decode_altivec.c

mips上编译过的mpg 运行正常 环境：AU12

			v8 = vec_mergel(v3,v4);
			
			vsum = vec_sub(v5,v6);
			v9 = vec_sub(v7,v8);
			vsum = vec_add(vsum,v9);
			
			v3 = (vector float)vec_cts(vsum,0);
			v1 = (vector float)vec_cmpgt(vsum,vmax);
			v2 = (vector float)vec_cmplt(vsum,vmin);
			vsample1 = vec_ld(0,samples);
			vsample2 = vec_ld(15,samples);
			v3 = (vector float)vec_packs((vector signed int)v3,(vector signed int)v3);
			v4 = (vector float)vec_perm(vsample1,vsample2,vperm3);
			v5 = (vector float)vec_perm(v3,v4,vperm5);
			v6 = (vector float)vec_perm(vsample2,vsample1,vperm3);
			v7 = (vector float)vec_perm(v5,v6,vperm4);
			v8 = (vector float)vec_perm(v6,v5,vperm4);
			vec_st((vector signed short)v7,15,samples);
			vec_st((vector signed short)v8,0,samples);
			samples += 8;
#ifdef __APPLE__
			v1 = (vector float)vec_sr((vector unsigned int)v1,(vector unsigned int)(31));
			v2 = (vector float)vec_sr((vector unsigned int)v2,(vector unsigned int)(31));
#else
			v1 = (vector float)vec_sr((vector unsigned int)v1,(vector unsigned int){31,31,31,31});
			v2 = (vector float)vec_sr((vector unsigned int)v2,(vector unsigned int){31,31,31,31});
#endif
			v5 = (vector float)vec_add((vector unsigned int)v1,(vector unsigned int)v2);
			vclip = vec_sums((vector signed int)v5,vclip);
		}
		
		{
			real sum;
			sum  = REAL_MUL(window[0x0], b0[0x0]);
			sum += REAL_MUL(window[0x2], b0[0x2]);
			sum += REAL_MUL(window[0x4], b0[0x4]);
			sum += REAL_MUL(window[0x6], b0[0x6]);
			sum += REAL_MUL(window[0x8], b0[0x8]);
			sum += REAL_MUL(window[0xA], b0[0xA]);
			sum += REAL_MUL(window[0xC], b0[0xC]);
			sum += REAL_MUL(window[0xE], b0[0xE]);
			WRITE_SAMPLE(samples,sum,clip);
			b0-=0x10,window-=0x20,samples+=step;
		}
		window += bo1<<1;
		
		vperm1 = vec_lvsl(0,window);
#ifdef __APPLE__
		vperm2 = vec_perm(vperm1,vperm1,(vector unsigned char)(12,13,14,15,8,9,10,11,4,5,6,7,0,1,2,3));
#else
		vperm2 = vec_perm(vperm1,vperm1,(vector unsigned char){12,13,14,15,8,9,10,11,4,5,6,7,0,1,2,3});
#endif
		vperm3 = vec_lvsl(0,samples);
		vperm4 = vec_lvsr(0,samples);
		for (j=3;j;j--)
		{
			vsum = vec_xor(vsum,vsum);
			vsum2 = vec_xor(vsum2,vsum2);
			vsum3 = vec_xor(vsum3,vsum3);
			vsum4 = vec_xor(vsum4,vsum4);
			v1 = vec_ld(-1,window);
			v2 = vec_ld(-16,window);
			v3 = vec_ld(-32,window);
			v4 = vec_ld(-48,window);
			v5 = vec_ld(-64,window);
			v1 = vec_perm(v2,v1,vperm2);
			v6 = vec_ld(0,b0);
			v2 = vec_perm(v3,v2,vperm2);
			v7 = vec_ld(16,b0);
			v3 = vec_perm(v4,v3,vperm2);
			v8 = vec_ld(32,b0);
			v4 = vec_perm(v5,v4,vperm2);
			v9 = vec_ld(48,b0);
			
			vsum = vec_nmsub(v1,v6,vsum);
			vsum = vec_nmsub(v2,v7,vsum);
			vsum = vec_nmsub(v3,v8,vsum);
			vsum = vec_nmsub(v4,v9,vsum);
			
			window -= 32;
			b0 -= 16;
			
			v1 = vec_ld(0,window);
			v2 = vec_ld(-16,window);
			v3 = vec_ld(-32,window);
			v4 = vec_ld(-48,window);
			v5 = vec_ld(-64,window);
			v1 = vec_perm(v2,v1,vperm2);
			v6 = vec_ld(0,b0);
			v2 = vec_perm(v3,v2,vperm2);
			v7 = vec_ld(16,b0);
			v3 = vec_perm(v4,v3,vperm2);
			v8 = vec_ld(32,b0);
			v4 = vec_perm(v5,v4,vperm2);
			v9 = vec_ld(48,b0);
			
