[my]; \ + COPY1_IF_LT( bsatd, dst ); \ + } \ + else \ + dst = COST_MAX; \ + } + +#define COST_MV_RD( mx, my, satd, do_dir, mdir ) \ + { \ + if( satd <= bsatd * SATD_THRESH ) \ + { \ + uint64_t cost; \ *(uint32_t*)cache_mv = *(uint32_t*)cache_mv2 = pack16to32_mask(mx,my); \ - cost = x264_rd_cost_part( h, i_lambda2, i4, m->i_pixel ); \ + cost = x264_rd_cost_part( h, i_lambda2, i4, m->i_pixel ); \ COPY4_IF_LT( bcost, cost, bmx, mx, bmy, my, dir, do_dir?mdir:dir ); \ - } \ -} + } \ + } void x264_me_refine_qpel_rd( x264_t *h, x264_me_t *m, int i_lambda2, int i4, int i_list ) { - // don't have to fill the whole mv cache rectangle - static const int pixel_mv_offs[] = { 0, 4, 4*8, 0, 2, 2*8, 0 }; - int16_t *cache_mv = h->mb.cache.mv[i_list][x264_scan8[i4]]; - int16_t *cache_mv2 = cache_mv + pixel_mv_offs[m->i_pixel]; - const uint16_t *p_cost_mvx, *p_cost_mvy; - const int bw = x264_pixel_size[m->i_pixel].w>>2; - const int bh = x264_pixel_size[m->i_pixel].h>>2; - const int i_pixel = m->i_pixel; - - ALIGNED_ARRAY_16( uint8_t, pix,[16*16] ); - uint64_t bcost = COST_MAX64; - int bmx = m->mv[0]; - int bmy = m->mv[1]; - int omx, omy, pmx, pmy, i, j; - unsigned bsatd; - int satd = 0; - int dir = -2; - int satds[8]; - - if( m->i_pixel != PIXEL_16x16 && i4 != 0 ) - x264_mb_predict_mv( h, i_list, i4, bw, m->mvp ); - pmx = m->mvp[0]; - pmy = m->mvp[1]; - p_cost_mvx = m->p_cost_mv - pmx; - p_cost_mvy = m->p_cost_mv - pmy; - COST_MV_SATD( bmx, bmy, bsatd, 0 ); - if( m->i_pixel != PIXEL_16x16 ) - COST_MV_RD( bmx, bmy, 0, 0, 0 ) + // don't have to fill the whole mv cache rectangle + static const int pixel_mv_offs[] = { 0, 4, 4*8, 0, 2, 2*8, 0 }; + int16_t *cache_mv = h->mb.cache.mv[i_list][x264_scan8[i4]]; + int16_t *cache_mv2 = cache_mv + pixel_mv_offs[m->i_pixel]; + const uint16_t *p_cost_mvx, *p_cost_mvy; + const int bw = x264_pixel_size[m->i_pixel].w>>2; + const int bh = x264_pixel_size[m->i_pixel].h>>2; + const int i_pixel = m->i_pixel; + + ALIGNED_ARRAY_16( uint8_t, pix,[16*16] ); + uint64_t bcost = COST_MAX64; + int bmx = m->mv[0]; + int bmy = m->mv[1]; + int omx, omy, pmx, pmy, i, j; + unsigned bsatd; + int satd = 0; + int dir = -2; + int satds[8]; + + if( m->i_pixel != PIXEL_16x16 && i4 != 0 ) + x264_mb_predict_mv( h, i_list, i4, bw, m->mvp ); + pmx = m->mvp[0]; + pmy = m->mvp[1]; + p_cost_mvx = m->p_cost_mv - pmx; + p_cost_mvy = m->p_cost_mv - pmy; + COST_MV_SATD( bmx, bmy, bsatd, 0 ); + if( m->i_pixel != PIXEL_16x16 ) + COST_MV_RD( bmx, bmy, 0, 0, 0 ) else - bcost = m->cost; + bcost = m->cost; - /* check the predicted mv */ - if( (bmx != pmx || bmy != pmy) - && pmx >= h->mb.mv_min_spel[0] && pmx <= h->mb.mv_max_spel[0] - && pmy >= h->mb.mv_min_spel[1] && pmy <= h->mb.mv_max_spel[1] ) + /* check the predicted mv */ + if( (bmx != pmx || bmy != pmy) + && pmx >= h->mb.mv_min_spel[0] && pmx <= h->mb.mv_max_spel[0] + && pmy >= h->mb.mv_min_spel[1] && pmy <= h->mb.mv_max_spel[1] ) { - COST_MV_SATD( pmx, pmy, satd, 0 ); - COST_MV_RD( pmx, pmy, satd, 0,0 ); - /* The hex motion search is guaranteed to not repeat the center candidate, - * so if pmv is chosen, set the "MV to avoid checking" to bmv instead. */ - if( bmx == pmx && bmy == pmy ) + COST_MV_SATD( pmx, pmy, satd, 0 ); + COST_MV_RD( pmx, pmy, satd, 0,0 ); + /* The hex motion search is guaranteed to not repeat the center candidate, + * so if pmv is chosen, set the "MV to avoid checking" to bmv instead. */ + if( bmx == pmx && bmy == pmy ) { - pmx = m->mv[0]; - pmy = m->mv[1]; + pmx = m->mv[0]; + pmy = m->mv[1]; } } - if( bmy < h->mb.mv_min_spel[1] + 3 || - bmy > h->mb.mv_max_spel[1] - 3 ) - return; - - /* subpel hex search, same pattern as ME HEX. */ - dir = -2; - omx = bmx; - omy = bmy; - for( j=0; j<6; j++ ) COST_MV_SATD( omx + hex2[j+1][0], omy + hex2[j+1][1], satds[j], 1 ); - for( j=0; j<6; j++ ) COST_MV_RD ( omx + hex2[j+1][0], omy + hex2[j+1][1], satds[j], 1,j ); + if( bmy < h->mb.mv_min_spel[1] + 3 || + bmy > h->mb.mv_max_spel[1] - 3 ) + return; + + /* subpel hex search, same pattern as ME HEX. */ + dir = -2; + omx = bmx; + omy = bmy; + for( j=0; j<6; j++ ) COST_MV_SATD( omx + hex2[j+1][0], omy + hex2[j+1][1], satds[j], 1 ); + for( j=0; j<6; j++ ) COST_MV_RD ( omx + hex2[j+1][0], omy + hex2[j+1][1], satds[j], 1,j ); - if( dir != -2 ) + if( dir != -2 ) { - /* half hexagon, not overlapping the previous iteration */ - for( i = 1; i < 10; i++ ) + /* half hexagon, not overlapping the previous iteration */ + for( i = 1; i < 10; i++ ) { - const int odir = mod6m1[dir+1]; - if( bmy < h->mb.mv_min_spel[1] + 3 || - bmy > h->mb.mv_max_spel[1] - 3 ) - break; - dir = -2; - omx = bmx; - omy = bmy; - for( j=0; j<3; j++ ) COST_MV_SATD( omx + hex2[odir+j][0], omy + hex2[odir+j][1], satds[j], 1 ); - for( j=0; j<3; j++ ) COST_MV_RD ( omx + hex2[odir+j][0], omy + hex2[odir+j][1], satds[j], 1, odir-1+j ); - if( dir == -2 ) - break; + const int odir = mod6m1[dir+1]; + if( bmy < h->mb.mv_min_spel[1] + 3 || + bmy > h->mb.mv_max_spel[1] - 3 ) + break; + dir = -2; + omx = bmx; + omy = bmy; + for( j=0; j<3; j++ ) COST_MV_SATD( omx + hex2[odir+j][0], omy + hex2[odir+j][1], satds[j], 1 ); + for( j=0; j<3; j++ ) COST_MV_RD ( omx + hex2[odir+j][0], omy + hex2[odir+j][1], satds[j], 1, odir-1+j ); + if( dir == -2 ) + break; } } - /* square refine, same pattern as ME HEX. */ - omx = bmx; - omy = bmy; - for( i=0; i<8; i++ ) COST_MV_SATD( omx + square1[i+1][0], omy + square1[i+1][1], satds[i], 1 ); - for( i=0; i<8; i++ ) COST_MV_RD ( omx + square1[i+1][0], omy + square1[i+1][1], satds[i], 0,0 ); - - m->cost = bcost; - m->mv[0] = bmx; - m->mv[1] = bmy; - x264_macroblock_cache_mv ( h, block_idx_x[i4], block_idx_y[i4], bw, bh, i_list, pack16to32_mask(bmx, bmy) ); - x264_macroblock_cache_mvd( h, block_idx_x[i4], block_idx_y[i4], bw, bh, i_list, pack16to32_mask(bmx - m->mvp[0], bmy - m->mvp[1]) ); + /* square refine, same pattern as ME HEX. */ + omx = bmx; + omy = bmy; + for( i=0; i<8; i++ ) COST_MV_SATD( omx + square1[i+1][0], omy + square1[i+1][1], satds[i], 1 ); + for( i=0; i<8; i++ ) COST_MV_RD ( omx + square1[i+1][0], omy + square1[i+1][1], satds[i], 0,0 ); + + m->cost = bcost; + m->mv[0] = bmx; + m->mv[1] = bmy; + x264_macroblock_cache_mv ( h, block_idx_x[i4], block_idx_y[i4], bw, bh, i_list, pack16to32_mask(bmx, bmy) ); + x264_macroblock_cache_mvd( h, block_idx_x[i4], block_idx_y[i4], bw, bh, i_list, pack16to32_mask(bmx - m->mvp[0], bmy - m->mvp[1]) ); } diff --exclude=.git --exclude=/gitignore -Naur x264/encoder/slicetype.c x264-or/encoder/slicetype.c --- x264/encoder/slicetype.c 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/encoder/slicetype.c 2009-10-28 15:05:29.000000000 +0100 @@ -909,7 +909,9 @@ /* Restore frametypes for all frames that haven't actually been decided yet. */ for( j = reset_start; j <= num_frames; j++ ) - frames[j]->i_type = X264_TYPE_AUTO; + frames[j]->i_type = X264_TYPE_AUTO; + + return; } void x264_slicetype_decide( x264_t *h ) diff --exclude=.git --exclude=/gitignore -Naur x264/link.ld x264-or/link.ld --- x264/link.ld 1970-01-01 01:00:00.000000000 +0100 +++ x264-or/link.ld 2009-10-28 15:05:29.000000000 +0100 @@ -0,0 +1,100 @@ +/* +MEMORY + { + vectors : ORIGIN = 0x00000000, LENGTH = 0x00002000 + flash : ORIGIN = 0x04000000, LENGTH = 0x00200000 + ram : ORIGIN = 0x00002000, LENGTH = 0x001fe000 + icm : ORIGIN = 0x00800000, LENGTH = 0x00004000 + } + */ + +MEMORY + { +/* + reset : ORIGIN = 0x00000000, LENGTH = 0x00000200 + vectors : ORIGIN = 0x00000200, LENGTH = 0x00001E00 + text : ORIGIN = 0x00002000, LENGTH = 0x000fe000 + data : ORIGIN = 0x00100000, LENGTH = 0x00fe0000 + stack : ORIGIN = 0x001fe000, LENGTH = 0x00010000 +*/ + yuv_data : ORIGIN = 25M, LENGTH = 7M + } + +/* +MEMORY + { + reset : ORIGIN = 0xc0000000, LENGTH = 0x00000200 + vectors : ORIGIN = 0xc0000200, LENGTH = 0x00001000 + ram : ORIGIN = 0xc0001200, LENGTH = 0x00FFED00 + } +*/ +SECTIONS +{ +/* + .reset : + { + *(.reset) + } > reset + + + + .vectors : + { + _vec_start = .; + *(.vectors) + _vec_end = .; + } > vectors + + .text : + { + *(.text) + } > text + + .rodata : + { + *(.rodata) + *(.rodata.*) + } > text + + .icm : + { + _icm_start = .; + *(.icm) + _icm_end = .; + } > data + + .data : + { + _dst_beg = .; + *(.data) + _dst_end = .; + } > data + + .bss : + { + *(.bss) + } > data + + .heap : + { + _heap_start = .; + *(.heap) + _heap_end = .; + } > data + + .stack (NOLOAD) : + { + *(.stack) + _src_addr = .; + } > data +*/ + .yuv_data : + { + _yuv_data_start = .; + __yuv_data_start = .; + *(.yuv_data) + _yuv_data_end = .; + __yuv_data_end = .; + } > yuv_data + +} diff --exclude=.git --exclude=/gitignore -Naur x264/Makefile x264-or/Makefile --- x264/Makefile 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/Makefile 2009-11-15 18:48:46.000000000 +0100 @@ -91,7 +91,66 @@ $(SONAME): .depend $(OBJS) $(OBJASM) $(CC) -shared -o $@ $(OBJS) $(OBJASM) $(SOFLAGS) $(LDFLAGS) -x264$(EXE): $(OBJCLI) libx264.a + +## OR32-specific build rules +## +## Essentially we create some sample YUV video data +## and embed it in an ELF file, in a section called yuv_data. +## In the linker script we then link this section to a place +## in ram and tell x264 to look here when it wants data. +## +## You can download sample h264 encoded CIF sequences +## here: http://www.tkn.tu-berlin.de/research/evalvid/cif.html +## +## +GEN_FILES=encoder/analyse_init_log2.c +VIDEO_DATA_FILE=yuv_data.elf +INPUT_VIDEO_FILE ?= ../test-sequences/football_cif.264 +INPUT_VIDEO_SIZE ?= cif +YUV_VIDEO_FILE ?= video_data.yuv + +# We will use a limit of about 5MB of video data to encode +ifeq ($(INPUT_VIDEO_SIZE), 4cif) +NUM_FRAMES ?=10 +endif +ifeq ($(INPUT_VIDEO_SIZE), cif) +NUM_FRAMES ?=30 +endif +ifeq ($(INPUT_VIDEO_SIZE), qcif) +NUM_FRAMES ?