libavcodec.hg: ppc/dsputil_altivec.c comparison

comparison ppc/dsputil_altivec.c @ 2979:bfabfdf9ce55 libavcodec

COSMETICS: tabs --> spaces, some prettyprinting

author	diego
date	Thu, 22 Dec 2005 01:10:11 +0000
parents	ef2149182f1c
children	0b546eab515d

comparison

equal deleted inserted replaced

-:403183bbb505
+:bfabfdf9ce55
 sad = (vector unsigned int)vec_splat_u32(0);
 for(i=0;i<h;i++) {
 /*
 Read unaligned pixels into our vectors. The vectors are as follows:
 pix1v: pix1[0]-pix1[15]
-pix2v: pix2[0]-pix2[15]	pix2iv: pix2[1]-pix2[16]
+pix2v: pix2[0]-pix2[15]      pix2iv: pix2[1]-pix2[16]
 */
 tv = (vector unsigned char *) pix1;
 pix1v = vec_perm(tv[0], tv[1], vec_lvsl(0, pix1));
 tv = (vector unsigned char *) &pix2[0];
 Due to the fact that pix3 = pix2 + line_size, the pix3 of one
 iteration becomes pix2 in the next iteration. We can use this
 fact to avoid a potentially expensive unaligned read, as well
 as some splitting, and vector addition each time around the loop.
 Read unaligned pixels into our vectors. The vectors are as follows:
-pix2v: pix2[0]-pix2[15]	pix2iv: pix2[1]-pix2[16]
+pix2v: pix2[0]-pix2[15]  pix2iv: pix2[1]-pix2[16]
 Split the pixel vectors into shorts
 */
 tv = (vector unsigned char *) &pix2[0];
 pix2v = vec_perm(tv[0], tv[1], vec_lvsl(0, &pix2[0]));
 for(i=0;i<h;i++) {
 /*
 Read unaligned pixels into our vectors. The vectors are as follows:
 pix1v: pix1[0]-pix1[15]
-pix3v: pix3[0]-pix3[15]	pix3iv: pix3[1]-pix3[16]
+pix3v: pix3[0]-pix3[15]      pix3iv: pix3[1]-pix3[16]
 */
 tv = (vector unsigned char *) pix1;
 pix1v = vec_perm(tv[0], tv[1], vec_lvsl(0, pix1));
 tv = (vector unsigned char *) &pix3[0];
 sad = (vector unsigned int)vec_splat_u32(0);
 for(i=0;i<h;i++) {
-	/* Read potentially unaligned pixels into t1 and t2 */
+/* Read potentially unaligned pixels into t1 and t2 */
 perm1 = vec_lvsl(0, pix1);
 pix1v = (vector unsigned char *) pix1;
 perm2 = vec_lvsl(0, pix2);
 pix2v = (vector unsigned char *) pix2;
 t1 = vec_perm(pix1v[0], pix1v[1], perm1);
 t2 = vec_perm(pix2v[0], pix2v[1], perm2);
-	/* Calculate a sum of abs differences vector */
+/* Calculate a sum of abs differences vector */
 t3 = vec_max(t1, t2);
 t4 = vec_min(t1, t2);
 t5 = vec_sub(t3, t4);
-	/* Add each 4 pixel group together and put 4 results into sad */
+/* Add each 4 pixel group together and put 4 results into sad */
 sad = vec_sum4s(t5, sad);
 pix1 += line_size;
 pix2 += line_size;
 }
 sad = (vector unsigned int)vec_splat_u32(0);
 permclear = (vector unsigned char)AVV(255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0);
 for(i=0;i<h;i++) {
-	/* Read potentially unaligned pixels into t1 and t2
+/* Read potentially unaligned pixels into t1 and t2
-	   Since we're reading 16 pixels, and actually only want 8,
+Since we're reading 16 pixels, and actually only want 8,
-	   mask out the last 8 pixels. The 0s don't change the sum. */
+mask out the last 8 pixels. The 0s don't change the sum. */
 perm1 = vec_lvsl(0, pix1);
 pix1v = (vector unsigned char *) pix1;
 perm2 = vec_lvsl(0, pix2);
 pix2v = (vector unsigned char *) pix2;
 t1 = vec_and(vec_perm(pix1v[0], pix1v[1], perm1), permclear);
 t2 = vec_and(vec_perm(pix2v[0], pix2v[1], perm2), permclear);
-	/* Calculate a sum of abs differences vector */
+/* Calculate a sum of abs differences vector */