			vsum2 = vec_nmsub(v1,v6,vsum2);
			vsum2 = vec_nmsub(v2,v7,vsum2);
			vsum2 = vec_nmsub(v3,v8,vsum2);
			vsum2 = vec_nmsub(v4,v9,vsum2);
			
			window -= 32;
			b0 -= 16;
			
			v1 = vec_ld(0,window);
			v2 = vec_ld(-16,window);
			v3 = vec_ld(-32,window);
			v4 = vec_ld(-48,window);
			v5 = vec_ld(-64,window);
			v1 = vec_perm(v2,v1,vperm2);
			v6 = vec_ld(0,b0);
			v2 = vec_perm(v3,v2,vperm2);
			v7 = vec_ld(16,b0);
			v3 = vec_perm(v4,v3,vperm2);
			v8 = vec_ld(32,b0);
			v4 = vec_perm(v5,v4,vperm2);
			v9 = vec_ld(48,b0);
			
			vsum3 = vec_nmsub(v1,v6,vsum3);
			vsum3 = vec_nmsub(v2,v7,vsum3);
			vsum3 = vec_nmsub(v3,v8,vsum3);
			vsum3 = vec_nmsub(v4,v9,vsum3);
			
			window -= 32;
			b0 -= 16;
			
			v1 = vec_ld(0,window);
			v2 = vec_ld(-16,window);
			v3 = vec_ld(-32,window);
			v4 = vec_ld(-48,window);
			v5 = vec_ld(-64,window);
			v1 = vec_perm(v2,v1,vperm2);
			v6 = vec_ld(0,b0);
			v2 = vec_perm(v3,v2,vperm2);
			v7 = vec_ld(16,b0);
			v3 = vec_perm(v4,v3,vperm2);
			v8 = vec_ld(32,b0);
			v4 = vec_perm(v5,v4,vperm2);
			v9 = vec_ld(48,b0);
			
			vsum4 = vec_nmsub(v1,v6,vsum4);
			vsum4 = vec_nmsub(v2,v7,vsum4);
			vsum4 = vec_nmsub(v3,v8,vsum4);
			vsum4 = vec_nmsub(v4,v9,vsum4);
			
			window -= 32;
			b0 -= 16;
			
			v1 = vec_mergeh(vsum,vsum3);
			v2 = vec_mergeh(vsum2,vsum4);
			v3 = vec_mergel(vsum,vsum3);
			v4 = vec_mergel(vsum2,vsum4);
			v5 = vec_mergeh(v1,v2);
			v6 = vec_mergel(v1,v2);
			v7 = vec_mergeh(v3,v4);
			v8 = vec_mergel(v3,v4);
			
			vsum = vec_add(v5,v6);
			v9 = vec_add(v7,v8);
			vsum = vec_add(vsum,v9);
			
			v3 = (vector float)vec_cts(vsum,0);
			v1 = (vector float)vec_cmpgt(vsum,vmax);
			v2 = (vector float)vec_cmplt(vsum,vmin);
			vsample1 = vec_ld(0,samples);
			vsample2 = vec_ld(15,samples);
			v3 = (vector float)vec_packs((vector signed int)v3,(vector signed int)v3);
			v4 = (vector float)vec_perm(vsample1,vsample2,vperm3);
			v5 = (vector float)vec_perm(v3,v4,vperm5);
			v6 = (vector float)vec_perm(vsample2,vsample1,vperm3);
			v7 = (vector float)vec_perm(v5,v6,vperm4);
			v8 = (vector float)vec_perm(v6,v5,vperm4);
			vec_st((vector signed short)v7,15,samples);
			vec_st((vector signed short)v8,0,samples);
			samples += 8;
#ifdef __APPLE__
			v1 = (vector float)vec_sr((vector unsigned int)v1,(vector unsigned int)(31));
			v2 = (vector float)vec_sr((vector unsigned int)v2,(vector unsigned int)(31));
#else
			v1 = (vector float)vec_sr((vector unsigned int)v1,(vector unsigned int){31,31,31,31});
			v2 = (vector float)vec_sr((vector unsigned int)v2,(vector unsigned int){31,31,31,31});
#endif
			v5 = (vector float)vec_add((vector unsigned int)v1,(vector unsigned int)v2);
			vclip = vec_sums((vector signed int)v5,vclip);
		}
#ifdef __APPLE__
		vperm5 = (vector unsigned char)(0,1,18,19,2,3,22,23,4,5,26,27,28,29,30,31);
#else
		vperm5 = (vector unsigned char){0,1,18,19,2,3,22,23,4,5,26,27,28,29,30,31};
#endif
		{
			vsum = vec_xor(vsum,vsum);
			vsum2 = vec_xor(vsum2,vsum2);
			vsum3 = vec_xor(vsum3,vsum3);
			vsum4 = vec_xor(vsum4,vsum4);
			v1 = vec_ld(-1,window);
			v2 = vec_ld(-16,window);
			v3 = vec_ld(-32,window);
			v4 = vec_ld(-48,window);
			v5 = vec_ld(-64,window);
			v1 = vec_perm(v2,v1,vperm2);
			v6 = vec_ld(0,b0);
			v2 = vec_perm(v3,v2,vperm2);
			v7 = vec_ld(16,b0);
			v3 = vec_perm(v4,v3,vperm2);
			v8 = vec_ld(32,b0);
			v4 = vec_perm(v5,v4,vperm2);
			v9 = vec_ld(48,b0);
			
