* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
-#include "avutil.h"
+#include "libavutil/avutil.h"
#define ALTIVEC_TRANSPOSE_8x8_SHORT(src_a,src_b,src_c,src_d,src_e,src_f,src_g,src_h) \
do { \
vector by assuming (stride % 16) == 0, unfortunately
this is not always true.
*/
- DECLARE_ALIGNED(16, short, data[8]) =
+ short data_0 = ((c->nonBQP*c->ppMode.baseDcDiff)>>8) + 1;
+ DECLARE_ALIGNED(16, short, data)[8] =
{
- ((c->nonBQP*c->ppMode.baseDcDiff)>>8) + 1,
- data[0] * 2 + 1,
+ data_0,
+ data_0 * 2 + 1,
c->QP * 2,
c->QP * 4
};
if (numEq > c->ppMode.flatnessThreshold){
const vector unsigned char mmoP1 = (const vector unsigned char)
- AVV(0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f,
- 0x00, 0x01, 0x12, 0x13, 0x08, 0x09, 0x1A, 0x1B);
+ {0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f,
+ 0x00, 0x01, 0x12, 0x13, 0x08, 0x09, 0x1A, 0x1B};
const vector unsigned char mmoP2 = (const vector unsigned char)
- AVV(0x04, 0x05, 0x16, 0x17, 0x0C, 0x0D, 0x1E, 0x1F,
- 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f);
+ {0x04, 0x05, 0x16, 0x17, 0x0C, 0x0D, 0x1E, 0x1F,
+ 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f};
const vector unsigned char mmoP = (const vector unsigned char)
vec_lvsl(8, (unsigned char*)0);
const vector signed int zero = vec_splat_s32(0);
const int properStride = (stride % 16);
const int srcAlign = ((unsigned long)src2 % 16);
- DECLARE_ALIGNED(16, short, qp[8]) = {c->QP};
+ DECLARE_ALIGNED(16, short, qp)[8] = {c->QP};
vector signed short vqp = vec_ld(0, qp);
vector signed short vb0, vb1, vb2, vb3, vb4, vb5, vb6, vb7, vb8, vb9;
- vector unsigned char vbA0, vbA1, vbA2, vbA3, vbA4, vbA5, vbA6, vbA7, vbA8, vbA9;
- vector unsigned char vbB0, vbB1, vbB2, vbB3, vbB4, vbB5, vbB6, vbB7, vbB8, vbB9;
+ vector unsigned char vbA0, av_uninit(vbA1), av_uninit(vbA2), av_uninit(vbA3), av_uninit(vbA4), av_uninit(vbA5), av_uninit(vbA6), av_uninit(vbA7), av_uninit(vbA8), vbA9;
+ vector unsigned char vbB0, av_uninit(vbB1), av_uninit(vbB2), av_uninit(vbB3), av_uninit(vbB4), av_uninit(vbB5), av_uninit(vbB6), av_uninit(vbB7), av_uninit(vbB8), vbB9;
vector unsigned char vbT0, vbT1, vbT2, vbT3, vbT4, vbT5, vbT6, vbT7, vbT8, vbT9;
vector unsigned char perml0, perml1, perml2, perml3, perml4,
perml5, perml6, perml7, perml8, perml9;
COMPUTE_VR(7, 8, 9);
const vector signed char neg1 = vec_splat_s8(-1);
- const vector unsigned char permHH = (const vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
- 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F);
+ const vector unsigned char permHH = (const vector unsigned char){0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+ 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F};
#define PACK_AND_STORE(i) \
-{ const vector unsigned char perms##i = \
+{ const vector unsigned char perms##i = \
vec_lvsr(i * stride, src2); \
const vector unsigned char vf##i = \
vec_packsu(vr##i, (vector signed short)zero); \
vec_st(svB##i, i * stride + 16, src2);}
#define PACK_AND_STORE_ALIGNED(i) \
-{ const vector unsigned char vf##i = \
+{ const vector unsigned char vf##i = \
vec_packsu(vr##i, (vector signed short)zero); \
const vector unsigned char vg##i = \
vec_perm(vf##i, vbT##i, permHH); \
*/
uint8_t *src2 = src + stride*3;
const vector signed int zero = vec_splat_s32(0);