=90 +endif + +$(INPUT_VIDEO_FILE): + @echo; echo; + @echo "\tNo sample video file to embed! Please edit the Makefile" + @echo "\tand set the variable INPUT_VIDEO_FILE to the location of" + @echo "\tsome sample video material." + @echo "\tOr, isntead of editing the makefile, specify it on the" + @echo "\tcommand line like so:" + @echo "\t\tINPUT_VIDEO_FILE=../coastguard_cif.yuv"; echo + @echo "\tYou can also specify the size of the video (cif, qcif etc.)" + @echo "\tby specifying INPUT_VIDEO_SIZE - the default is cif, however" + @echo "\tyou will need to change the program's hardcoded resolution" + @echo + exit 1 +$(YUV_VIDEO_FILE): $(INPUT_VIDEO_FILE) +# First convert it to raw YUV, only about 5MB large, so for cif take 30 frames, qcif is 90 frames + ffmpeg -s $(INPUT_VIDEO_SIZE) -i $(INPUT_VIDEO_FILE) -vframes $(NUM_FRAMES) $(YUV_VIDEO_FILE) +$(VIDEO_DATA_FILE): $(YUV_VIDEO_FILE) + or32-elf-ld -r -b binary -o yuv_data.o $(YUV_VIDEO_FILE) + or32-elf-objcopy --rename-section .data=.yuv_data yuv_data.o + mv yuv_data.o $(VIDEO_DATA_FILE) +encoder/analyse_init_log2.c: encoder/analyse_gen_init_array.sh + cd encoder && chmod a+x analyse_gen_init_array.sh && ./analyse_gen_init_array.sh +#OR32_DEPS= reset.o except.o uart.o syscalls.o $(VIDEO_DATA_FILE) + +OR32_DEPS= $(VIDEO_DATA_FILE) $(GEN_FILES) + +sim: x264$(EXE) + or32-elf-sim -f or1ksim_x264.cfg $< + + +x264$(EXE): $(OR32_DEPS) $(OBJCLI) libx264.a $(CC) -o $@ $+ $(LDFLAGS) checkasm: tools/checkasm.o libx264.a @@ -147,12 +206,13 @@ endif clean: - rm -f $(OBJS) $(OBJASM) $(OBJCLI) $(SONAME) *.a x264 x264.exe .depend TAGS + rm -f $(OBJS) $(OBJASM) $(OBJCLI) $(SONAME) *.a x264 x264.exe .depend TAGS $(YUV_VIDEO_FILE) rm -f checkasm checkasm.exe tools/checkasm.o tools/checkasm-a.o rm -f $(SRC2:%.c=%.gcda) $(SRC2:%.c=%.gcno) - sed -e 's/ *-fprofile-\(generate\|use\)//g' config.mak > config.mak2 && mv config.mak2 config.mak distclean: clean + rm -f $(OR32_DEPS) uart0* $(GEN_FILES) rm -f config.mak config.h x264.pc rm -rf test/ diff --exclude=.git --exclude=/gitignore -Naur x264/muxers.c x264-or/muxers.c --- x264/muxers.c 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/muxers.c 2009-11-15 18:45:09.000000000 +0100 @@ -52,15 +52,32 @@ typedef struct { - FILE *fh; - int width, height; - int next_frame; + //FILE *fh; + char *fh; + + int width, height; + int next_frame; + } yuv_input_t; +void init_yuv_dataspace(char* yuv_dat_addr, hnd_t *p_handle, x264_param_t *p_param) +{ + yuv_input_t *h = malloc( sizeof(yuv_input_t) ); + V(fprintf( stderr, "init_yuv_dataspace: h addr = 0x%.8x\n", h)); + h->fh = yuv_dat_addr; + V(fprintf( stderr, "init_yuv_dataspace: h->fh 0x%.8x\n",h->fh)); + h->width = p_param->i_width; + V(fprintf( stderr, "init_yuv_dataspace: h->width %d\n",h->width)); + h->height = p_param->i_height; + V(fprintf( stderr, "init_yuv_dataspace: h->height %d\n",h->height)); + h->next_frame = 0; + *p_handle = (hnd_t *)h; +} + /* raw 420 yuv file operation */ int open_file_yuv( char *psz_filename, hnd_t *p_handle, x264_param_t *p_param ) { - yuv_input_t *h = malloc( sizeof(yuv_input_t) ); + yuv_input_t *h = malloc( sizeof(yuv_input_t) ); if( !h ) return -1; h->width = p_param->i_width; @@ -80,7 +97,10 @@ int get_frame_total_yuv( hnd_t handle ) { - yuv_input_t *h = handle; + yuv_input_t *h = handle; + + V(fprintf(stderr, "get_frame_total_yuv args: handle: 0x%.8x, h: 0x%.8x\n", handle, h)); + /* int i_frame_total = 0; if( !fseek( h->fh, 0, SEEK_END ) ) @@ -88,24 +108,45 @@ uint64_t i_size = ftell( h->fh ); fseek( h->fh, 0, SEEK_SET ); i_frame_total = (int)(i_size / ( h->width * h->height * 3 / 2 )); - } + }*/ + +#ifdef USE_HARDCODED_FRAME_NUM + return (int)HARDCODED_FRAME_NUM; +#else + fprintf( stderr, "get_frame_total_yuv: %d %d %d %d\n", + h->width, h->height, + YUV_DATA_SIZE, + (int)(YUV_DATA_SIZE / ( h->width * h->height * 3 / 2 ))); + return (int)(YUV_DATA_SIZE / ( h->width * h->height * 3 / 2 ));//i_frame_total; + +#endif - return i_frame_total; } int read_frame_yuv( x264_picture_t *p_pic, hnd_t handle, int i_frame ) { yuv_input_t *h = handle; - + int i; + if( i_frame != h->next_frame ) - if( fseek( h->fh, (uint64_t)i_frame * h->width * h->height * 3 / 2, SEEK_SET ) ) - return -1; - - if( fread( p_pic->img.plane[0], 1, h->width * h->height, h->fh ) <= 0 - || fread( p_pic->img.plane[1], 1, h->width * h->height / 4, h->fh ) <= 0 - || fread( p_pic->img.plane[2], 1, h->width * h->height / 4, h->fh ) <= 0 ) - return -1; + //if( fseek( h->fh, (uint64_t)i_frame * h->width * h->height * 3 / 2, SEEK_SET ) ) + // Progress pointer to beginning of this frame + h->fh = (void*) YUV_DATA_ADDR + i_frame * h->width * h->height * 3 / 2; + //return -1; + + //if( fread( p_pic->img.plane[0], 1, h->width * h->height, h->fh ) <= 0 + for (i=0;i<h->width * h->height;i++) p_pic->img.plane[0][i] = h->fh[i]; + h->fh += h->width * h->height; + //|| fread( p_pic->img.plane[1], 1, h->width * h->height / 4, h->fh ) <= 0 + for (i=0;i<(h->width * h->height) / 4;i++) p_pic->img.plane[1][i] = h->fh[i]; + h->fh += (h->width * h->height) / 4; + + //|| fread( p_pic->img.plane[2], 1, h->width * h->height / 4, h->fh ) <= 0 ) + for (i=0;i<(h->width * h->height) / 4;i++) p_pic->img.plane[2][i] = h->fh[i]; + h->fh += (h->width * h->height) / 4; + //return -1; + h->next_frame = i_frame+1; return 0; @@ -113,12 +154,15 @@ int close_file_yuv(hnd_t handle) { + return 0; + /* // -- jb yuv_input_t *h = handle; if( !h || !h->fh ) return 0; fclose( h->fh ); free( h ); return 0; + */ } /* YUV4MPEG2 raw 420 yuv file operation */ @@ -530,11 +574,21 @@ } #endif +#ifdef ENC_OUT_ADDR +char * enc_out_pos; +#endif int open_file_bsf( char *psz_filename, hnd_t *p_handle ) { +#ifdef ENC_OUT_ADDR + V(fprintf(stderr, "open_file_bsf\n")); + // Use a hardcoded memory location for outputting encoded data to + p_handle = (void *) ENC_OUT_ADDR; + enc_out_pos = (char *) ENC_OUT_ADDR; +#else if( !(*p_handle = fopen(psz_filename, "w+b")) ) return -1; +#endif return 0; } @@ -546,9 +600,16 @@ int write_nalu_bsf( hnd_t handle, uint8_t *p_nalu, int i_size ) { +#ifdef ENC_OUT_ADDR + V(fprintf(stderr, "write_nalu_bsf: writing %d bytes from 0x%.8x\n", i_size, (unsigned long) enc_out_pos)); + // Just write to the spot in memory and increment the pointer + int i; for (i=0;i<i_size;i++) *enc_out_pos++ = p_nalu[i]; + return i_size; +#else if( fwrite( p_nalu, i_size, 1, (FILE*)handle ) > 0 ) return i_size; return -1; +#endif } int set_eop_bsf( hnd_t handle, x264_picture_t *p_picture ) @@ -558,6 +619,15 @@ int close_file_bsf( hnd_t handle ) { +#ifdef ENC_OUT_ADDR + // Let's calculate a checksum for the written data + int i; char* d = (char*)ENC_OUT_ADDR; unsigned int cksum; + unsigned int size = (unsigned int)enc_out_pos - (unsigned int)ENC_OUT_ADDR; + for(i=0;i<size;i++)cksum+=d[i++]; + fprintf(stderr, "close_file_bsf: wrote %d bytes from 0x%.8x to 0x%.8x\n", size, (unsigned int) ENC_OUT_ADDR, (unsigned int) enc_out_pos); + fprintf(stderr, "close_file_bsf: cksum 0x%.8x\n", cksum); + return 0; +#endif if( !handle || handle == stdout ) return 0; diff --exclude=.git --exclude=/gitignore -Naur x264/muxers.h x264-or/muxers.h --- x264/muxers.h 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/muxers.h 2009-10-28 15:05:29.000000000 +0100 @@ -27,6 +27,7 @@ typedef void *hnd_t; int open_file_yuv( char *psz_filename, hnd_t *p_handle, x264_param_t *p_param ); +void init_yuv_dataspace(char* yuv_dat_addr, hnd_t *p_handle, x264_param_t *p_param); int get_frame_total_yuv( hnd_t handle ); int read_frame_yuv( x264_picture_t *p_pic, hnd_t handle, int i_frame ); int close_file_yuv( hnd_t handle ); diff --exclude=.git --exclude=/gitignore -Naur x264/or1ksim_x264.cfg x264-or/or1ksim_x264.cfg --- x264/or1ksim_x264.cfg 1970-01-01 01:00:00.000000000 +0100 +++ x264-or/or1ksim_x264.cfg 2009-11-15 19:57:09.000000000 +0100 @@ -0,0 +1,886 @@ +/* sim.cfg -- Simulator configuration script file + Copyright (C) 2001-2002, Marko Mlinar, markom@opencores.org + +This file is part of OpenRISC 1000 Architectural Simulator. +It contains the default configuration and help about configuring +the simulator. + +This program 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. + +This program 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 this program; if not, write to the Free Software +Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ + + +/* INTRODUCTION + + The ork1sim has various parameters, that are set in configuration files + like this one. The user can switch between configurations at startup by + specifying the required configuration file with the -f <filename.cfg> option. + If no configuration file is specified or1ksim searches for the default + configuration file sim.cfg. First it searches for './sim.cfg'. If this + file is not found, it searches for '~/or1k/sim.cfg'. If this file is + not found too, it reverts to the built-in default configuration. + + NOTE: Users should not rely on the built-in configuration, since the + default configuration may differ between version. + Rather create a configuration file that sets all critical values. + + This file may contain (standard C) comments only - no // support. + + Configure files may be be included, using: + include "file_name_to_include" + + Like normal configuration files, the included file is divided into + sections. Each section is described in detail also. + + Some section have subsections. One example of such a subsection is: + + device <index> + instance specific parameters... + enddevice + + which creates a device instance. +*/ + + +/* MEMORY SECTION + + This section specifies how the memory is generated and the blocks + it consists of. + + type = random/unknown/pattern + Specifies the initial memory values. + 'random' generates random memory using seed 'random_seed'. + 'pattern' fills memory with 'pattern'. + 'unknown' does not specify