 t3 = vec_max(t1, t2);
 t4 = vec_min(t1, t2);
 t5 = vec_sub(t3, t4);
-	/* Add each 4 pixel group together and put 4 results into sad */
+/* Add each 4 pixel group together and put 4 results into sad */
 sad = vec_sum4s(t5, sad);
 pix1 += line_size;
 pix2 += line_size;
 }
 permclear = (vector unsigned char)AVV(255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0);
 for(i=0;i<h;i++) {
-	/* Read potentially unaligned pixels into t1 and t2
+/* Read potentially unaligned pixels into t1 and t2
-	   Since we're reading 16 pixels, and actually only want 8,
+Since we're reading 16 pixels, and actually only want 8,
-	   mask out the last 8 pixels. The 0s don't change the sum. */
+mask out the last 8 pixels. The 0s don't change the sum. */
 perm1 = vec_lvsl(0, pix1);
 pix1v = (vector unsigned char *) pix1;
 perm2 = vec_lvsl(0, pix2);
 pix2v = (vector unsigned char *) pix2;
 t1 = vec_and(vec_perm(pix1v[0], pix1v[1], perm1), permclear);
 /*
 Since we want to use unsigned chars, we can take advantage
 of the fact that abs(a-b)^2 = (a-b)^2.
 */
-	/* Calculate abs differences vector */
+/* Calculate abs differences vector */
 t3 = vec_max(t1, t2);
 t4 = vec_min(t1, t2);
 t5 = vec_sub(t3, t4);
 /* Square the values and add them to our sum */
 vector signed int sumsqr;
 sum = (vector unsigned int)vec_splat_u32(0);
 for(i=0;i<h;i++) {
-	/* Read potentially unaligned pixels into t1 and t2 */
+/* Read potentially unaligned pixels into t1 and t2 */
 perm1 = vec_lvsl(0, pix1);
 pix1v = (vector unsigned char *) pix1;
 perm2 = vec_lvsl(0, pix2);
 pix2v = (vector unsigned char *) pix2;
 t1 = vec_perm(pix1v[0], pix1v[1], perm1);
 /*
 Since we want to use unsigned chars, we can take advantage
 of the fact that abs(a-b)^2 = (a-b)^2.
 */
-	/* Calculate abs differences vector */
+/* Calculate abs differences vector */
 t3 = vec_max(t1, t2);
 t4 = vec_min(t1, t2);
 t5 = vec_sub(t3, t4);
 /* Square the values and add them to our sum */
 int s __attribute__((aligned(16)));
 sad = (vector unsigned int)vec_splat_u32(0);
 for (i = 0; i < 16; i++) {
-	/* Read the potentially unaligned 16 pixels into t1 */
+/* Read the potentially unaligned 16 pixels into t1 */
 perm = vec_lvsl(0, pix);
 pixv = (vector unsigned char *) pix;
 t1 = vec_perm(pixv[0], pixv[1], perm);
-	/* Add each 4 pixel group together and put 4 results into sad */
+/* Add each 4 pixel group together and put 4 results into sad */
 sad = vec_sum4s(t1, sad);
 pix += line_size;
 }
 AVV(0x08, 0x09, 0x0A, 0x0B,
 0x0C, 0x0D, 0x0E, 0x0F,
 0x00, 0x01, 0x02, 0x03,
 0x04, 0x05, 0x06, 0x07);
-#define ONEITERBUTTERFLY(i, res)					\
+#define ONEITERBUTTERFLY(i, res)                                        \
-{									\
+{                                                                   \
-register vector unsigned char src1, src2, srcO;		       	\
+register vector unsigned char src1, src2, srcO;                   \
-register vector unsigned char dst1, dst2, dstO;		       	\
+register vector unsigned char dst1, dst2, dstO;                   \
-src1 = vec_ld(stride * i, src);					\
+src1 = vec_ld(stride * i, src);                                   \
-if ((((stride * i) + (unsigned long)src) & 0x0000000F) > 8)	\
+if ((((stride * i) + (unsigned long)src) & 0x0000000F) > 8)       \
