mirrors/gstreamer
mirrors/gstreamer
1
1
Fork 0
mirror of https://gitlab.freedesktop.org/gstreamer/gstreamer.git synced 2024-11-14 13:21:28 +00:00
Code Issues 1 Projects Releases Wiki Activity
gstreamer/gst/bayer/gstbayerorc-dist.c

3253 lines
92 KiB
C
Raw Normal View History

bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
/* autogenerated from gstbayerorc.orc */
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <glib.h>
#ifndef _ORC_INTEGER_TYPEDEFS_
#define _ORC_INTEGER_TYPEDEFS_
#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
#include <stdint.h>
typedef int8_t orc_int8;
typedef int16_t orc_int16;
typedef int32_t orc_int32;
typedef int64_t orc_int64;
typedef uint8_t orc_uint8;
typedef uint16_t orc_uint16;
typedef uint32_t orc_uint32;
typedef uint64_t orc_uint64;
#define ORC_UINT64_C(x) UINT64_C(x)
#elif defined(_MSC_VER)
typedef signed __int8 orc_int8;
typedef signed __int16 orc_int16;
typedef signed __int32 orc_int32;
typedef signed __int64 orc_int64;
typedef unsigned __int8 orc_uint8;
typedef unsigned __int16 orc_uint16;
typedef unsigned __int32 orc_uint32;
typedef unsigned __int64 orc_uint64;
#define ORC_UINT64_C(x) (x##Ui64)
#define inline __inline
#else
#include <limits.h>
typedef signed char orc_int8;
typedef short orc_int16;
typedef int orc_int32;
typedef unsigned char orc_uint8;
typedef unsigned short orc_uint16;
typedef unsigned int orc_uint32;
#if INT_MAX == LONG_MAX
typedef long long orc_int64;
typedef unsigned long long orc_uint64;
#define ORC_UINT64_C(x) (x##ULL)
#else
typedef long orc_int64;
typedef unsigned long orc_uint64;
#define ORC_UINT64_C(x) (x##UL)
#endif
#endif
typedef union
{
orc_int16 i;
orc_int8 x2[2];
} orc_union16;
typedef union
{
orc_int32 i;
float f;
orc_int16 x2[2];
orc_int8 x4[4];
} orc_union32;
typedef union
{
orc_int64 i;
double f;
orc_int32 x2[2];
float x2f[2];
orc_int16 x4[4];
} orc_union64;
#endif
#ifndef ORC_RESTRICT
#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
#define ORC_RESTRICT restrict
#elif defined(__GNUC__) && __GNUC__ >= 4
#define ORC_RESTRICT __restrict__
#else
#define ORC_RESTRICT
#endif
#endif
#ifndef DISABLE_ORC
#include <orc/orc.h>
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_horiz_upsample_unaligned (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
guint8 * ORC_RESTRICT d2, const guint8 * ORC_RESTRICT s1, int n);
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_horiz_upsample (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
guint8 * ORC_RESTRICT d2, const guint8 * ORC_RESTRICT s1, int n);
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_merge_bg_bgra (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n);
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_merge_gr_bgra (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n);
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_merge_bg_abgr (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n);
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_merge_gr_abgr (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n);
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_merge_bg_rgba (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n);
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_merge_gr_rgba (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n);
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_merge_bg_argb (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n);