- DECLARE_ALIGNED(16, short, qp[8]) = {8*c->QP};
+ DECLARE_ALIGNED(16, short, qp)[8] = {8*c->QP};
vector signed short vqp = vec_splat(
(vector signed short)vec_ld(0, qp), 0);
const vector unsigned char st5 = vec_packsu(vb5plusd, (vector signed short)zero);
const vector signed char neg1 = vec_splat_s8(-1);
- const vector unsigned char permHH = (const vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
- 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F);
+ const vector unsigned char permHH = (const vector unsigned char){0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+ 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F};
#define STORE(i) \
-{ const vector unsigned char perms##i = \
+{ const vector unsigned char perms##i = \
vec_lvsr(i * stride, src2); \
const vector unsigned char vg##i = \
vec_perm(st##i, vbT##i, permHH); \
src & stride :-(
*/
uint8_t *srcCopy = src;
- DECLARE_ALIGNED(16, uint8_t, dt[16]);
+ DECLARE_ALIGNED(16, uint8_t, dt)[16];
const vector signed int zero = vec_splat_s32(0);
vector unsigned char v_dt;
dt[0] = deringThreshold;
v_dt = vec_splat(vec_ld(0, dt), 0);
-#define LOAD_LINE(i) \
- const vector unsigned char perm##i = \
- vec_lvsl(i * stride, srcCopy); \
- vector unsigned char sA##i = vec_ld(i * stride, srcCopy); \
- vector unsigned char sB##i = vec_ld(i * stride + 16, srcCopy); \
+#define LOAD_LINE(i) \
+ const vector unsigned char perm##i = \
+ vec_lvsl(i * stride, srcCopy); \
+ vector unsigned char sA##i = vec_ld(i * stride, srcCopy); \
+ vector unsigned char sB##i = vec_ld(i * stride + 16, srcCopy); \
vector unsigned char src##i = vec_perm(sA##i, sB##i, perm##i)
LOAD_LINE(0);
vector unsigned char v_avg;
{
const vector unsigned char trunc_perm = (vector unsigned char)
- AVV(0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08,
- 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18);
+ {0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08,
+ 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18};
const vector unsigned char trunc_src12 = vec_perm(src1, src2, trunc_perm);
const vector unsigned char trunc_src34 = vec_perm(src3, src4, trunc_perm);
const vector unsigned char trunc_src56 = vec_perm(src5, src6, trunc_perm);
v_avg = vec_avg(v_min, v_max);
}
- DECLARE_ALIGNED(16, signed int, S[8]);
+ DECLARE_ALIGNED(16, signed int, S)[8];
{
const vector unsigned short mask1 = (vector unsigned short)
- AVV(0x0001, 0x0002, 0x0004, 0x0008,
- 0x0010, 0x0020, 0x0040, 0x0080);
+ {0x0001, 0x0002, 0x0004, 0x0008,
+ 0x0010, 0x0020, 0x0040, 0x0080};
const vector unsigned short mask2 = (vector unsigned short)
- AVV(0x0100, 0x0200, 0x0000, 0x0000,
- 0x0000, 0x0000, 0x0000, 0x0000);
+ {0x0100, 0x0200, 0x0000, 0x0000,
+ 0x0000, 0x0000, 0x0000, 0x0000};
const vector unsigned int vuint32_16 = vec_sl(vec_splat_u32(1), vec_splat_u32(4));
const vector unsigned int vuint32_1 = vec_splat_u32(1);
/* I'm not sure the following is actually faster
than straight, unvectorized C code :-( */
- DECLARE_ALIGNED(16, int, tQP2[4]);
+ DECLARE_ALIGNED(16, int, tQP2)[4];
tQP2[0]= c->QP/2 + 1;
vector signed int vQP2 = vec_ld(0, tQP2);
vQP2 = vec_splat(vQP2, 0);
const vector unsigned int vuint32_4 = vec_splat_u32(4);
const vector unsigned char permA1 = (vector unsigned char)
