Originally committed as revision 22593 to svn://svn.ffmpeg.org/ffmpeg/trunktags/v0.6
@@ -837,7 +837,6 @@ INCLUDE_FILE_PATTERNS = | |||
PREDEFINED = __attribute__(x)="" "RENAME(x)=x ## _TMPL" "DEF(x)=x ## _TMPL" \ | |||
HAVE_AV_CONFIG_H HAVE_MMX HAVE_MMX2 HAVE_AMD3DNOW \ | |||
ATTR_ALIGN(x)="" | |||
# If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then | |||
# this tag can be used to specify a list of macro names that should be expanded. | |||
@@ -51,20 +51,18 @@ | |||
* fpsub16 f12, f10, f10 | |||
*/ | |||
#define ATTR_ALIGN(alignd) __attribute__ ((aligned(alignd))) | |||
#define DUP4(x) {x, x, x, x} | |||
#define DUP8(x) {x, x, x, x, x, x, x, x} | |||
static const int16_t constants1[] ATTR_ALIGN(8) = DUP4 (1); | |||
static const int16_t constants2[] ATTR_ALIGN(8) = DUP4 (2); | |||
static const int16_t constants3[] ATTR_ALIGN(8) = DUP4 (3); | |||
static const int16_t constants6[] ATTR_ALIGN(8) = DUP4 (6); | |||
static const int8_t constants_fe[] ATTR_ALIGN(8) = DUP8 (0xfe); | |||
static const int8_t constants_7f[] ATTR_ALIGN(8) = DUP8 (0x7f); | |||
static const int8_t constants128[] ATTR_ALIGN(8) = DUP8 (128); | |||
static const int16_t constants256_512[] ATTR_ALIGN(8) = | |||
DECLARE_ALIGNED(8, static const int16_t, constants1)[] = DUP4 (1); | |||
DECLARE_ALIGNED(8, static const int16_t, constants2)[] = DUP4 (2); | |||
DECLARE_ALIGNED(8, static const int16_t, constants3)[] = DUP4 (3); | |||
DECLARE_ALIGNED(8, static const int16_t, constants6)[] = DUP4 (6); | |||
DECLARE_ALIGNED(8, static const int8_t, constants_fe)[] = DUP8 (0xfe); | |||
DECLARE_ALIGNED(8, static const int8_t, constants_7f)[] = DUP8 (0x7f); | |||
DECLARE_ALIGNED(8, static const int8_t, constants128)[] = DUP8 (128); | |||
DECLARE_ALIGNED(8, static const int16_t, constants256_512)[] = | |||
{256, 512, 256, 512}; | |||
static const int16_t constants256_1024[] ATTR_ALIGN(8) = | |||
DECLARE_ALIGNED(8, static const int16_t, constants256_1024)[] = | |||
{256, 1024, 256, 1024}; | |||
#define REF_0 0 | |||
@@ -33,8 +33,6 @@ | |||
#include "libavutil/common.h" | |||
#include "libavcodec/dsputil.h" | |||
#define ATTR_ALIGN(align) __attribute__ ((__aligned__ (align))) | |||
////////////////////////////////////////////////////////////////////// | |||
// | |||
// constants for the forward DCT | |||
@@ -55,30 +53,30 @@ | |||
#define X8(x) x,x,x,x,x,x,x,x | |||
//concatenated table, for forward DCT transformation | |||
static const int16_t fdct_tg_all_16[24] ATTR_ALIGN(16) = { | |||
DECLARE_ALIGNED(16, static const int16_t, fdct_tg_all_16)[24] = { | |||
X8(13036), // tg * (2<<16) + 0.5 | |||
X8(27146), // tg * (2<<16) + 0.5 | |||
X8(-21746) // tg * (2<<16) + 0.5 | |||
}; | |||
static const int16_t ocos_4_16[8] ATTR_ALIGN(16) = { | |||
DECLARE_ALIGNED(16, static const int16_t, ocos_4_16)[8] = { | |||
X8(23170) //cos * (2<<15) + 0.5 | |||
}; | |||
static const int16_t fdct_one_corr[8] ATTR_ALIGN(16) = { X8(1) }; | |||
DECLARE_ALIGNED(16, static const int16_t, fdct_one_corr)[8] = { X8(1) }; | |||