how memory should be generated, + leaving the memory in a undefined state. This is the fastest + option. + + random_seed = <value> + random seed for randomizer, used if type = 'random'. + + pattern = <value> + pattern to fill memory, used if type = 'pattern'. + + nmemories = <value> + number of memory instances connected + + baseaddr = <hex_value> + memory start address + + size = <hex_value> + memory size + + name = "<string>" + memory block name + + ce = <value> + chip enable index of the memory instance + + mc = <value> + memory controller this memory is connected to + + delayr = <value> + cycles, required for read access, -1 if instance does not support reading + + delayw = <value> + cycles, required for write access, -1 if instance does not support writing + + log = "<filename>" + filename, where to log memory accesses to, no log, if log command is not specified +*/ + + +section memory + /*random_seed = 12345 + type = random*/ + pattern = 0x00 + type = unknown /* Fastest */ + + name = "FLASH" + ce = 0 + mc = 0 + baseaddr = 0xf0000000 + size = 0x01000000 + delayr = 10 + delayw = -1 +end + +section memory + /*random_seed = 12345 + type = random*/ + pattern = 0x00 + type = unknown /* Fastest */ + + name = "RAM" + ce = 1 + mc = 0 + baseaddr = 0x00000000 + size = 0x02000000 + delayr = 20 + delayw = 25 +end + +section memory + /*random_seed = 12345 + type = random*/ + pattern = 0x00 + type = unknown /* Fastest */ + + name = "SRAM" + mc = 0 + ce = 2 + baseaddr = 0xa4000000 + size = 0x00100000 + delayr = 1 + delayw = 2 +end + + +/* IMMU SECTION + + This section configures the Instruction Memory Manangement Unit + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of ITLB sets; must be power of two + + nways = <value> + number of ITLB ways + + pagesize = <value> + instruction page size; must be power of two + + entrysize = <value> + instruction entry size in bytes + + ustates = <value> + number of ITLB usage states (2, 3, 4 etc., max is 4) + + hitdelay = <value> + number of cycles immu hit costs + + missdelay = <value> + number of cycles immu miss costs +*/ + +section immu + enabled = 1 + nsets = 64 + nways = 1 + pagesize = 8192 + hitdelay = 0 + missdelay = 0 +end + + +/* DMMU SECTION + + This section configures the Data Memory Manangement Unit + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of DTLB sets; must be power of two + + nways = <value> + number of DTLB ways + + pagesize = <value> + data page size; must be power of two + + entrysize = <value> + data entry size in bytes + + ustates = <value> + number of DTLB usage states (2, 3, 4 etc., max is 4) + + hitdelay = <value> + number of cycles dmmu hit costs + + missdelay = <value> + number of cycles dmmu miss costs +*/ + +section dmmu + enabled = 1 + nsets = 64 + nways = 1 + pagesize = 8192 + hitdelay = 0 + missdelay = 0 +end + + +/* IC SECTION + + This section configures the Instruction Cache + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of IC sets; must be power of two + + nways = <value> + number of IC ways + + blocksize = <value> + IC block size in bytes; must be power of two + + ustates = <value> + number of IC usage states (2, 3, 4 etc., max is 4) + + hitdelay = <value> + number of cycles ic hit costs + + missdelay = <value> + number of cycles ic miss costs +*/ + +section ic + enabled = 0 + nsets = 512 + nways = 1 + blocksize = 16 + hitdelay = 20 + missdelay = 20 +end + + +/* DC SECTION + + This section configures the Data Cache + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of DC sets; must be power of two + + nways = <value> + number of DC ways + + blocksize = <value> + DC block size in bytes; must be power of two + + ustates = <value> + number of DC usage states (2, 3, 4 etc., max is 4) + + load_hitdelay = <value> + number of cycles dc load hit costs + + load_missdelay = <value> + number of cycles dc load miss costs + + store_hitdelay = <value> + number of cycles dc load hit costs + + store_missdelay = <value> + number of cycles dc load miss costs +*/ + +section dc + enabled = 0 + nsets = 512 + nways = 1 + blocksize = 16 + load_hitdelay = 20 + load_missdelay = 20 + store_hitdelay = 20 + store_missdelay = 20 +end + + +/* SIM SECTION + + This section specifies how or1ksim should behave. + + verbose = 0/1 + '0': don't print extra messages + '1': print extra messages + + debug = 0-9 + 0 : no debug messages + 1-9: debug message level. + higher numbers produce more messages + + profile = 0/1 + '0': don't generate profiling file 'sim.profile' + '1': don't generate profiling file 'sim.profile' + + prof_fn = "<filename>" + optional filename for the profiling file. + valid only if 'profile' is set + + mprofile = 0/1 + '0': don't generate memory profiling file 'sim.mprofile' + '1': generate memory profiling file 'sim.mprofile' + + mprof_fn = "<filename>" + optional filename for the memory profiling file. + valid only if 'mprofile' is set + + history = 0/1 + '0': don't track execution flow + '1': track execution flow + Execution flow can be tracked for the simulator's + 'hist' command. Useful for back-trace debugging. + + iprompt = 0/1 + '0': start in <not interactive prompt> (so what do we start in ???) + '1': start in interactive prompt. + + exe_log = 0/1 + '0': don't generate execution log. + '1': generate execution log. + + exe_log = default/hardware/simple/software + type of execution log, default is used when not specified + + exe_log_start = <value> + index of first instruction to start logging, default = 0 + + exe_log_end = <value> + index of last instruction to end logging; not limited, if omitted + + exe_log_marker = <value> + <value> specifies number of instructions before horizontal marker is + printed; if zero, markers are disabled (default) + + exe_log_fn = "<filename>" + filename for the exection log file. + valid only if 'exe_log' is set + + clkcycle = <value>[ps|ns|us|ms] + specifies time measurement for one cycle +*/ + +section sim + verbose = 1 + debug = 0 + profile = 0 + history = 0 + + clkcycle = 10ns +end + + +/* SECTION VAPI + + This section configures the Verification API, used for Advanced + Core Verification. + + enabled = 0/1 + '0': disbable VAPI server + '1': enable/start VAPI server + + server_port = <value> + TCP/IP port to start VAPI server on + + log_enabled = 0/1 + '0': disable VAPI requests logging + '1': enable VAPI requests logging + + hide_device_id = 0/1 + '0': don't log device id (for compatability with old version) + '1': log device id + + + vapi_fn = <filename> + filename for the log file. + valid only if log_enabled is set +*/ + +section VAPI + enabled = 0 + server_port = 9998 + log_enabled = 0 + vapi_log_fn = "vapi.log" +end + + +/* CPU SECTION + + This section specifies various CPU parameters. + + ver = <value> + rev = <value> + specifies version and revision of the CPU used + + upr = <value> + changes the upr register + + sr = <value> + sets the initial Supervision Register value + + superscalar = 0/1 + '0': CPU is scalar + '1': CPU is superscalar + (modify cpu/or32/execute.c to tune superscalar model) + + hazards = 0/1 + '0': don't track data hazards in superscalar CPU + '1': track data hazards in superscalar CPU + If tracked, data hazards can be displayed using the + simulator's 'r' command. + + dependstats = 0/1 + '0': don't calculate inter-instruction dependencies. + '1': calculate inter-instruction dependencies. + If calculated, inter-instruction dependencies can be + displayed using the simulator's 'stat' command. + + sbuf_len = <value> + length of store buffer (<= 256), 0 = disabled +*/ + +section cpu + ver = 0x12 + cfg = 0x00 + rev = 0x01 + /* upr = */ + superscalar = 0 + hazards = 0 + dependstats = 0 + sbuf_len = 0 + hardfloat = 1 +end + + +/* PM SECTION + + This section specifies Power Management parameters + + enabled = 0/1 + '0': disable power management + '1': enable power management +*/ + +section pm + enabled = 0 +end + + +/* BPB SECTION + + This section specifies how branch prediction should behave. + + enabled = 0/1 + '0': disable branch prediction + '1': enable branch prediction + + btic = 0/1 + '0': disable branch target instruction cache model + '1': enable branch target instruction cache model + + sbp_bf_fwd = 0/1 + Static branch prediction for 'l.bf' + '0': don't use forward prediction + '1': use forward prediction + + sbp_bnf_fwd = 0/1 + Static branch prediction for 'l.bnf' + '0': don't use forward prediction + '1': use forward prediction + + hitdelay = <value> + number of cycles bpb hit costs + + missdelay = <value> + number of cycles bpb miss costs +*/ + +section bpb + enabled = 0 + btic = 0 + sbp_bf_fwd = 0 + sbp_bnf_fwd = 0 + hitdelay = 0 + missdelay = 0 +end + + +/* DEBUG SECTION + + This sections specifies how the debug unit should behave. + + enabled = 0/1 + '0': disable debug unit + '1': enable debug unit + + gdb_enabled = 0/1 + '0': don't start gdb server + '1': start gdb server at port 'server_port' + + server_port = <value> + TCP/IP port to start gdb server on + valid only if gdb_enabled is set + + vapi_id = <hex_value> + Used to create "fake" vapi log file containing the JTAG proxy messages. +*/ +section debug + enabled = 0 + rsp_enabled = 0 + rsp_port = 5554 + /*server_port = 9999*/ +end + + +/* MC SECTION + + This section configures the memory controller + + enabled = 0/1 + '0': disable memory controller + '1': enable memory controller + + baseaddr = <hex_value> + address of first MC register + + POC = <hex_value> + Power On Configuration register + + index = <value> + Index of this memory controller amongst all the memory controllers +*/ + +section mc + enabled = 0 + baseaddr = 0x93000000 + POC = 0x00000008 /* Power on configuration register */ + index = 0 +end + + +/* UART SECTION + + This section configures the