-	src2 = vec_ld((stride * i) + 16, src);				\
+src2 = vec_ld((stride * i) + 16, src);                          \
-srcO = vec_perm(src1, src2, vec_lvsl(stride * i, src));		\
+srcO = vec_perm(src1, src2, vec_lvsl(stride * i, src));           \
-dst1 = vec_ld(stride * i, dst);					\
+dst1 = vec_ld(stride * i, dst);                                   \
-if ((((stride * i) + (unsigned long)dst) & 0x0000000F) > 8)	\
+if ((((stride * i) + (unsigned long)dst) & 0x0000000F) > 8)       \
-	dst2 = vec_ld((stride * i) + 16, dst);				\
+dst2 = vec_ld((stride * i) + 16, dst);                          \
-dstO = vec_perm(dst1, dst2, vec_lvsl(stride * i, dst));		\
+dstO = vec_perm(dst1, dst2, vec_lvsl(stride * i, dst));           \
-/* promote the unsigned chars to signed shorts */			\
+/* promote the unsigned chars to signed shorts */                 \
-/* we're in the 8x8 function, we only care for the first 8 */	\
+/* we're in the 8x8 function, we only care for the first 8 */     \
-register vector signed short srcV =			       	\
+register vector signed short srcV =                               \
-	(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
+(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
-register vector signed short dstV =			       	\
+register vector signed short dstV =                               \
-	(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
+(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
-/* substractions inside the first butterfly */			\
+/* substractions inside the first butterfly */                    \
-register vector signed short but0 = vec_sub(srcV, dstV);	       	\
+register vector signed short but0 = vec_sub(srcV, dstV);          \
-register vector signed short op1 = vec_perm(but0, but0, perm1);  	\
+register vector signed short op1 = vec_perm(but0, but0, perm1);   \
-register vector signed short but1 = vec_mladd(but0, vprod1, op1);	\
+register vector signed short but1 = vec_mladd(but0, vprod1, op1); \
-register vector signed short op2 = vec_perm(but1, but1, perm2);  	\
+register vector signed short op2 = vec_perm(but1, but1, perm2);   \
-register vector signed short but2 = vec_mladd(but1, vprod2, op2);	\
+register vector signed short but2 = vec_mladd(but1, vprod2, op2); \
-register vector signed short op3 = vec_perm(but2, but2, perm3);  	\
+register vector signed short op3 = vec_perm(but2, but2, perm3);   \
-res = vec_mladd(but2, vprod3, op3);				\
+res = vec_mladd(but2, vprod3, op3);                               \
 }
 ONEITERBUTTERFLY(0, temp0);
 ONEITERBUTTERFLY(1, temp1);
 ONEITERBUTTERFLY(2, temp2);
 ONEITERBUTTERFLY(3, temp3);
 AVV(0x08, 0x09, 0x0A, 0x0B,
 0x0C, 0x0D, 0x0E, 0x0F,
 0x00, 0x01, 0x02, 0x03,
 0x04, 0x05, 0x06, 0x07);
-#define ONEITERBUTTERFLY(i, res1, res2)					\
+#define ONEITERBUTTERFLY(i, res1, res2)                                 \
-{									\
+{                                                                   \
 register vector unsigned char src1 asm ("v22"), src2 asm ("v23"); \
 register vector unsigned char dst1 asm ("v24"), dst2 asm ("v25"); \
-src1 = vec_ld(stride * i, src);					\
+src1 = vec_ld(stride * i, src);                                   \
-src2 = vec_ld((stride * i) + 16, src);				\
+src2 = vec_ld((stride * i) + 16, src);                            \