- AVV(0x00, 0x01, 0x02, 0x10, 0x11, 0x12, 0x1F, 0x1F,
- 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F);
+ {0x00, 0x01, 0x02, 0x10, 0x11, 0x12, 0x1F, 0x1F,
+ 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F};
const vector unsigned char permA2 = (vector unsigned char)
- AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x10, 0x11,
- 0x12, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F);
+ {0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x10, 0x11,
+ 0x12, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F, 0x1F};
const vector unsigned char permA1inc = (vector unsigned char)
- AVV(0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00);
+ {0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00};
const vector unsigned char permA2inc = (vector unsigned char)
- AVV(0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x01,
- 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00);
+ {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x01,
+ 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00};
const vector unsigned char magic = (vector unsigned char)
- AVV(0x01, 0x02, 0x01, 0x02, 0x04, 0x02, 0x01, 0x02,
- 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00);
+ {0x01, 0x02, 0x01, 0x02, 0x04, 0x02, 0x01, 0x02,
+ 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00};
const vector unsigned char extractPerm = (vector unsigned char)
- AVV(0x10, 0x10, 0x10, 0x01, 0x10, 0x10, 0x10, 0x01,
- 0x10, 0x10, 0x10, 0x01, 0x10, 0x10, 0x10, 0x01);
+ {0x10, 0x10, 0x10, 0x01, 0x10, 0x10, 0x10, 0x01,
+ 0x10, 0x10, 0x10, 0x01, 0x10, 0x10, 0x10, 0x01};
const vector unsigned char extractPermInc = (vector unsigned char)
- AVV(0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01,
- 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01);
+ {0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01,
+ 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01};
const vector unsigned char identity = vec_lvsl(0,(unsigned char *)0);
const vector unsigned char tenRight = (vector unsigned char)
- AVV(0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00);
+ {0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00};
const vector unsigned char eightLeft = (vector unsigned char)
- AVV(0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x08);
+ {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x08};
#define F_INIT(i) \
#undef LOAD_LINE
#define ACCUMULATE_DIFFS(i) \
- vector signed short v_d##i = vec_sub(v_tempBlurredAss##i, \
+ vector signed short v_d##i = vec_sub(v_tempBlurredAss##i, \
v_srcAss##i); \
v_dp = vec_msums(v_d##i, v_d##i, v_dp); \
v_sysdp = vec_msums(v_d##i, vsint16_1, v_sysdp)
#define OP(i) \
const vector signed short v_temp##i = \
- vec_mladd(v_tempBlurredAss##i, \
+ vec_mladd(v_tempBlurredAss##i, \
vsint16_7, v_srcAss##i); \
const vector signed short v_temp2##i = \
vec_add(v_temp##i, vsint16_4); \
#define OP(i) \
const vector signed short v_temp##i = \
- vec_mladd(v_tempBlurredAss##i, \
+ vec_mladd(v_tempBlurredAss##i, \
vsint16_3, v_srcAss##i); \
const vector signed short v_temp2##i = \
vec_add(v_temp##i, vsint16_2); \
}
const vector signed char neg1 = vec_splat_s8(-1);
- const vector unsigned char permHH = (const vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