UARTs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first UART register for this device + + + channel = <channeltype>:<args> + + The channel parameter indicates the source of received UART characters + and the sink for transmitted UART characters. + + The <channeltype> can be either "file", "xterm", "tcp", "fd", or "tty" + (without quotes). + + A) To send/receive characters from a pair of files, use a file + channel: + + channel=file:<rxfile>,<txfile> + + B) To create an interactive terminal window, use an xterm channel: + + channel=xterm:[<xterm_arg>]* + + C) To create a bidirectional tcp socket which one could, for example, + access via telnet, use a tcp channel: + + channel=tcp:<port number> + + D) To cause the UART to read/write from existing numeric file + descriptors, use an fd channel: + + channel=fd:<rx file descriptor num>,<tx file descriptor num> + + E) To connect the UART to a physical serial port, create a tty + channel: + + channel=tty:device=/dev/ttyS0,baud=9600 + + irq = <value> + irq number for this device + + 16550 = 0/1 + '0': this device is a UART16450 + '1': this device is a UART16550 + + jitter = <value> + in msecs... time to block, -1 to disable it + + vapi_id = <hex_value> + VAPI id of this instance +*/ + +section uart + enabled = 1 + baseaddr = 0x90000000 + irq = 2 + channel = "file:uart0.rx,uart0.tx" + /* channel = "tcp:10084" */ + /* channel = "xterm:" */ + jitter = -1 /* async behaviour */ + 16550 = 1 +end + + +/* DMA SECTION + + This section configures the DMAs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first DMA register for this device + + irq = <value> + irq number for this device + + vapi_id = <hex_value> + VAPI id of this instance +*/ + +section dma + enabled = 1 + baseaddr = 0x9a000000 + irq = 11 +end + + +/* ETHERNET SECTION + + This section configures the ETHERNETs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first ethernet register for this device + + dma = <value> + which controller is this ethernet "connected" to + + irq = <value> + ethernet mac IRQ level + + rtx_type = <value> + use 0 - file interface, 1 - socket interface + + rx_channel = <value> + DMA channel used for RX + + tx_channel = <value> + DMA channel used for TX + + rxfile = "<filename>" + filename, where to read data from + + txfile = "<filename>" + filename, where to write data to + + sockif = "<ifacename>" + interface name of ethernet socket + + vapi_id = <hex_value> + VAPI id of this instance +*/ + +section ethernet + enabled = 0 + baseaddr = 0x92000000 + /* dma = 0 */ + irq = 4 + rtx_type = 0 + /* tx_channel = 0 */ + /* rx_channel = 1 */ + /*rxfile = "eth0.rx"*/ + txfile = "eth0.tx" + sockif = "eth0" +end + + +/* GPIO SECTION + + This section configures the GPIOs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first GPIO register for this device + + irq = <value> + irq number for this device + + base_vapi_id = <hex_value> + first VAPI id of this instance + GPIO uses 8 consecutive VAPI IDs +*/ + +section gpio + enabled = 0 + baseaddr = 0x91000000 + irq = 3 + base_vapi_id = 0x0200 +end + +/* VGA SECTION + + This section configures the VGA/LCD controller + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first VGA register + + irq = <value> + irq number for this device + + refresh_rate = <value> + number of cycles between screen dumps + + filename = "<filename>" + template name for generated names (e.g. "primary" produces "primary0023.bmp") +*/ + +section vga + enabled = 1 + baseaddr = 0x97100000 + irq = 8 + refresh_rate = 100000 + filename = "primary" +end + + +/* TICK TIMER SECTION + + This section configures tick timer + + enabled = 0/1 + whether tick timer is enabled +*/ + +section pic + enabled = 1 + edge_trigger = 1 +end + +/* FB SECTION + + This section configures the frame buffer + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + base address of frame buffer + + paladdr = <hex_value> + base address of first palette entry + + refresh_rate = <value> + number of cycles between screen dumps + + filename = "<filename>" + template name for generated names (e.g. "primary" produces "primary0023.bmp") +*/ + +section fb + enabled = 1 + baseaddr = 0x97000000 + refresh_rate = 1000000 + filename = "primary" +end + + +/* KBD SECTION + + This section configures the PS/2 compatible keyboard + + baseaddr = <hex_value> + base address of the keyboard device + + rxfile = "<filename>" + filename, where to read data from +*/ + +section kbd + enabled = 1 + irq = 5 + baseaddr = 0x94000000 + rxfile = "kbd.rx" +end + + +/* ATA SECTION + + This section configures the ATA/ATAPI host controller + + baseaddr = <hex_value> + address of first ATA register + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + irq = <value> + irq number for this device + + debug = <value> + debug level for ata models. + 0: no debug messages + 1: verbose messages + 3: normal messages (more messages than verbose) + 5: debug messages (normal debug messages) + 7: flow control messages (debug statemachine flows) + 9: low priority message (display everything the code does) + + dev_type0/1 = <value> + ata device 0 type + 0: NO_CONNeCT: none (not connected) + 1: FILE : simulated harddisk + 2: LOCAL : local system harddisk + + dev_file0/1 = "<filename>" + filename for simulated ATA device + valid only if dev_type0 == 1 + + dev_size0/1 = <value> + size of simulated hard-disk (in MBytes) + valid only if dev_type0 == 1 + + dev_packet0/1 = <value> + 0: simulated ATA device does NOT implement PACKET command feature set + 1: simulated ATA device does implement PACKET command feature set + + FIXME: irq number +*/ + +section ata + enabled = 0 + baseaddr = 0x9e000000 + irq = 15 + +end + + diff --exclude=.git --exclude=/gitignore -Naur x264/or1ksim_x264_sadssdmod.cfg x264-or/or1ksim_x264_sadssdmod.cfg --- x264/or1ksim_x264_sadssdmod.cfg 1970-01-01 01:00:00.000000000 +0100 +++ x264-or/or1ksim_x264_sadssdmod.cfg 2009-11-17 00:15:03.000000000 +0100 @@ -0,0 +1,896 @@ +/* sim.cfg -- Simulator configuration script file + Copyright (C) 2001-2002, Marko Mlinar, markom@opencores.org + +This file is part of OpenRISC 1000 Architectural Simulator. +It contains the default configuration and help about configuring +the simulator. + +This program 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. + +This program 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 this program; if not, write to the Free Software +Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ + + +/* INTRODUCTION + + The ork1sim has various parameters, that are set in configuration files + like this one. The user can switch between configurations at startup by + specifying the required configuration file with the -f <filename.cfg> option. + If no configuration file is specified or1ksim searches for the default + configuration file sim.cfg. First it searches for './sim.cfg'. If this + file is not found, it searches for '~/or1k/sim.cfg'. If this file is + not found too, it reverts to the built-in default configuration. + + NOTE: Users should not rely on the built-in configuration, since the + default configuration may differ between version. + Rather create a configuration file that sets all critical values. + + This file may contain (standard C) comments only - no // support. + + Configure files may be be included, using: + include "file_name_to_include" + + Like normal configuration files, the included file is divided into + sections. Each section is described in detail also. + + Some section have subsections. One example of such a subsection is: + + device <index> + instance specific parameters... + enddevice + + which creates a device instance. +*/ + + +/* MEMORY SECTION + + This section specifies how the memory is generated and the blocks + it consists of. + + type = random/unknown/pattern + Specifies the initial memory values. + 'random' generates random memory using seed 'random_seed'. + 'pattern' fills memory with 'pattern'. + 'unknown' does not specify how memory should be generated, + leaving the memory in a undefined state. This is the fastest + option. + + random_seed = <value> + random seed for randomizer, used if type = 'random'. + + pattern = <value> + pattern to fill memory, used if type = 'pattern'. + + nmemories = <value> + number of memory instances connected + + baseaddr = <hex_value> + memory start address + + size = <hex_value> + memory size + + name = "<string>" + memory block name + + ce = <value> + chip enable index of the memory instance + + mc = <value> + memory controller this memory is connected to + + delayr = <value> + cycles, required for read access, -1 if instance does not support reading + + delayw = <value> + cycles, required for write access, -1 if instance does not support writing + + log = "<filename>" + filename, where to log memory accesses to, no log, if log command is not specified +*/ + + +section memory + /*random_seed = 12345 + type = random*/ + pattern = 0x00 + type = unknown /* Fastest */ + + name = "FLASH" + ce = 0 + mc = 0 + baseaddr = 0xf0000000 + size = 0x01000000 + delayr = 10 + delayw = -1 +end + +section memory + /*random_seed = 12345 + type = random*/ + pattern = 0x00 + type = unknown /* Fastest */ + + name = "RAM" + ce = 1 + mc = 0 + baseaddr = 0x00000000 + size = 0x02000000 + delayr = 20 + delayw = 25 +end + +section memory + /*random_seed = 12345 + type = random*/ + pattern = 0x00 + type = unknown /* Fastest */ + + name = "SRAM" + mc = 0 + ce = 2 + baseaddr = 0xa4000000 + size = 0x00100000 + delayr = 1 + delayw = 2 +end + + +/* IMMU SECTION + + This section configures the Instruction Memory Manangement Unit + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of ITLB sets; must be power of two + + nways = <value> + number of ITLB ways + + pagesize = <value> + instruction