static const int32_t fdct_r_row[2] ATTR_ALIGN(8) = {RND_FRW_ROW, RND_FRW_ROW }; | |||
DECLARE_ALIGNED(8, static const int32_t, fdct_r_row)[2] = {RND_FRW_ROW, RND_FRW_ROW }; | |||
static struct | |||
{ | |||
const int32_t fdct_r_row_sse2[4] ATTR_ALIGN(16); | |||
} fdct_r_row_sse2 ATTR_ALIGN(16)= | |||
DECLARE_ALIGNED(16, const int32_t, fdct_r_row_sse2)[4]; | |||
} fdct_r_row_sse2 = | |||
{{ | |||
RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW | |||
}}; | |||
//static const long fdct_r_row_sse2[4] ATTR_ALIGN(16) = {RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW}; | |||
//DECLARE_ALIGNED(16, static const long, fdct_r_row_sse2)[4] = {RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW, RND_FRW_ROW}; | |||
static const int16_t tab_frw_01234567[] ATTR_ALIGN(8) = { // forward_dct coeff table | |||
DECLARE_ALIGNED(8, static const int16_t, tab_frw_01234567)[] = { // forward_dct coeff table | |||
16384, 16384, 22725, 19266, | |||
16384, 16384, 12873, 4520, | |||
21407, 8867, 19266, -4520, | |||
@@ -154,10 +152,10 @@ static const int16_t tab_frw_01234567[] ATTR_ALIGN(8) = { // forward_dct coeff | |||
static struct | |||
{ | |||
const int16_t tab_frw_01234567_sse2[256] ATTR_ALIGN(16); | |||
} tab_frw_01234567_sse2 ATTR_ALIGN(16) = | |||
DECLARE_ALIGNED(16, const int16_t, tab_frw_01234567_sse2)[256]; | |||
} tab_frw_01234567_sse2 = | |||
{{ | |||
//static const int16_t tab_frw_01234567_sse2[] ATTR_ALIGN(16) = { // forward_dct coeff table | |||
//DECLARE_ALIGNED(16, static const int16_t, tab_frw_01234567_sse2)[] = { // forward_dct coeff table | |||
#define TABLE_SSE2 C4, C4, C1, C3, -C6, -C2, -C1, -C5, \ | |||
C4, C4, C5, C7, C2, C6, C3, -C7, \ | |||
-C4, C4, C7, C3, C6, -C2, C7, -C5, \ | |||
@@ -535,7 +533,7 @@ static av_always_inline void fdct_row_mmx(const int16_t *in, int16_t *out, const | |||
void ff_fdct_mmx(int16_t *block) | |||
{ | |||
int64_t align_tmp[16] ATTR_ALIGN(8); | |||
DECLARE_ALIGNED(8, int64_t, align_tmp)[16]; | |||
int16_t * block1= (int16_t*)align_tmp; | |||
const int16_t *table= tab_frw_01234567; | |||
int i; | |||
@@ -553,7 +551,7 @@ void ff_fdct_mmx(int16_t *block) | |||
void ff_fdct_mmx2(int16_t *block) | |||
{ | |||
int64_t align_tmp[16] ATTR_ALIGN(8); | |||
DECLARE_ALIGNED(8, int64_t, align_tmp)[16]; | |||
int16_t *block1= (int16_t*)align_tmp; | |||
const int16_t *table= tab_frw_01234567; | |||
int i; | |||
@@ -571,7 +569,7 @@ void ff_fdct_mmx2(int16_t *block) | |||
void ff_fdct_sse2(int16_t *block) | |||
{ | |||
int64_t align_tmp[16] ATTR_ALIGN(16); | |||
DECLARE_ALIGNED(16, int64_t, align_tmp)[16]; | |||
int16_t * const block1= (int16_t*)align_tmp; | |||
fdct_col_sse2(block, block1, 0); | |||
@@ -26,8 +26,6 @@ | |||
#include "dsputil_mmx.h" | |||
#include "mmx.h" | |||
#define ATTR_ALIGN(align) __attribute__ ((__aligned__ (align))) | |||
#define ROW_SHIFT 11 | |||
#define COL_SHIFT 6 | |||
@@ -400,10 +398,10 @@ static inline void idct_col (int16_t * const col, const int offset) | |||
#define T3 43790 | |||
#define C4 23170 | |||
static const short t1_vector[] ATTR_ALIGN(8) = {T1,T1,T1,T1}; | |||