- 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F);
+ const vector unsigned char permHH = (const vector unsigned char){0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+ 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F};
#define PACK_AND_STORE(src, i) \
const vector unsigned char perms##src##i = \
vector unsigned char tempO = vec_mergeh(src7, zero);
vector unsigned char tempP = vec_mergel(src7, zero);
- vector unsigned char temp0 = vec_mergeh(tempA, tempI);
- vector unsigned char temp1 = vec_mergel(tempA, tempI);
- vector unsigned char temp2 = vec_mergeh(tempB, tempJ);
- vector unsigned char temp3 = vec_mergel(tempB, tempJ);
- vector unsigned char temp4 = vec_mergeh(tempC, tempK);
- vector unsigned char temp5 = vec_mergel(tempC, tempK);
- vector unsigned char temp6 = vec_mergeh(tempD, tempL);
- vector unsigned char temp7 = vec_mergel(tempD, tempL);
- vector unsigned char temp8 = vec_mergeh(tempE, tempM);
- vector unsigned char temp9 = vec_mergel(tempE, tempM);
+ vector unsigned char temp0 = vec_mergeh(tempA, tempI);
+ vector unsigned char temp1 = vec_mergel(tempA, tempI);
+ vector unsigned char temp2 = vec_mergeh(tempB, tempJ);
+ vector unsigned char temp3 = vec_mergel(tempB, tempJ);
+ vector unsigned char temp4 = vec_mergeh(tempC, tempK);
+ vector unsigned char temp5 = vec_mergel(tempC, tempK);
+ vector unsigned char temp6 = vec_mergeh(tempD, tempL);
+ vector unsigned char temp7 = vec_mergel(tempD, tempL);
+ vector unsigned char temp8 = vec_mergeh(tempE, tempM);
+ vector unsigned char temp9 = vec_mergel(tempE, tempM);
vector unsigned char temp10 = vec_mergeh(tempF, tempN);
vector unsigned char temp11 = vec_mergel(tempF, tempN);
vector unsigned char temp12 = vec_mergeh(tempG, tempO);
tempO = vec_mergeh(temp7, temp15);
tempP = vec_mergel(temp7, temp15);
- temp0 = vec_mergeh(tempA, tempI);
- temp1 = vec_mergel(tempA, tempI);
- temp2 = vec_mergeh(tempB, tempJ);
- temp3 = vec_mergel(tempB, tempJ);
- temp4 = vec_mergeh(tempC, tempK);
- temp5 = vec_mergel(tempC, tempK);
- temp6 = vec_mergeh(tempD, tempL);
- temp7 = vec_mergel(tempD, tempL);
- temp8 = vec_mergeh(tempE, tempM);
- temp9 = vec_mergel(tempE, tempM);
+ temp0 = vec_mergeh(tempA, tempI);
+ temp1 = vec_mergel(tempA, tempI);
+ temp2 = vec_mergeh(tempB, tempJ);
+ temp3 = vec_mergel(tempB, tempJ);
+ temp4 = vec_mergeh(tempC, tempK);
+ temp5 = vec_mergel(tempC, tempK);
+ temp6 = vec_mergeh(tempD, tempL);
+ temp7 = vec_mergel(tempD, tempL);
+ temp8 = vec_mergeh(tempE, tempM);
+ temp9 = vec_mergel(tempE, tempM);
temp10 = vec_mergeh(tempF, tempN);
temp11 = vec_mergel(tempF, tempN);
temp12 = vec_mergeh(tempG, tempO);
temp14 = vec_mergeh(tempH, tempP);
temp15 = vec_mergel(tempH, tempP);
- vec_st(temp0, 0, dst);
- vec_st(temp1, 16, dst);
- vec_st(temp2, 32, dst);
- vec_st(temp3, 48, dst);
- vec_st(temp4, 64, dst);
- vec_st(temp5, 80, dst);
- vec_st(temp6, 96, dst);
- vec_st(temp7, 112, dst);
- vec_st(temp8, 128, dst);
- vec_st(temp9, 144, dst);
+ vec_st(temp0, 0, dst);
+ vec_st(temp1, 16, dst);
+ vec_st(temp2, 32, dst);
+ vec_st(temp3, 48, dst);
+ vec_st(temp4, 64, dst);
+ vec_st(temp5, 80, dst);
+ vec_st(temp6, 96, dst);
+ vec_st(temp7, 112, dst);
+ vec_st(temp8, 128, dst);
+ vec_st(temp9, 144, dst);
vec_st(temp10, 160, dst);
vec_st(temp11, 176, dst);
vec_st(temp12, 192, dst);