altivec_mlib.c

基于linux的DVD播放器程序

			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,4\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += stride, ref_block += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,3\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += stride, ref_block += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,4\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += stride, ref_block += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,3\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += stride, ref_block += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,4\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += stride, ref_block += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,3\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += stride, ref_block += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,4\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
	} else {
		int i0 = 0, i1 = 4;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
	}
}

void 
mlib_VideoCopyRefAve_U8_U8_8x4(uint8_t *curr_block,
			       const uint8_t *ref_block,
			       int32_t stride)
{
	ASSERT(((int)curr_block & 7) == 0);

	if ((((int)ref_block ^ (int)curr_block) & 15) != 0) {
		const int i0 = 0, i1 = 16, i2 = 4;
		asm(""
			"lvsl 3,%1,%2\n"
			"lvsl 4,%1,%3\n"
			"lvsr 5,%0,%2\n"
			"lvsr 6,%0,%3\n"
			"vperm 3,3,3,5\n"
			"vperm 4,4,4,6\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i0 + stride));
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,3\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += stride, ref_block += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,4\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += stride, ref_block += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,3\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += stride, ref_block += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"lvx 2,%0,%2\n"
			"vperm 0,0,1,4\n"
			"vavgub 0,0,2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
	} else {
		int i0 = 0, i1 = 4;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%0,%2\n"
			"lvx 1,%1,%2\n"
			"vavgub 0,0,1\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
	}
}

void
mlib_VideoCopyRef_U8_U8_16x16_multiple(uint8_t *curr_block,
				       const uint8_t *ref_block,
				       int32_t stride,
				       int32_t count)
{
	ASSERT(((int)curr_block & 15) == 0);
	ASSERT(((int)ref_block & 15) == 0);
	
	while (count--) {
		int i0 = 0;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));

		curr_block += 16, ref_block += 16;
	}
}

void
mlib_VideoCopyRef_U8_U8_16x16(uint8_t *curr_block,
			      const uint8_t *ref_block,
			      int32_t stride)
{
	ASSERT(((int)curr_block & 15) == 0);
	
	if (((int)ref_block & 15) != 0) {
		int i0 = 0, i1 = 16;
		asm(""
			"lvsl 2,%0,%1\n"
		"" : : "b" (ref_block), "b" (i0));
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
		i0 += stride, i1 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
	} else {
		int i0 = 0;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
		i0 += stride;
		asm(""
			"lvx 0,%1,%2\n"
			"stvx 0,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0));
	}
}

void 
mlib_VideoCopyRef_U8_U8_16x8(uint8_t *curr_block,