Libav
|
00001 /* 00002 * GMC (Global Motion Compensation) 00003 * AltiVec-enabled 00004 * Copyright (c) 2003 Romain Dolbeau <romain@dolbeau.org> 00005 * 00006 * This file is part of FFmpeg. 00007 * 00008 * FFmpeg is free software; you can redistribute it and/or 00009 * modify it under the terms of the GNU Lesser General Public 00010 * License as published by the Free Software Foundation; either 00011 * version 2.1 of the License, or (at your option) any later version. 00012 * 00013 * FFmpeg is distributed in the hope that it will be useful, 00014 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00015 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 00016 * Lesser General Public License for more details. 00017 * 00018 * You should have received a copy of the GNU Lesser General Public 00019 * License along with FFmpeg; if not, write to the Free Software 00020 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00021 */ 00022 00023 #include "libavcodec/dsputil.h" 00024 #include "dsputil_ppc.h" 00025 #include "util_altivec.h" 00026 #include "types_altivec.h" 00027 #include "dsputil_altivec.h" 00028 00029 /* 00030 altivec-enhanced gmc1. ATM this code assume stride is a multiple of 8, 00031 to preserve proper dst alignment. 00032 */ 00033 #define GMC1_PERF_COND (h==8) 00034 void gmc1_altivec(uint8_t *dst /* align 8 */, uint8_t *src /* align1 */, int stride, int h, int x16, int y16, int rounder) 00035 { 00036 POWERPC_PERF_DECLARE(altivec_gmc1_num, GMC1_PERF_COND); 00037 const DECLARE_ALIGNED(16, unsigned short, rounder_a) = rounder; 00038 const DECLARE_ALIGNED(16, unsigned short, ABCD)[8] = 00039 { 00040 (16-x16)*(16-y16), /* A */ 00041 ( x16)*(16-y16), /* B */ 00042 (16-x16)*( y16), /* C */ 00043 ( x16)*( y16), /* D */ 00044 0, 0, 0, 0 /* padding */ 00045 }; 00046 register const vector unsigned char vczero = (const vector unsigned char)vec_splat_u8(0); 00047 register const vector unsigned short vcsr8 = (const vector unsigned short)vec_splat_u16(8); 00048 register vector unsigned char dstv, dstv2, src_0, src_1, srcvA, srcvB, srcvC, srcvD; 00049 register vector unsigned short Av, Bv, Cv, Dv, rounderV, tempA, tempB, tempC, tempD; 00050 int i; 00051 unsigned long dst_odd = (unsigned long)dst & 0x0000000F; 00052 unsigned long src_really_odd = (unsigned long)src & 0x0000000F; 00053 00054 00055 POWERPC_PERF_START_COUNT(altivec_gmc1_num, GMC1_PERF_COND); 00056 00057 tempA = vec_ld(0, (unsigned short*)ABCD); 00058 Av = vec_splat(tempA, 0); 00059 Bv = vec_splat(tempA, 1); 00060 Cv = vec_splat(tempA, 2); 00061 Dv = vec_splat(tempA, 3); 00062 00063 rounderV = vec_splat((vec_u16)vec_lde(0, &rounder_a), 0); 00064 00065 // we'll be able to pick-up our 9 char elements 00066 // at src from those 32 bytes 00067 // we load the first batch here, as inside the loop 00068 // we can re-use 'src+stride' from one iteration 00069 // as the 'src' of the next. 00070 src_0 = vec_ld(0, src); 00071 src_1 = vec_ld(16, src); 00072 srcvA = vec_perm(src_0, src_1, vec_lvsl(0, src)); 00073 00074 if (src_really_odd != 0x0000000F) { 00075 // if src & 0xF == 0xF, then (src+1) is properly aligned 00076 // on the second vector. 00077 srcvB = vec_perm(src_0, src_1, vec_lvsl(1, src)); 00078 } else { 00079 srcvB = src_1; 00080 } 00081 srcvA = vec_mergeh(vczero, srcvA); 00082 srcvB = vec_mergeh(vczero, srcvB); 00083 00084 for(i=0; i<h; i++) { 00085 dst_odd = (unsigned long)dst & 0x0000000F; 00086 src_really_odd = (((unsigned long)src) + stride) & 0x0000000F; 00087 00088 dstv = vec_ld(0, dst); 00089 00090 // we we'll be able to pick-up our 9 char elements 00091 // at src + stride from those 32 bytes 00092 // then reuse the resulting 2 vectors srvcC and srcvD 00093 // as the next srcvA and srcvB 00094 src_0 = vec_ld(stride + 0, src); 00095 src_1 = vec_ld(stride + 16, src); 00096 srcvC = vec_perm(src_0, src_1, vec_lvsl(stride + 0, src)); 00097 00098 if (src_really_odd != 0x0000000F) { 00099 // if src & 0xF == 0xF, then (src+1) is properly aligned 00100 // on the second vector. 00101 srcvD = vec_perm(src_0, src_1, vec_lvsl(stride + 1, src)); 00102 } else { 00103 srcvD = src_1; 00104 } 00105 00106 srcvC = vec_mergeh(vczero, srcvC); 00107 srcvD = vec_mergeh(vczero, srcvD); 00108 00109 00110 // OK, now we (finally) do the math :-) 00111 // those four instructions replaces 32 int muls & 32 int adds. 00112 // isn't AltiVec nice ? 00113 tempA = vec_mladd((vector unsigned short)srcvA, Av, rounderV); 00114 tempB = vec_mladd((vector unsigned short)srcvB, Bv, tempA); 00115 tempC = vec_mladd((vector unsigned short)srcvC, Cv, tempB); 00116 tempD = vec_mladd((vector unsigned short)srcvD, Dv, tempC); 00117 00118 srcvA = srcvC; 00119 srcvB = srcvD; 00120 00121 tempD = vec_sr(tempD, vcsr8); 00122 00123 dstv2 = vec_pack(tempD, (vector unsigned short)vczero); 00124 00125 if (dst_odd) { 00126 dstv2 = vec_perm(dstv, dstv2, vcprm(0,1,s0,s1)); 00127 } else { 00128 dstv2 = vec_perm(dstv, dstv2, vcprm(s0,s1,2,3)); 00129 } 00130 00131 vec_st(dstv2, 0, dst); 00132 00133 dst += stride; 00134 src += stride; 00135 } 00136 00137 POWERPC_PERF_STOP_COUNT(altivec_gmc1_num, GMC1_PERF_COND); 00138 }