static const short t2_vector[] ATTR_ALIGN(8) = {T2,T2,T2,T2}; | |||
static const short t3_vector[] ATTR_ALIGN(8) = {T3,T3,T3,T3}; | |||
static const short c4_vector[] ATTR_ALIGN(8) = {C4,C4,C4,C4}; | |||
DECLARE_ALIGNED(8, static const short, t1_vector)[] = {T1,T1,T1,T1}; | |||
DECLARE_ALIGNED(8, static const short, t2_vector)[] = {T2,T2,T2,T2}; | |||
DECLARE_ALIGNED(8, static const short, t3_vector)[] = {T3,T3,T3,T3}; | |||
DECLARE_ALIGNED(8, static const short, c4_vector)[] = {C4,C4,C4,C4}; | |||
/* column code adapted from Peter Gubanov */ | |||
/* http://www.elecard.com/peter/idct.shtml */ | |||
@@ -542,20 +540,20 @@ static inline void idct_col (int16_t * const col, const int offset) | |||
} | |||
static const int32_t rounder0[] ATTR_ALIGN(8) = | |||
DECLARE_ALIGNED(8, static const int32_t, rounder0)[] = | |||
rounder ((1 << (COL_SHIFT - 1)) - 0.5); | |||
static const int32_t rounder4[] ATTR_ALIGN(8) = rounder (0); | |||
static const int32_t rounder1[] ATTR_ALIGN(8) = | |||
DECLARE_ALIGNED(8, static const int32_t, rounder4)[] = rounder (0); | |||
DECLARE_ALIGNED(8, static const int32_t, rounder1)[] = | |||
rounder (1.25683487303); /* C1*(C1/C4+C1+C7)/2 */ | |||
static const int32_t rounder7[] ATTR_ALIGN(8) = | |||
DECLARE_ALIGNED(8, static const int32_t, rounder7)[] = | |||
rounder (-0.25); /* C1*(C7/C4+C7-C1)/2 */ | |||
static const int32_t rounder2[] ATTR_ALIGN(8) = | |||
DECLARE_ALIGNED(8, static const int32_t, rounder2)[] = | |||
rounder (0.60355339059); /* C2 * (C6+C2)/2 */ | |||
static const int32_t rounder6[] ATTR_ALIGN(8) = | |||
DECLARE_ALIGNED(8, static const int32_t, rounder6)[] = | |||
rounder (-0.25); /* C2 * (C6-C2)/2 */ | |||
static const int32_t rounder3[] ATTR_ALIGN(8) = | |||
DECLARE_ALIGNED(8, static const int32_t, rounder3)[] = | |||
rounder (0.087788325588); /* C3*(-C3/C4+C3+C5)/2 */ | |||
static const int32_t rounder5[] ATTR_ALIGN(8) = | |||
DECLARE_ALIGNED(8, static const int32_t, rounder5)[] = | |||
rounder (-0.441341716183); /* C3*(-C5/C4+C5-C3)/2 */ | |||
#undef COL_SHIFT | |||
@@ -564,13 +562,13 @@ static const int32_t rounder5[] ATTR_ALIGN(8) = | |||
#define declare_idct(idct,table,idct_row_head,idct_row,idct_row_tail,idct_row_mid) \ | |||
void idct (int16_t * const block) \ | |||
{ \ | |||
static const int16_t table04[] ATTR_ALIGN(16) = \ | |||
DECLARE_ALIGNED(16, static const int16_t, table04)[] = \ | |||
table (22725, 21407, 19266, 16384, 12873, 8867, 4520); \ | |||
static const int16_t table17[] ATTR_ALIGN(16) = \ | |||
DECLARE_ALIGNED(16, static const int16_t, table17)[] = \ | |||
table (31521, 29692, 26722, 22725, 17855, 12299, 6270); \ | |||
static const int16_t table26[] ATTR_ALIGN(16) = \ | |||
DECLARE_ALIGNED(16, static const int16_t, table26)[] = \ | |||
table (29692, 27969, 25172, 21407, 16819, 11585, 5906); \ | |||
static const int16_t table35[] ATTR_ALIGN(16) = \ | |||
DECLARE_ALIGNED(16, static const int16_t, table35)[] = \ | |||
table (26722, 25172, 22654, 19266, 15137, 10426, 5315); \ | |||
\ | |||
idct_row_head (block, 0*8, table04); \ | |||