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
void bayer_orc_merge_gr_argb (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n);
/* begin Orc C target preamble */
#define ORC_CLAMP(x,a,b) ((x)<(a) ? (a) : ((x)>(b) ? (b) : (x)))
#define ORC_ABS(a) ((a)<0 ? -(a) : (a))
#define ORC_MIN(a,b) ((a)<(b) ? (a) : (b))
#define ORC_MAX(a,b) ((a)>(b) ? (a) : (b))
#define ORC_SB_MAX 127
#define ORC_SB_MIN (-1-ORC_SB_MAX)
#define ORC_UB_MAX 255
#define ORC_UB_MIN 0
#define ORC_SW_MAX 32767
#define ORC_SW_MIN (-1-ORC_SW_MAX)
#define ORC_UW_MAX 65535
#define ORC_UW_MIN 0
#define ORC_SL_MAX 2147483647
#define ORC_SL_MIN (-1-ORC_SL_MAX)
#define ORC_UL_MAX 4294967295U
#define ORC_UL_MIN 0
#define ORC_CLAMP_SB(x) ORC_CLAMP(x,ORC_SB_MIN,ORC_SB_MAX)
#define ORC_CLAMP_UB(x) ORC_CLAMP(x,ORC_UB_MIN,ORC_UB_MAX)
#define ORC_CLAMP_SW(x) ORC_CLAMP(x,ORC_SW_MIN,ORC_SW_MAX)
#define ORC_CLAMP_UW(x) ORC_CLAMP(x,ORC_UW_MIN,ORC_UW_MAX)
#define ORC_CLAMP_SL(x) ORC_CLAMP(x,ORC_SL_MIN,ORC_SL_MAX)
#define ORC_CLAMP_UL(x) ORC_CLAMP(x,ORC_UL_MIN,ORC_UL_MAX)
#define ORC_SWAP_W(x) ((((x)&0xff)<<8) | (((x)&0xff00)>>8))
#define ORC_SWAP_L(x) ((((x)&0xff)<<24) | (((x)&0xff00)<<8) | (((x)&0xff0000)>>8) | (((x)&0xff000000)>>24))
#define ORC_SWAP_Q(x) ((((x)&ORC_UINT64_C(0xff))<<56) | (((x)&ORC_UINT64_C(0xff00))<<40) | (((x)&ORC_UINT64_C(0xff0000))<<24) | (((x)&ORC_UINT64_C(0xff000000))<<8) | (((x)&ORC_UINT64_C(0xff00000000))>>8) | (((x)&ORC_UINT64_C(0xff0000000000))>>24) | (((x)&ORC_UINT64_C(0xff000000000000))>>40) | (((x)&ORC_UINT64_C(0xff00000000000000))>>56))
#define ORC_PTR_OFFSET(ptr,offset) ((void *)(((unsigned char *)(ptr)) + (offset)))
#define ORC_DENORMAL(x) ((x) & ((((x)&0x7f800000) == 0) ? 0xff800000 : 0xffffffff))
#define ORC_ISNAN(x) ((((x)&0x7f800000) == 0x7f800000) && (((x)&0x007fffff) != 0))
#define ORC_DENORMAL_DOUBLE(x) ((x) & ((((x)&ORC_UINT64_C(0x7ff0000000000000)) == 0) ? ORC_UINT64_C(0xfff0000000000000) : ORC_UINT64_C(0xffffffffffffffff)))
#define ORC_ISNAN_DOUBLE(x) ((((x)&ORC_UINT64_C(0x7ff0000000000000)) == ORC_UINT64_C(0x7ff0000000000000)) && (((x)&ORC_UINT64_C(0x000fffffffffffff)) != 0))
#ifndef ORC_RESTRICT
#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
#define ORC_RESTRICT restrict
#elif defined(__GNUC__) && __GNUC__ >= 4
#define ORC_RESTRICT __restrict__
#else
#define ORC_RESTRICT
#endif
#endif
/* end Orc C target preamble */
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_horiz_upsample_unaligned */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_horiz_upsample_unaligned (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
guint8 * ORC_RESTRICT d2, const guint8 * ORC_RESTRICT s1, int n)
{
int i;
orc_union16 *ORC_RESTRICT ptr0;
orc_union16 *ORC_RESTRICT ptr1;
const orc_union16 *ORC_RESTRICT ptr4;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_int8 var40;
orc_int8 var41;
orc_union16 var42;
orc_int8 var43;
orc_int8 var44;
orc_int8 var45;
orc_int8 var46;
ptr0 = (orc_union16 *) d1;
ptr1 = (orc_union16 *) d2;
ptr4 = (orc_union16 *) s1;
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: splitwb */
{
orc_union16 _src;
_src.i = var37.i;
var40 = _src.x2[1];
var41 = _src.x2[0];
}
/* 2: loadoffw */
var42 = ptr4[i + 1];
/* 3: splitwb */
{
orc_union16 _src;
_src.i = var42.i;
var43 = _src.x2[1];
var44 = _src.x2[0];
}
/* 4: avgub */