page size; must be power of two + + entrysize = <value> + instruction entry size in bytes + + ustates = <value> + number of ITLB usage states (2, 3, 4 etc., max is 4) + + hitdelay = <value> + number of cycles immu hit costs + + missdelay = <value> + number of cycles immu miss costs +*/ + +section immu + enabled = 0 + nsets = 64 + nways = 1 + pagesize = 8192 + hitdelay = 0 + missdelay = 0 +end + + +/* DMMU SECTION + + This section configures the Data Memory Manangement Unit + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of DTLB sets; must be power of two + + nways = <value> + number of DTLB ways + + pagesize = <value> + data page size; must be power of two + + entrysize = <value> + data entry size in bytes + + ustates = <value> + number of DTLB usage states (2, 3, 4 etc., max is 4) + + hitdelay = <value> + number of cycles dmmu hit costs + + missdelay = <value> + number of cycles dmmu miss costs +*/ + +section dmmu + enabled = 0 + nsets = 64 + nways = 1 + pagesize = 8192 + hitdelay = 0 + missdelay = 0 +end + + +/* IC SECTION + + This section configures the Instruction Cache + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of IC sets; must be power of two + + nways = <value> + number of IC ways + + blocksize = <value> + IC block size in bytes; must be power of two + + ustates = <value> + number of IC usage states (2, 3, 4 etc., max is 4) + + hitdelay = <value> + number of cycles ic hit costs + + missdelay = <value> + number of cycles ic miss costs +*/ + +section ic + enabled = 0 + nsets = 512 + nways = 1 + blocksize = 16 + hitdelay = 20 + missdelay = 20 +end + + +/* DC SECTION + + This section configures the Data Cache + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of DC sets; must be power of two + + nways = <value> + number of DC ways + + blocksize = <value> + DC block size in bytes; must be power of two + + ustates = <value> + number of DC usage states (2, 3, 4 etc., max is 4) + + load_hitdelay = <value> + number of cycles dc load hit costs + + load_missdelay = <value> + number of cycles dc load miss costs + + store_hitdelay = <value> + number of cycles dc load hit costs + + store_missdelay = <value> + number of cycles dc load miss costs +*/ + +section dc + enabled = 0 + nsets = 512 + nways = 1 + blocksize = 16 + load_hitdelay = 20 + load_missdelay = 20 + store_hitdelay = 20 + store_missdelay = 20 +end + + +/* SIM SECTION + + This section specifies how or1ksim should behave. + + verbose = 0/1 + '0': don't print extra messages + '1': print extra messages + + debug = 0-9 + 0 : no debug messages + 1-9: debug message level. + higher numbers produce more messages + + profile = 0/1 + '0': don't generate profiling file 'sim.profile' + '1': don't generate profiling file 'sim.profile' + + prof_fn = "<filename>" + optional filename for the profiling file. + valid only if 'profile' is set + + mprofile = 0/1 + '0': don't generate memory profiling file 'sim.mprofile' + '1': generate memory profiling file 'sim.mprofile' + + mprof_fn = "<filename>" + optional filename for the memory profiling file. + valid only if 'mprofile' is set + + history = 0/1 + '0': don't track execution flow + '1': track execution flow + Execution flow can be tracked for the simulator's + 'hist' command. Useful for back-trace debugging. + + iprompt = 0/1 + '0': start in <not interactive prompt> (so what do we start in ???) + '1': start in interactive prompt. + + exe_log = 0/1 + '0': don't generate execution log. + '1': generate execution log. + + exe_log = default/hardware/simple/software + type of execution log, default is used when not specified + + exe_log_start = <value> + index of first instruction to start logging, default = 0 + + exe_log_end = <value> + index of last instruction to end logging; not limited, if omitted + + exe_log_marker = <value> + <value> specifies number of instructions before horizontal marker is + printed; if zero, markers are disabled (default) + + exe_log_fn = "<filename>" + filename for the exection log file. + valid only if 'exe_log' is set + + clkcycle = <value>[ps|ns|us|ms] + specifies time measurement for one cycle +*/ + +section sim + verbose = 1 + debug = 0 + profile = 0 + history = 0 + + clkcycle = 10ns +end + + +/* SECTION VAPI + + This section configures the Verification API, used for Advanced + Core Verification. + + enabled = 0/1 + '0': disbable VAPI server + '1': enable/start VAPI server + + server_port = <value> + TCP/IP port to start VAPI server on + + log_enabled = 0/1 + '0': disable VAPI requests logging + '1': enable VAPI requests logging + + hide_device_id = 0/1 + '0': don't log device id (for compatability with old version) + '1': log device id + + + vapi_fn = <filename> + filename for the log file. + valid only if log_enabled is set +*/ + +section VAPI + enabled = 0 + server_port = 9998 + log_enabled = 0 + vapi_log_fn = "vapi.log" +end + + +/* CPU SECTION + + This section specifies various CPU parameters. + + ver = <value> + rev = <value> + specifies version and revision of the CPU used + + upr = <value> + changes the upr register + + sr = <value> + sets the initial Supervision Register value + + superscalar = 0/1 + '0': CPU is scalar + '1': CPU is superscalar + (modify cpu/or32/execute.c to tune superscalar model) + + hazards = 0/1 + '0': don't track data hazards in superscalar CPU + '1': track data hazards in superscalar CPU + If tracked, data hazards can be displayed using the + simulator's 'r' command. + + dependstats = 0/1 + '0': don't calculate inter-instruction dependencies. + '1': calculate inter-instruction dependencies. + If calculated, inter-instruction dependencies can be + displayed using the simulator's 'stat' command. + + sbuf_len = <value> + length of store buffer (<= 256), 0 = disabled +*/ + +section cpu + ver = 0x12 + cfg = 0x00 + rev = 0x01 + /* upr = */ + superscalar = 0 + hazards = 0 + dependstats = 0 + sbuf_len = 0 + hardfloat = 1 +end + + +/* PM SECTION + + This section specifies Power Management parameters + + enabled = 0/1 + '0': disable power management + '1': enable power management +*/ + +section pm + enabled = 0 +end + + +/* BPB SECTION + + This section specifies how branch prediction should behave. + + enabled = 0/1 + '0': disable branch prediction + '1': enable branch prediction + + btic = 0/1 + '0': disable branch target instruction cache model + '1': enable branch target instruction cache model + + sbp_bf_fwd = 0/1 + Static branch prediction for 'l.bf' + '0': don't use forward prediction + '1': use forward prediction + + sbp_bnf_fwd = 0/1 + Static branch prediction for 'l.bnf' + '0': don't use forward prediction + '1': use forward prediction + + hitdelay = <value> + number of cycles bpb hit costs + + missdelay = <value> + number of cycles bpb miss costs +*/ + +section bpb + enabled = 0 + btic = 0 + sbp_bf_fwd = 0 + sbp_bnf_fwd = 0 + hitdelay = 0 + missdelay = 0 +end + + +/* DEBUG SECTION + + This sections specifies how the debug unit should behave. + + enabled = 0/1 + '0': disable debug unit + '1': enable debug unit + + gdb_enabled = 0/1 + '0': don't start gdb server + '1': start gdb server at port 'server_port' + + server_port = <value> + TCP/IP port to start gdb server on + valid only if gdb_enabled is set + + vapi_id = <hex_value> + Used to create "fake" vapi log file containing the JTAG proxy messages. +*/ +section debug + enabled = 0 + rsp_enabled = 0 + rsp_port = 5554 + /*server_port = 9999*/ +end + + +/* MC SECTION + + This section configures the memory controller + + enabled = 0/1 + '0': disable memory controller + '1': enable memory controller + + baseaddr = <hex_value> + address of first MC register + + POC = <hex_value> + Power On Configuration register + + index = <value> + Index of this memory controller amongst all the memory controllers +*/ + +section mc + enabled = 0 + baseaddr = 0x93000000 + POC = 0x00000008 /* Power on configuration register */ + index = 0 +end + + +/* UART SECTION + + This section configures the UARTs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first UART register for this device + + + channel = <channeltype>:<args> + + The channel parameter indicates the source of received UART characters + and the sink for transmitted UART characters. + + The <channeltype> can be either "file", "xterm", "tcp", "fd", or "tty" + (without quotes). + + A) To send/receive characters from a pair of files, use a file + channel: + + channel=file:<rxfile>,<txfile> + + B) To create an interactive terminal window, use an xterm channel: + + channel=xterm:[<xterm_arg>]* + + C) To create a bidirectional tcp socket which one could, for example, + access via telnet, use a tcp channel: + + channel=tcp:<port number> + + D) To cause the UART to read/write from existing numeric file + descriptors, use an fd channel: + + channel=fd:<rx file descriptor num>,<tx file descriptor num> + + E) To connect the UART to a physical serial port, create a tty + channel: + + channel=tty:device=/dev/ttyS0,baud=9600 + + irq = <value> + irq number for this device + + 16550 = 0/1 + '0': this device is a UART16450 + '1': this device is a UART16550 + + jitter = <value> + in msecs... time to block, -1 to disable it + + vapi_id = <hex_value> + VAPI id of this instance +*/ + +section uart + enabled = 1 + baseaddr = 0x90000000 + irq = 2 + channel = "file:uart0.rx,uart0.tx" + /* channel = "tcp:10084" */ + /* channel = "xterm:" */ + jitter = -1 /* async behaviour */ + 16550 = 1 +end + + +/* DMA SECTION + + This section configures the DMAs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first DMA register for this device + + irq = <value> + irq number for this device + + vapi_id = <hex_value> + VAPI id of this instance +*/ + +section dma + enabled = 0 + baseaddr = 0x9a000000 + irq = 11 +end + + +/* ETHERNET SECTION + + This