 register vector unsigned char srcO asm ("v22") = vec_perm(src1, src2, vec_lvsl(stride * i, src)); \
-dst1 = vec_ld(stride * i, dst);					\
+dst1 = vec_ld(stride * i, dst);                                   \
-dst2 = vec_ld((stride * i) + 16, dst);				\
+dst2 = vec_ld((stride * i) + 16, dst);                            \
 register vector unsigned char dstO asm ("v23") = vec_perm(dst1, dst2, vec_lvsl(stride * i, dst)); \
-/* promote the unsigned chars to signed shorts */			\
+/* promote the unsigned chars to signed shorts */                 \
 register vector signed short srcV asm ("v24") =                   \
-	(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
+(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
 register vector signed short dstV asm ("v25") =                   \
-	(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
+(vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
 register vector signed short srcW asm ("v26") =                   \
-	(vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)srcO); \
+(vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)srcO); \
 register vector signed short dstW asm ("v27") =                   \
-	(vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)dstO); \
+(vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)dstO); \
-/* substractions inside the first butterfly */			\
+/* substractions inside the first butterfly */                    \
 register vector signed short but0 asm ("v28") = vec_sub(srcV, dstV); \
 register vector signed short but0S asm ("v29") = vec_sub(srcW, dstW); \
 register vector signed short op1 asm ("v30") = vec_perm(but0, but0, perm1); \
 register vector signed short but1 asm ("v22") = vec_mladd(but0, vprod1, op1); \
 register vector signed short op1S asm ("v23") = vec_perm(but0S, but0S, perm1); \
 register vector signed short op2 asm ("v25") = vec_perm(but1, but1, perm2); \
 register vector signed short but2 asm ("v26") = vec_mladd(but1, vprod2, op2); \
 register vector signed short op2S asm ("v27") = vec_perm(but1S, but1S, perm2); \
 register vector signed short but2S asm ("v28") = vec_mladd(but1S, vprod2, op2S); \
 register vector signed short op3 asm ("v29") = vec_perm(but2, but2, perm3); \
-res1 = vec_mladd(but2, vprod3, op3);				\
+res1 = vec_mladd(but2, vprod3, op3);                              \
 register vector signed short op3S asm ("v30") = vec_perm(but2S, but2S, perm3); \
-res2 = vec_mladd(but2S, vprod3, op3S);				\
+res2 = vec_mladd(but2S, vprod3, op3S);                            \
 }
 ONEITERBUTTERFLY(0, temp0, temp0S);
 ONEITERBUTTERFLY(1, temp1, temp1S);
 ONEITERBUTTERFLY(2, temp2, temp2S);
 ONEITERBUTTERFLY(3, temp3, temp3S);
 #endif //CONFIG_DARWIN
 int has_altivec(void)
 {
 #ifdef __AMIGAOS4__
-	ULONG result = 0;
+ULONG result = 0;
-	extern struct ExecIFace *IExec;
+extern struct ExecIFace *IExec;
-	IExec->GetCPUInfoTags(GCIT_VectorUnit, &result, TAG_DONE);
+IExec->GetCPUInfoTags(GCIT_VectorUnit, &result, TAG_DONE);
-	if (result == VECTORTYPE_ALTIVEC) return 1;
+if (result == VECTORTYPE_ALTIVEC) return 1;
-	return 0;
+return 0;
 #else /* __AMIGAOS4__ */
 #ifdef CONFIG_DARWIN
 int sels[2] = {CTL_HW, HW_VECTORUNIT};
 int has_vu = 0;

Mercurial > libavcodec.hg

comparison ppc/dsputil_altivec.c @ 2979:bfabfdf9ce55 libavcodec