var45 = ((orc_uint8) var40 + (orc_uint8) var43 + 1) >> 1;
/* 5: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var40;
_dest.x2[1] = var45;
var38.i = _dest.i;
}
/* 6: storew */
ptr0[i] = var38;
/* 7: avgub */
var46 = ((orc_uint8) var41 + (orc_uint8) var44 + 1) >> 1;
/* 8: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var46;
_dest.x2[1] = var44;
var39.i = _dest.i;
}
/* 9: storew */
ptr1[i] = var39;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_horiz_upsample_unaligned (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union16 *ORC_RESTRICT ptr0;
orc_union16 *ORC_RESTRICT ptr1;
const orc_union16 *ORC_RESTRICT ptr4;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_int8 var40;
orc_int8 var41;
orc_union16 var42;
orc_int8 var43;
orc_int8 var44;
orc_int8 var45;
orc_int8 var46;
ptr0 = (orc_union16 *) ex->arrays[0];
ptr1 = (orc_union16 *) ex->arrays[1];
ptr4 = (orc_union16 *) ex->arrays[4];
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: splitwb */
{
orc_union16 _src;
_src.i = var37.i;
var40 = _src.x2[1];
var41 = _src.x2[0];
}
/* 2: loadoffw */
var42 = ptr4[i + 1];
/* 3: splitwb */
{
orc_union16 _src;
_src.i = var42.i;
var43 = _src.x2[1];
var44 = _src.x2[0];
}
/* 4: avgub */
var45 = ((orc_uint8) var40 + (orc_uint8) var43 + 1) >> 1;
/* 5: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var40;
_dest.x2[1] = var45;
var38.i = _dest.i;
}
/* 6: storew */
ptr0[i] = var38;
/* 7: avgub */
var46 = ((orc_uint8) var41 + (orc_uint8) var44 + 1) >> 1;
/* 8: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var46;
_dest.x2[1] = var44;
var39.i = _dest.i;
}
/* 9: storew */
ptr1[i] = var39;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_horiz_upsample_unaligned (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
guint8 * ORC_RESTRICT d2, const guint8 * ORC_RESTRICT s1, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_horiz_upsample_unaligned");
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_set_backup_function (p,
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_horiz_upsample_unaligned);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 2, "d1");
orc_program_add_destination (p, 2, "d2");
orc_program_add_source (p, 2, "s1");
orc_program_add_constant (p, 4, 0x00000001, "c1");
orc_program_add_temporary (p, 2, "t1");
orc_program_add_temporary (p, 1, "t2");
orc_program_add_temporary (p, 1, "t3");
orc_program_add_temporary (p, 1, "t4");
orc_program_add_temporary (p, 1, "t5");
orc_program_append_2 (p, "splitwb", 0, ORC_VAR_T3, ORC_VAR_T2, ORC_VAR_S1,
ORC_VAR_D1);
orc_program_append_2 (p, "loadoffw", 0, ORC_VAR_T1, ORC_VAR_S1,
ORC_VAR_C1, ORC_VAR_D1);
orc_program_append_2 (p, "splitwb", 0, ORC_VAR_T5, ORC_VAR_T4, ORC_VAR_T1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 0, ORC_VAR_T5, ORC_VAR_T3, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 0, ORC_VAR_D1, ORC_VAR_T3, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 0, ORC_VAR_T2, ORC_VAR_T2, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 0, ORC_VAR_D2, ORC_VAR_T2, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_D2] = d2;
ex->arrays[ORC_VAR_S1] = (void *) s1;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_horiz_upsample */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_horiz_upsample (guint8 * ORC_RESTRICT d1, guint8 * ORC_RESTRICT d2,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, int n)
{
int i;