section configures the ETHERNETs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first ethernet register for this device + + dma = <value> + which controller is this ethernet "connected" to + + irq = <value> + ethernet mac IRQ level + + rtx_type = <value> + use 0 - file interface, 1 - socket interface + + rx_channel = <value> + DMA channel used for RX + + tx_channel = <value> + DMA channel used for TX + + rxfile = "<filename>" + filename, where to read data from + + txfile = "<filename>" + filename, where to write data to + + sockif = "<ifacename>" + interface name of ethernet socket + + vapi_id = <hex_value> + VAPI id of this instance +*/ + +section ethernet + enabled = 0 + baseaddr = 0x92000000 + /* dma = 0 */ + irq = 4 + rtx_type = 0 + /* tx_channel = 0 */ + /* rx_channel = 1 */ + /*rxfile = "eth0.rx"*/ + txfile = "eth0.tx" + sockif = "eth0" +end + + +/* GPIO SECTION + + This section configures the GPIOs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first GPIO register for this device + + irq = <value> + irq number for this device + + base_vapi_id = <hex_value> + first VAPI id of this instance + GPIO uses 8 consecutive VAPI IDs +*/ + +section gpio + enabled = 0 + baseaddr = 0x91000000 + irq = 3 + base_vapi_id = 0x0200 +end + +/* VGA SECTION + + This section configures the VGA/LCD controller + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first VGA register + + irq = <value> + irq number for this device + + refresh_rate = <value> + number of cycles between screen dumps + + filename = "<filename>" + template name for generated names (e.g. "primary" produces "primary0023.bmp") +*/ + +section vga + enabled = 0 + baseaddr = 0x97100000 + irq = 8 + refresh_rate = 100000 + filename = "primary" +end + + +/* TICK TIMER SECTION + + This section configures tick timer + + enabled = 0/1 + whether tick timer is enabled +*/ + +section pic + enabled = 1 + edge_trigger = 1 +end + +/* FB SECTION + + This section configures the frame buffer + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + base address of frame buffer + + paladdr = <hex_value> + base address of first palette entry + + refresh_rate = <value> + number of cycles between screen dumps + + filename = "<filename>" + template name for generated names (e.g. "primary" produces "primary0023.bmp") +*/ + +section fb + enabled = 0 + baseaddr = 0x97000000 + refresh_rate = 1000000 + filename = "primary" +end + + +/* KBD SECTION + + This section configures the PS/2 compatible keyboard + + baseaddr = <hex_value> + base address of the keyboard device + + rxfile = "<filename>" + filename, where to read data from +*/ + +section kbd + enabled = 0 + irq = 5 + baseaddr = 0x94000000 + rxfile = "kbd.rx" +end + + +/* ATA SECTION + + This section configures the ATA/ATAPI host controller + + baseaddr = <hex_value> + address of first ATA register + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + irq = <value> + irq number for this device + + debug = <value> + debug level for ata models. + 0: no debug messages + 1: verbose messages + 3: normal messages (more messages than verbose) + 5: debug messages (normal debug messages) + 7: flow control messages (debug statemachine flows) + 9: low priority message (display everything the code does) + + dev_type0/1 = <value> + ata device 0 type + 0: NO_CONNeCT: none (not connected) + 1: FILE : simulated harddisk + 2: LOCAL : local system harddisk + + dev_file0/1 = "<filename>" + filename for simulated ATA device + valid only if dev_type0 == 1 + + dev_size0/1 = <value> + size of simulated hard-disk (in MBytes) + valid only if dev_type0 == 1 + + dev_packet0/1 = <value> + 0: simulated ATA device does NOT implement PACKET command feature set + 1: simulated ATA device does implement PACKET command feature set + + FIXME: irq number +*/ + +section ata + enabled = 0 + baseaddr = 0x9e000000 + irq = 15 + +end + + + +/* SAD/SSD MODULE SECTION + This module calculates SAD/SSD on a set of data, specifically + for use with the x264 H.264 encoding software. +*/ +section x264_sadssdmod + enabled = 1 + baseaddr = 0x26400000 + name = "x264_sadssdmod0" +end diff --exclude=.git --exclude=/gitignore -Naur x264/rsp_or1ksim_x264.cfg x264-or/rsp_or1ksim_x264.cfg --- x264/rsp_or1ksim_x264.cfg 1970-01-01 01:00:00.000000000 +0100 +++ x264-or/rsp_or1ksim_x264.cfg 2009-11-15 19:56:58.000000000 +0100 @@ -0,0 +1,886 @@ +/* sim.cfg -- Simulator configuration script file + Copyright (C) 2001-2002, Marko Mlinar, markom@opencores.org + +This file is part of OpenRISC 1000 Architectural Simulator. +It contains the default configuration and help about configuring +the simulator. + +This program 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. + +This program 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 this program; if not, write to the Free Software +Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ + + +/* INTRODUCTION + + The ork1sim has various parameters, that are set in configuration files + like this one. The user can switch between configurations at startup by + specifying the required configuration file with the -f <filename.cfg> option. + If no configuration file is specified or1ksim searches for the default + configuration file sim.cfg. First it searches for './sim.cfg'. If this + file is not found, it searches for '~/or1k/sim.cfg'. If this file is + not found too, it reverts to the built-in default configuration. + + NOTE: Users should not rely on the built-in configuration, since the + default configuration may differ between version. + Rather create a configuration file that sets all critical values. + + This file may contain (standard C) comments only - no // support. + + Configure files may be be included, using: + include "file_name_to_include" + + Like normal configuration files, the included file is divided into + sections. Each section is described in detail also. + + Some section have subsections. One example of such a subsection is: + + device <index> + instance specific parameters... + enddevice + + which creates a device instance. +*/ + + +/* MEMORY SECTION + + This section specifies how the memory is generated and the blocks + it consists of. + + type = random/unknown/pattern + Specifies the initial memory values. + 'random' generates random memory using seed 'random_seed'. + 'pattern' fills memory with 'pattern'. + 'unknown' does not specify how memory should be generated, + leaving the memory in a undefined state. This is the fastest + option. + + random_seed = <value> + random seed for randomizer, used if type = 'random'. + + pattern = <value> + pattern to fill memory, used if type = 'pattern'. + + nmemories = <value> + number of memory instances connected + + baseaddr = <hex_value> + memory start address + + size = <hex_value> + memory size + + name = "<string>" + memory block name + + ce = <value> + chip enable index of the memory instance + + mc = <value> + memory controller this memory is connected to + + delayr = <value> + cycles, required for read access, -1 if instance does not support reading + + delayw = <value> + cycles, required for write access, -1 if instance does not support writing + + log = "<filename>" + filename, where to log memory accesses to, no log, if log command is not specified +*/ + + +section memory + /*random_seed = 12345 + type = random*/ + pattern = 0x00 + type = unknown /* Fastest */ + + name = "FLASH" + ce = 0 + mc = 0 + baseaddr = 0xf0000000 + size = 0x01000000 + delayr = 10 + delayw = -1 +end + +section memory + /*random_seed = 12345 + type = random*/ + pattern = 0x00 + type = unknown /* Fastest */ + + name = "RAM" + ce = 1 + mc = 0 + baseaddr = 0x00000000 + size = 0x02000000 + delayr = 20 + delayw = 25 +end + +section memory + /*random_seed = 12345 + type = random*/ + pattern = 0x00 + type = unknown /* Fastest */ + + name = "SRAM" + mc = 0 + ce = 2 + baseaddr = 0xa4000000 + size = 0x00100000 + delayr = 1 + delayw = 2 +end + + +/* IMMU SECTION + + This section configures the Instruction Memory Manangement Unit + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of ITLB sets; must be power of two + + nways = <value> + number of ITLB ways + + pagesize = <value> + instruction page size; must be power of two + + entrysize = <value> + instruction entry size in bytes + + ustates = <value> + number of ITLB usage states (2, 3, 4 etc., max is 4) + + hitdelay = <value> + number of cycles immu hit costs + + missdelay = <value> + number of cycles immu miss costs +*/ + +section immu + enabled = 1 + nsets = 64 + nways = 1 + pagesize = 8192 + hitdelay = 0 + missdelay = 0 +end + + +/* DMMU SECTION + + This section configures the Data Memory Manangement Unit + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of DTLB sets; must be power of two + + nways = <value> + number of DTLB ways + + pagesize = <value> + data page size; must be power of two + + entrysize = <value> + data entry size in bytes + + ustates = <value> + number of DTLB usage states (2, 3, 4 etc., max is 4) + + hitdelay = <value> + number of cycles dmmu hit costs + + missdelay = <value> + number of cycles dmmu miss costs +*/ + +section dmmu + enabled = 1 + nsets = 64 + nways = 1 + pagesize = 8192 + hitdelay = 0 + missdelay = 0 +end + + +/* IC SECTION + + This section configures the Instruction Cache + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of IC sets; must be power of two + + nways = <value> + number of IC ways + + blocksize = <value> + IC block size in bytes; must be power of two + + ustates = <value> + number of IC usage states (2, 3, 4 etc., max is 4) + + hitdelay = <value> + number of cycles ic hit costs + + missdelay = <value> + number of cycles ic miss costs +*/ + +section ic + enabled = 0 + nsets = 512 + nways = 1 + blocksize = 16 + hitdelay = 20 + missdelay = 20 +end + + +/* DC SECTION + + This section configures the Data Cache + + enabled = 0/1 + '0': disabled + '1': enabled + (NOTE: UPR bit is set) + + nsets = <value> + number of DC sets; must be power of two + + nways = <value> + number of DC ways + + blocksize = <value> + DC block size in bytes; must be power of two + + ustates = <value> + number of DC usage states (2, 3, 4 etc., max is 4) + + load_hitdelay = <value> + number of cycles dc load hit costs + + load_missdelay = <value> + number of cycles dc load miss costs + + store_hitdelay = <value> + number of cycles dc load hit costs + + store_missdelay = <value> + number of cycles dc load miss costs +*/ + +section dc + enabled = 0 + nsets = 512 + nways = 1 + blocksize = 16 + load_hitdelay = 20 + load_missdelay = 20 + store_hitdelay = 20 + store_missdelay = 20 +end + + +/* SIM SECTION + + This section specifies how or1ksim should behave. + + verbose = 0/1 + '0': don't print extra messages + '1': print extra messages + + debug = 0-9 + 0 : no debug messages + 1-9: debug message level. + higher numbers produce more messages + + profile = 0/1 + '0': don't generate profiling file 'sim.profile' + '1': don't generate profiling file 'sim.profile' + + prof_fn = "<filename>" + optional filename for the profiling file. + valid only if 'profile' is set + + mprofile = 0/1 + '0': don't generate memory profiling file 'sim.mprofile' + '1': generate memory profiling file 'sim.mprofile' + + mprof_fn = "<filename>" + optional filename for the memory profiling file. + valid only if 'mprofile' is set + + history = 0/1 + '0': don't track execution flow + '1': track execution flow + Execution flow can be tracked for the simulator's + 'hist' command. Useful for back-trace debugging. + + iprompt = 0/1 + '0': start in <not interactive prompt> (so what do we start in ???) + '1': start in interactive prompt. + + exe_log = 0/1 + '0': don't generate execution log. + '1': generate execution log. + + exe_log = default/hardware/simple/software + type of execution log, default is used when not specified + + exe_log_start = <value> + index of first instruction to start logging, default = 0 + + exe_log_end = <value> + index of last instruction to end logging; not limited, if omitted + + exe_log_marker = <value> + <value> specifies number of instructions before horizontal marker is + printed; if zero, markers are disabled (default) + + exe_log_fn = "<filename>" + filename for the exection log file. + valid only if 'exe_log' is set + + clkcycle = <value>[ps|ns|us|ms] + specifies time measurement for one cycle +*/ + +section sim + verbose = 1 + debug = 0 + profile = 0 + history = 0 + + clkcycle = 10ns +end + + +/* SECTION VAPI + + This section configures the Verification API, used for Advanced + Core Verification. + + enabled = 0/1 + '0': disbable VAPI server + '1': enable/start VAPI server + + server_port = <value> + TCP/IP port to start VAPI server on + + log_enabled = 0/1 + '0': disable VAPI requests logging + '1': enable VAPI requests logging + + hide_device_id = 0/1 + '0': don't log device id (for compatability with old version) + '1': log device id + + + vapi_fn = <filename> + filename for the log file. + valid only if log_enabled is set +*/ + +section VAPI + enabled = 0 + server_port = 9998 + log_enabled = 0 + vapi_log_fn = "vapi.log" +end + + +/* CPU SECTION + + This section specifies various CPU parameters. + + ver = <value> + rev = <value> + specifies version and revision of the CPU used + + upr = <value> + changes the upr register + + sr = <value> + sets the initial Supervision Register value + + superscalar = 0/1 + '0': CPU is scalar + '1': CPU is superscalar + (modify cpu/or32/execute.c to tune superscalar model) + + hazards = 0/1 + '0': don't track data hazards in superscalar CPU + '1': track data hazards in superscalar CPU + If tracked, data hazards can be displayed using the + simulator's 'r' command. + + dependstats = 0/1 + '0': don't calculate inter-instruction dependencies. + '1': calculate inter-instruction dependencies. + If calculated, inter-instruction dependencies can be + displayed using the simulator's 'stat' command. + + sbuf_len = <value> + length of store buffer (<= 256), 0 = disabled +*/ + +section cpu + ver = 0x12 + cfg = 0x00 + rev = 0x01 + /* upr = */ + superscalar = 0 + hazards = 0 + dependstats = 0 + sbuf_len = 0 + hardfloat = 1 +end + + +/* PM SECTION + + This section specifies Power Management parameters + + enabled = 0/1 + '0': disable power management + '1': enable power management +*/ + +section pm + enabled = 0 +end + + +/* BPB SECTION + + This section specifies how branch prediction should behave. + + enabled = 0/1 + '0': disable branch prediction + '1': enable branch prediction + + btic = 0/1 + '0': disable branch target instruction cache model + '1': enable branch target instruction cache model + + sbp_bf_fwd = 0/1 + Static branch prediction for 'l.bf' + '0': don't use forward prediction + '1': use forward prediction + + sbp_bnf_fwd = 0/1 + Static branch prediction for 'l.bnf' + '0': don't use forward prediction + '1': use forward prediction + + hitdelay = <value> + number of cycles bpb hit costs + + missdelay = <value> + number of cycles bpb miss costs +*/ + +section bpb + enabled = 0 + btic = 0 + sbp_bf_fwd = 0 + sbp_bnf_fwd = 0 + hitdelay = 0 + missdelay = 0 +end + + +/* DEBUG SECTION + + This sections specifies how the debug unit should behave. + + enabled = 0/1 + '0': disable debug unit + '1': enable debug unit + + gdb_enabled = 0/1 + '0': don't start gdb server + '1': start gdb server at port 'server_port' + + server_port = <value> + TCP/IP port to start gdb server on + valid only if gdb_enabled is set + + vapi_id = <hex_value> + Used to create "fake" vapi log file containing the JTAG proxy messages. +*/ +section debug + enabled = 1 + rsp_enabled = 1 + rsp_port = 5554 + /*server_port = 9999*/ +end + + +/* MC SECTION + + This section configures the memory controller + + enabled = 0/1 + '0': disable memory controller + '1': enable memory controller + + baseaddr = <hex_value> + address of first MC register + + POC = <hex_value> + Power On Configuration register + + index = <value> + Index of this memory controller amongst all the memory controllers +*/ + +section mc + enabled = 0 + baseaddr = 0x93000000 + POC = 0x00000008 /* Power on configuration register */ + index = 0 +end + + +/* UART SECTION + + This section configures the UARTs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first UART register for this device + + + channel = <channeltype>:<args> + + The channel parameter indicates the source of received UART characters + and the sink for transmitted UART characters. + + The <channeltype> can be either "file", "xterm", "tcp", "fd", or "tty" + (without quotes). + + A) To send/receive characters from a pair of files, use a file + channel: + + channel=file:<rxfile>,<txfile> + + B) To create an interactive terminal window, use an xterm channel: + + channel=xterm:[<xterm_arg>]* + + C) To create a bidirectional tcp socket which one could, for example, + access via telnet, use a tcp channel: + + channel=tcp:<port number> + + D) To cause the UART to read/write from existing numeric file + descriptors, use an fd channel: + + channel=fd:<rx file descriptor num>,<tx file descriptor num> + + E) To connect the UART to a physical serial port, create a tty + channel: + + channel=tty:device=/dev/ttyS0,baud=9600 + + irq = <value> + irq number for this device + + 16550 = 0/1 + '0': this device is a UART16450 + '1': this device is a UART16550 + + jitter = <value> + in msecs... time to block, -1 to disable it + + vapi_id = <hex_value> + VAPI id of this instance +*/ + +section uart + enabled = 1 + baseaddr = 0x90000000 + irq = 2 + channel = "file:uart0.rx,uart0.tx" + /* channel = "tcp:10084" */ + /* channel = "xterm:" */ + jitter = -1 /* async behaviour */ + 16550 = 1 +end + + +/* DMA SECTION + + This section configures the DMAs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first DMA register for this device + + irq = <value> + irq number for this device + + vapi_id = <hex_value> + VAPI id of this instance +*/ + +section dma + enabled = 1 + baseaddr = 0x9a000000 + irq = 11 +end + + +/* ETHERNET SECTION + + This section configures the ETHERNETs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first ethernet register for this device + + dma = <value> + which controller is this ethernet "connected" to + + irq = <value> + ethernet mac IRQ level + + rtx_type = <value> + use 0 - file interface, 1 - socket interface + + rx_channel = <value> + DMA channel used for RX + + tx_channel = <value> + DMA channel used for TX + + rxfile = "<filename>" + filename, where to read data from + + txfile = "<filename>" + filename, where to write data to + + sockif = "<ifacename>" + interface name of ethernet socket + + vapi_id = <hex_value> + VAPI id of this instance +*/ + +section ethernet + enabled = 0 + baseaddr = 0x92000000 + /* dma = 0 */ + irq = 4 + rtx_type = 0 + /* tx_channel = 0 */ + /* rx_channel = 1 */ + /*rxfile = "eth0.rx"*/ + txfile = "eth0.tx" + sockif = "eth0" +end + + +/* GPIO SECTION + + This section configures the GPIOs + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first GPIO register for this device + + irq = <value> + irq number for this device + + base_vapi_id = <hex_value> + first VAPI id of this instance + GPIO uses 8 consecutive VAPI IDs +*/ + +section gpio + enabled = 0 + baseaddr = 0x91000000 + irq = 3 + base_vapi_id = 0x0200 +end + +/* VGA SECTION + + This section configures the VGA/LCD controller + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + address of first VGA register + + irq = <value> + irq number for this device + + refresh_rate = <value> + number of cycles between screen dumps + + filename = "<filename>" + template name for generated names (e.g. "primary" produces "primary0023.bmp") +*/ + +section vga + enabled = 1 + baseaddr = 0x97100000 + irq = 8 + refresh_rate = 100000 + filename = "primary" +end + + +/* TICK TIMER SECTION + + This section configures tick timer + + enabled = 0/1 + whether tick timer is enabled +*/ + +section pic + enabled = 1 + edge_trigger = 1 +end + +/* FB SECTION + + This section configures the frame buffer + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + baseaddr = <hex_value> + base address of frame buffer + + paladdr = <hex_value> + base address of first palette entry + + refresh_rate = <value> + number of cycles between screen dumps + + filename = "<filename>" + template name for generated names (e.g. "primary" produces "primary0023.bmp") +*/ + +section fb + enabled = 1 + baseaddr = 0x97000000 + refresh_rate = 1000000 + filename = "primary" +end + + +/* KBD SECTION + + This section configures the PS/2 compatible keyboard + + baseaddr = <hex_value> + base address of the keyboard device + + rxfile = "<filename>" + filename, where to read data from +*/ + +section kbd + enabled = 1 + irq = 5 + baseaddr = 0x94000000 + rxfile = "kbd.rx" +end + + +/* ATA SECTION + + This section configures the ATA/ATAPI host controller + + baseaddr = <hex_value> + address of first ATA register + + enabled = <0|1> + Enable/disable the peripheral. By default if it is enabled. + + irq = <value> + irq number for this device + + debug = <value> + debug level for ata models. + 0: no debug messages + 1: verbose messages + 3: normal messages (more messages than verbose) + 5: debug messages (normal debug messages) + 7: flow control messages (debug statemachine flows) + 9: low priority message (display everything the code does) + + dev_type0/1 = <value> + ata device 0 type + 0: NO_CONNeCT: none (not connected) + 1: FILE : simulated harddisk + 2: LOCAL : local system harddisk + + dev_file0/1 = "<filename>" + filename for simulated ATA device + valid only if dev_type0 == 1 + + dev_size0/1 = <value> + size of simulated hard-disk (in MBytes) + valid only if dev_type0 == 1 + + dev_packet0/1 = <value> + 0: simulated ATA device does NOT implement PACKET command feature set + 1: simulated ATA device does implement PACKET command feature set + + FIXME: irq number +*/ + +section ata + enabled = 0 + baseaddr = 0x9e000000 + irq = 15 + +end + + diff --exclude=.git --exclude=/gitignore -Naur x264/x264.c x264-or/x264.c --- x264/x264.c 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/x264.c 2009-11-15 17:35:30.000000000 +0100 @@ -40,6 +40,8 @@ #define SetConsoleTitle(t) #endif +//#include "uart.h" + uint8_t *mux_buffer = NULL; int mux_buffer_size = 0; @@ -54,11 +56,11 @@ } typedef struct { - int b_progress; - int i_seek; - hnd_t hin; - hnd_t hout; - FILE *qpfile; + int b_progress; + int i_seek; + hnd_t hin; /* hnd_t is a void* */ + hnd_t hout; + FILE *qpfile; } cli_opt_t; /* input file operation function pointers */ @@ -78,15 +80,28 @@ static int Parse( int argc, char **argv, x264_param_t *param, cli_opt_t *opt ); static int Encode( x264_param_t *param, cli_opt_t *opt ); - +int yuv_data_size; /**************************************************************************** * main: ****************************************************************************/ int main( int argc, char **argv ) { - x264_param_t param; - cli_opt_t opt; - int ret; + char* new_argv ="./x264 --profile=baseline"; + argc = 2; + + x264_param_t param; + cli_opt_t opt; + int ret; + + //uart_init() ; + + extern int _end; // start of free memory + extern int _stack; // end of free memory + + V(fprintf( stderr, "start of heap at 0x%.8x, size of heap: %d bytes\n", + (unsigned int) &_end, (unsigned int) ((unsigned int) &_stack) - ((unsigned int) &_end))); + + //fprintf(stderr,"start\n"); #ifdef PTW32_STATIC_LIB pthread_win32_process_attach_np(); @@ -100,14 +115,90 @@ x264_param_default( ¶m ); - /* Parse command line */ - if( Parse( argc, argv, ¶m, &opt ) < 0 ) - return -1; + // Baseline parameters + param.analyse.b_transform_8x8 = 0; + param.b_cabac = 0; + param.i_cqm_preset = X264_CQM_FLAT; + param.i_bframe = 0; + /* + // Ultrafast preset + param.i_frame_reference = 1; + param.i_scenecut_threshold = 0; + param.b_deblocking_filter = 0; + param.b_cabac = 0; + param.i_bframe = 0; + param.analyse.intra = 0; + param.analyse.inter = 0; + param.analyse.b_transform_8x8 = 0; + param.analyse.i_me_method = X264_ME_DIA; + param.analyse.i_subpel_refine = 0; + param.rc.i_aq_mode = 0; + param.analyse.b_mixed_references = 0; + param.analyse.i_trellis = 0; + param.i_bframe_adaptive = X264_B_ADAPT_NONE; + param.rc.b_mb_tree = 0; + */ + // Fast preset, standard + /* + param.i_frame_reference = 2; + param.analyse.i_subpel_refine = 6; + param.rc.i_lookahead = 30; + */ + // Fast preset, but with 0 lookahead + param.i_frame_reference = 2; + param.analyse.i_subpel_refine = 6; + param.rc.i_lookahead = 0; + + // Verbose (info per frame encoded) + param.i_log_level = X264_LOG_DEBUG; + + // Set a bitrate + param.rc.i_bitrate = 500; // kbps + param.rc.i_rc_method = X264_RC_ABR; + + // A smaller range of QP (qp_max - qp_min) means less startup time for computing motion vectors costs (not sure exactly how this is done) --jb + // Defaults here were i_qp_min = 10, i_qp_max = 51 + param.rc.i_qp_min = 10; + param.rc.i_qp_max = 51; + param.rc.i_qp_step = 4; + + + // Picture height and width + // CIF = 352x288 + param.i_width = 352; + param.i_height = 288; + V(fprintf( stderr, "x264 [info]: %dx%d @ %.2f fps\n", + param.i_width, param.i_height, + (float)param.i_fps_num / (float)param.i_fps_den)); + + // VBV buffer size in kbits + param.rc.i_vbv_buffer_size = (352*288*12*4)/1024; // 4 frames of VBV (?!) + + + + V(printf("video file in memory from 0x%.8x, size %d bytes\n", YUV_DATA_ADDR, YUV_DATA_SIZE)); + /* Must define a few things for use of static CIF data + YUV_DATA_ADDR - address where the data starts + YUV_DATA_SIZE - size of data in bytes + ENC_OUT_ADDR - where we'll store the encoded data + */ + + /* Parse command line */ + Parse( argc, &new_argv, ¶m, &opt ); + //if( Parse( argc, &new_argv, ¶m, &opt ) < 0 ) + //return -1; + + // We specify a place in memory of some YUV CIF data instad of opening a file + init_yuv_dataspace((char *)YUV_DATA_ADDR, (hnd_t *)&opt.hin, ¶m); +#ifdef ENC_OUT_ADDR + p_open_outfile( 0, &opt.hout ); // Setup the out "file" which is really just a spot in memory +#endif + //opt.hout = (void*) ENC_OUT_ADDR; // We specify a place in memory of where we'll dump the encoded data - /* Control-C handler */ - signal( SIGINT, SigIntHandler ); + /* Control-C handler */ + //signal( SIGINT, SigIntHandler ); - ret = Encode( ¶m, &opt ); + ret = Encode( ¶m, &opt ); #ifdef PTW32_STATIC_LIB pthread_win32_thread_detach_np(); @@ -555,10 +646,12 @@ /* Default output file driver */ p_open_outfile = open_file_bsf; p_set_outfile_param = set_param_bsf; - p_write_nalu = write_nalu_bsf; + p_write_nalu = write_nalu_bsf; //--jb p_set_eop = set_eop_bsf; p_close_outfile = close_file_bsf; + return 0; // -- added jb + /* Presets are applied before all other options. */ for( optind = 0;; ) { @@ -978,9 +1071,11 @@ } else { - sscanf( argv[optind++], "%ux%u", ¶m->i_width, ¶m->i_height ); - if( param->i_log_level >= X264_LOG_INFO ) - fprintf( stderr, "x264 [info]: %dx%d @ %.2f fps\n", param->i_width, param->i_height, (double)param->i_fps_num / (double)param->i_fps_den); + sscanf( argv[optind++], "%ux%u", ¶m->i_width, ¶m->i_height ); + if( param->i_log_level >= X264_LOG_INFO ) + fprintf( stderr, "x264 [info]: %dx%d @ %.2f fps\n", + param->i_width, param->i_height, + (double)param->i_fps_num / (double)param->i_fps_den); } } @@ -1127,7 +1222,10 @@ { i_nalu_size = p_write_nalu( hout, nal[i].p_payload, nal[i].i_payload ); if( i_nalu_size < 0 ) + { + fprintf(stderr, "Encode_frame: p_write_nalu() returned i_nalu_size < 0: %d\n", i_nalu_size); return -1; + } i_file += i_nalu_size; } if (i_nal) @@ -1172,12 +1270,15 @@ opt->b_progress &= param->i_log_level < X264_LOG_DEBUG; i_frame_total = p_get_frame_total( opt->hin ); i_frame_total -= opt->i_seek; + if( ( i_frame_total == 0 || param->i_frame_total < i_frame_total ) && param->i_frame_total > 0 ) i_frame_total = param->i_frame_total; + param->i_frame_total = i_frame_total; + V(fprintf(stderr, "Encode: i_frame_total:%d\n", param->i_frame_total)); i_update_interval = i_frame_total ? x264_clip3( i_frame_total / 1000, 1, 10 ) : 10; - + V(fprintf(stderr, "Encode: i_update_interval:%d\n", i_update_interval)); if( ( h = x264_encoder_open( param ) ) == NULL ) { fprintf( stderr, "x264 [error]: x264_encoder_open failed\n" ); @@ -1202,6 +1303,7 @@ i_start = x264_mdate(); + V(fprintf(stderr, "Starting Encoder loop: i_frame_total %d\n", i_frame_total)); /* Encode frames */ for( i_frame = 0, i_file = 0, i_frame_output = 0; b_ctrl_c == 0 && (i_frame < i_frame_total || i_frame_total == 0); ) { diff --exclude=.git --exclude=/gitignore -Naur x264/x264.h x264-or/x264.h --- x264/x264.h 2009-10-25 17:41:22.000000000 +0100 +++ x264-or/x264.h 2009-10-28 15:05:29.000000000 +0100 @@ -24,6 +24,29 @@ #ifndef X264_X264_H #define X264_X264_H +extern int yuv_data_start; +extern int yuv_data_end; +//#define YUV_DATA_ADDR 0x01000000 +#define YUV_DATA_ADDR &yuv_data_start +#define YUV_DATA_SIZE ((int)(&yuv_data_end) - (int)(YUV_DATA_ADDR)) +#define ENC_OUT_ADDR &yuv_data_end + +//#define USE_HARDCODED_FRAME_NUM +#define HARDCODED_FRAME_NUM 10 + + + +/* Enable for verbose output during runtime */ +//#define X264_VERBOSE +#ifdef X264_VERBOSE +#define V(x) x +#else +#define V(x) +#endif + + +extern int yuv_data_size; + #if !defined(_STDINT_H) && !defined(_STDINT_H_) && \ !defined(_INTTYPES_H) && !defined(_INTTYPES_H_) # ifdef _MSC_VER