			vsum = vec_nmsub(v1,v6,vsum);
			vsum = vec_nmsub(v2,v7,vsum);
			vsum = vec_nmsub(v3,v8,vsum);
			vsum = vec_nmsub(v4,v9,vsum);
			
			window -= 32;
			b0 -= 16;
			
			v1 = vec_ld(0,window);
			v2 = vec_ld(-16,window);
			v3 = vec_ld(-32,window);
			v4 = vec_ld(-48,window);
			v5 = vec_ld(-64,window);
			v1 = vec_perm(v2,v1,vperm2);
			v6 = vec_ld(0,b0);
			v2 = vec_perm(v3,v2,vperm2);
			v7 = vec_ld(16,b0);
			v3 = vec_perm(v4,v3,vperm2);
			v8 = vec_ld(32,b0);
			v4 = vec_perm(v5,v4,vperm2);
			v9 = vec_ld(48,b0);
			
			vsum2 = vec_nmsub(v1,v6,vsum2);
			vsum2 = vec_nmsub(v2,v7,vsum2);
			vsum2 = vec_nmsub(v3,v8,vsum2);
			vsum2 = vec_nmsub(v4,v9,vsum2);
			
			window -= 32;
			b0 -= 16;
			
			v1 = vec_ld(0,window);
			v2 = vec_ld(-16,window);
			v3 = vec_ld(-32,window);
			v4 = vec_ld(-48,window);
			v5 = vec_ld(-64,window);
			v1 = vec_perm(v2,v1,vperm2);
			v6 = vec_ld(0,b0);
			v2 = vec_perm(v3,v2,vperm2);
			v7 = vec_ld(16,b0);
			v3 = vec_perm(v4,v3,vperm2);
			v8 = vec_ld(32,b0);
			v4 = vec_perm(v5,v4,vperm2);
			v9 = vec_ld(48,b0);
			
			vsum3 = vec_nmsub(v1,v6,vsum3);
			vsum3 = vec_nmsub(v2,v7,vsum3);
			vsum3 = vec_nmsub(v3,v8,vsum3);
			vsum3 = vec_nmsub(v4,v9,vsum3);
			
			v1 = vec_mergeh(vsum,vsum3);
			v2 = vec_mergeh(vsum2,vsum2);
			v3 = vec_mergel(vsum,vsum3);
			v4 = vec_mergel(vsum2,vsum2);
			v5 = vec_mergeh(v1,v2);
			v6 = vec_mergel(v1,v2);
			v7 = vec_mergeh(v3,v4);
			v8 = vec_mergel(v3,v4);
			
			vsum = vec_add(v5,v6);
			v9 = vec_add(v7,v8);
			vsum = vec_add(vsum,v9);
			
			v3 = (vector float)vec_cts(vsum,0);
			v1 = (vector float)vec_cmpgt(vsum,vmax);
			v2 = (vector float)vec_cmplt(vsum,vmin);
			vsample1 = vec_ld(0,samples);
			vsample2 = vec_ld(15,samples);
			v3 = (vector float)vec_packs((vector signed int)v3,(vector signed int)v3);
			v4 = (vector float)vec_perm(vsample1,vsample2,vperm3);
			v5 = (vector float)vec_perm(v3,v4,vperm5);
			v6 = (vector float)vec_perm(vsample2,vsample1,vperm3);
			v7 = (vector float)vec_perm(v5,v6,vperm4);
			v8 = (vector float)vec_perm(v6,v5,vperm4);
			vec_st((vector signed short)v7,15,samples);
			vec_st((vector signed short)v8,0,samples);
			samples += 6;
#ifdef __APPLE__
			v1 = (vector float)vec_sr((vector unsigned int)v1,(vector unsigned int)(31,31,31,32));
			v2 = (vector float)vec_sr((vector unsigned int)v2,(vector unsigned int)(31,31,31,32));
#else
			v1 = (vector float)vec_sr((vector unsigned int)v1,(vector unsigned int){31,31,31,32});
			v2 = (vector float)vec_sr((vector unsigned int)v2,(vector unsigned int){31,31,31,32});
#endif
			v5 = (vector float)vec_add((vector unsigned int)v1,(vector unsigned int)v2);
			vclip = vec_sums((vector signed int)v5,vclip);
			vec_st(vclip,0,clip_tmp);
			clip += clip_tmp[3];
		}
  }
  if(final) fr->buffer.fill += 128;

  return clip;
}