orc_union16 *ORC_RESTRICT ptr0;
orc_union16 *ORC_RESTRICT ptr1;
const orc_union16 *ORC_RESTRICT ptr4;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_int8 var43;
orc_int8 var44;
orc_int8 var45;
orc_int8 var46;
orc_union16 var47;
orc_int8 var48;
orc_int8 var49;
orc_int8 var50;
orc_int8 var51;
ptr0 = (orc_union16 *) d1;
ptr1 = (orc_union16 *) d2;
ptr4 = (orc_union16 *) s1;
for (i = 0; i < n; i++) {
/* 0: loadoffw */
var42 = ptr4[i + -1];
/* 1: splitwb */
{
orc_union16 _src;
_src.i = var42.i;
var43 = _src.x2[1];
var44 = _src.x2[0];
}
/* 2: loadw */
var39 = ptr4[i];
/* 3: splitwb */
{
orc_union16 _src;
_src.i = var39.i;
var45 = _src.x2[1];
var46 = _src.x2[0];
}
/* 4: loadoffw */
var47 = ptr4[i + 1];
/* 5: splitwb */
{
orc_union16 _src;
_src.i = var47.i;
var48 = _src.x2[1];
var49 = _src.x2[0];
}
/* 6: avgub */
var50 = ((orc_uint8) var46 + (orc_uint8) var49 + 1) >> 1;
/* 7: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var46;
_dest.x2[1] = var50;
var40.i = _dest.i;
}
/* 8: storew */
ptr0[i] = var40;
/* 9: avgub */
var51 = ((orc_uint8) var43 + (orc_uint8) var45 + 1) >> 1;
/* 10: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var51;
_dest.x2[1] = var45;
var41.i = _dest.i;
}
/* 11: storew */
ptr1[i] = var41;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_horiz_upsample (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union16 *ORC_RESTRICT ptr0;
orc_union16 *ORC_RESTRICT ptr1;
const orc_union16 *ORC_RESTRICT ptr4;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_int8 var43;
orc_int8 var44;
orc_int8 var45;
orc_int8 var46;
orc_union16 var47;
orc_int8 var48;
orc_int8 var49;
orc_int8 var50;
orc_int8 var51;
ptr0 = (orc_union16 *) ex->arrays[0];
ptr1 = (orc_union16 *) ex->arrays[1];
ptr4 = (orc_union16 *) ex->arrays[4];
for (i = 0; i < n; i++) {
/* 0: loadoffw */
var42 = ptr4[i + -1];
/* 1: splitwb */
{
orc_union16 _src;
_src.i = var42.i;
var43 = _src.x2[1];
var44 = _src.x2[0];
}
/* 2: loadw */
var39 = ptr4[i];
/* 3: splitwb */
{
orc_union16 _src;
_src.i = var39.i;
var45 = _src.x2[1];
var46 = _src.x2[0];
}
/* 4: loadoffw */
var47 = ptr4[i + 1];
/* 5: splitwb */
{
orc_union16 _src;
_src.i = var47.i;
var48 = _src.x2[1];
var49 = _src.x2[0];
}
/* 6: avgub */
var50 = ((orc_uint8) var46 + (orc_uint8) var49 + 1) >> 1;
/* 7: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var46;
_dest.x2[1] = var50;
var40.i = _dest.i;
}
/* 8: storew */
ptr0[i] = var40;
/* 9: avgub */
var51 = ((orc_uint8) var43 + (orc_uint8) var45 + 1) >> 1;
/* 10: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var51;
_dest.x2[1] = var45;
var41.i = _dest.i;
}
/* 11: storew */
ptr1[i] = var41;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_horiz_upsample (guint8 * ORC_RESTRICT d1, guint8 * ORC_RESTRICT d2,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_horiz_upsample");
orc_program_set_backup_function (p, _backup_bayer_orc_horiz_upsample);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 2, "d1");
orc_program_add_destination (p, 2, "d2");
orc_program_add_source (p, 2, "s1");
orc_program_add_constant (p, 4, 0xffffffff, "c1");
orc_program_add_constant (p, 4, 0x00000001, "c2");
orc_program_add_temporary (p, 2, "t1");
orc_program_add_temporary (p, 1, "t2");
orc_program_add_temporary (p, 1, "t3");
orc_program_add_temporary (p, 1, "t4");
orc_program_add_temporary (p, 1, "t5");
orc_program_add_temporary (p, 1, "t6");
