altivec_mlib.c

基于linux的DVD播放器程序

			     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));
	} 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));
	}
}

void 
mlib_VideoCopyRef_U8_U8_8x8_multiple(uint8_t *curr_block,
				     uint8_t *ref_block,
				     int32_t stride,
				     int32_t count)
{
	ASSERT(((int)curr_block & 7) == 0);

	while (count--) {
		int i0 = 0, i1 = 4;
		asm(""
			"lvx 0,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));

		curr_block += 8, ref_block += 8;
	}
}

void 
mlib_VideoCopyRef_U8_U8_8x8(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 2,%1,%2\n"
			"lvsl 3,%1,%3\n"
			"lvsr 4,%0,%2\n"
			"lvsr 5,%0,%3\n"
			"vperm 2,2,2,4\n"
			"vperm 3,3,3,5\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i0 + stride));
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\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"
			"vperm 0,0,1,3\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"
			"vperm 0,0,1,2\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"
			"vperm 0,0,1,3\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"
			"vperm 0,0,1,2\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"
			"vperm 0,0,1,3\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"
			"vperm 0,0,1,2\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"
			"vperm 0,0,1,3\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\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_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 2,%1,%2\n"
			"lvsl 3,%1,%3\n"
			"lvsr 4,%0,%2\n"
			"lvsr 5,%0,%3\n"
			"vperm 2,2,2,4\n"
			"vperm 3,3,3,5\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i0 + stride));
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 0,0,1,2\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"
			"vperm 0,0,1,3\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"
			"vperm 0,0,1,2\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"
			"vperm 0,0,1,3\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,%1,%2\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,%1,%2\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,%1,%2\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,%1,%2\n"
			"stvewx 0,%0,%2\n"
			"stvewx 0,%0,%3\n"
		"" : : "b" (curr_block), "b" (ref_block), "b" (i0), "b" (i1));
	}
}

void
mlib_VideoInterpAveX_U8_U8_16x16(uint8_t *curr_block,
                                 const uint8_t *ref_block,
                                 int32_t frame_stride,
                                 int32_t field_stride)
{
	int i;
	int i0 = 0, i1 = 16;

	ASSERT(((int)curr_block & 15) == 0);

	asm(""
		"vspltisb 0,1\n"
		"lvsl 2,%0,%1\n"
		"vaddubs 3,2,0\n"
	"" : : "b" (ref_block), "b" (i0));
	for (i = 0; i < 16; i++) {
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 4,0,1,2\n"
			"vperm 5,0,1,3\n"
			"lvx 6,%0,%2\n"
			"vavgub 4,4,5\n"
			"vavgub 4,4,6\n"
			"stvx 4,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1));
		i0 += frame_stride, i1 += frame_stride;
	}
}

void 
mlib_VideoInterpAveX_U8_U8_16x8(uint8_t *curr_block,
				const uint8_t *ref_block,
				int32_t frame_stride,
				int32_t field_stride)
{
	int i;
	int i0 = 0, i1 = 16;

	ASSERT(((int)curr_block & 15) == 0);

	asm(""
		"vspltisb 0,1\n"
		"lvsl 2,%0,%1\n"
		"vaddubs 3,2,0\n"
	"" : : "b" (ref_block), "b" (i0), "b" (i1));
	for (i = 0; i < 8; i++) {
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 4,0,1,2\n"
			"vperm 5,0,1,3\n"
			"lvx 6,%0,%2\n"
			"vavgub 4,4,5\n"
			"vavgub 4,4,6\n"
			"stvx 4,%0,%2\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1));
		i0 += frame_stride, i1 += frame_stride;
	}
}

void 
mlib_VideoInterpAveX_U8_U8_8x8(uint8_t *curr_block,
			       const uint8_t *ref_block,
			       int32_t frame_stride,
			       int32_t field_stride)
{
	int i;
	const int i0 = 0, i1 = 16, i2 = 4;

	ASSERT(((int)curr_block & 7) == 0);

	asm(""
		"vspltisb 0,1\n"
		"lvsl 2,%1,%2\n"
		"lvsr 3,%0,%2\n"
		"lvsl 4,%1,%3\n"
		"lvsr 5,%0,%3\n"
		"vperm 2,2,2,3\n"
		"vperm 4,4,4,5\n"
		"vaddubs 3,2,0\n"
		"vaddubs 5,4,0\n"
	"" : : "b" (curr_block), "b" (ref_block),
	      "b" (i0), "b" (i0 + frame_stride));
	for (i = 0; i < 4; i++) {
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 6,0,1,2\n"
			"vperm 7,0,1,3\n"
			"lvx 8,%0,%2\n"
			"vavgub 6,6,7\n"
			"vavgub 6,6,8\n"
			"stvewx 6,%0,%2\n"
			"stvewx 6,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += frame_stride, ref_block += frame_stride;
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 6,0,1,4\n"
			"vperm 7,0,1,5\n"
			"lvx 8,%0,%2\n"
			"vavgub 6,6,7\n"
			"vavgub 6,6,8\n"
			"stvewx 6,%0,%2\n"
			"stvewx 6,%0,%4\n"
		"" : : "b" (curr_block), "b" (ref_block),
		      "b" (i0), "b" (i1), "b" (i2));
		curr_block += frame_stride, ref_block += frame_stride;
	}
}

void 
mlib_VideoInterpAveX_U8_U8_8x4(uint8_t *curr_block,
			       const uint8_t *ref_block,
			       int32_t frame_stride,
			       int32_t field_stride)
{
	int i;
	const int i0 = 0, i1 = 16, i2 = 4;

	ASSERT(((int)curr_block & 7) == 0);

	asm(""
		"vspltisb 0,1\n"
		"lvsl 2,%1,%2\n"
		"lvsr 3,%0,%2\n"
		"lvsl 4,%1,%3\n"
		"lvsr 5,%0,%3\n"
		"vperm 2,2,2,3\n"
		"vperm 4,4,4,5\n"
		"vaddubs 3,2,0\n"
		"vaddubs 5,4,0\n"
	"" : : "b" (curr_block), "b" (ref_block),
	      "b" (i0), "b" (i0 + frame_stride));
	for (i = 0; i < 2; i++) {
		asm(""
			"lvx 0,%1,%2\n"
			"lvx 1,%1,%3\n"
			"vperm 6,0,1,2\n"
			"vperm 7,0,1,3\n"
			"lvx 8,%0,%2\n"
			"vavgub 6,6,7\n"
			"vavgub 6,6,8\n"
			"stvewx 6,%0,%2\n"