orc_program_add_temporary (p, 1, "t7");
orc_program_append_2 (p, "loadoffw", 0, ORC_VAR_T1, ORC_VAR_S1,
ORC_VAR_C1, ORC_VAR_D1);
orc_program_append_2 (p, "splitwb", 0, ORC_VAR_T3, ORC_VAR_T2, ORC_VAR_T1,
ORC_VAR_D1);
orc_program_append_2 (p, "splitwb", 0, ORC_VAR_T5, ORC_VAR_T4, ORC_VAR_S1,
ORC_VAR_D1);
orc_program_append_2 (p, "loadoffw", 0, ORC_VAR_T1, ORC_VAR_S1,
ORC_VAR_C2, ORC_VAR_D1);
orc_program_append_2 (p, "splitwb", 0, ORC_VAR_T7, ORC_VAR_T6, ORC_VAR_T1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 0, ORC_VAR_T6, ORC_VAR_T4, ORC_VAR_T6,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 0, ORC_VAR_D1, ORC_VAR_T4, ORC_VAR_T6,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 0, ORC_VAR_T3, ORC_VAR_T3, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 0, ORC_VAR_D2, ORC_VAR_T3, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_D2] = d2;
ex->arrays[ORC_VAR_S1] = (void *) s1;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_merge_bg_bgra */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_bg_bgra (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
int i;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) d1;
ptr4 = (orc_union16 *) s1;
ptr5 = (orc_union16 *) s2;
ptr6 = (orc_union16 *) s3;
ptr7 = (orc_union16 *) s4;
ptr8 = (orc_union16 *) s5;
ptr9 = (orc_union16 *) s6;
/* 9: loadpw */
var42.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 11: loadpw */
var43.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 16: loadpb */
var45.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var45.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr5[i];
/* 1: loadw */
var38 = ptr9[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr4[i];
/* 4: loadw */
var40 = ptr8[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr7[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 14: loadw */
var44 = ptr6[i];
/* 15: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var53.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var53.x2[1];
var54.x2[1] = _dest.i;
}
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_merge_bg_bgra (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) ex->arrays[0];
ptr4 = (orc_union16 *) ex->arrays[4];
ptr5 = (orc_union16 *) ex->arrays[5];
ptr6 = (orc_union16 *) ex->arrays[6];
ptr7 = (orc_union16 *) ex->arrays[7];
ptr8 = (orc_union16 *) ex->arrays[8];
ptr9 = (orc_union16 *) ex->arrays[9];
/* 9: loadpw */
var42.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 11: loadpw */
var43.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 16: loadpb */
var45.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var45.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr5[i];
/* 1: loadw */
var38 = ptr9[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr4[i];
/* 4: loadw */
var40 = ptr8[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr7[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 14: loadw */
var44 = ptr6[i];
/* 15: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var53.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var53.x2[1];
var54.x2[1] = _dest.i;
}
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_bg_bgra (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_merge_bg_bgra");
orc_program_set_backup_function (p, _backup_bayer_orc_merge_bg_bgra);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 8, "d1");
orc_program_add_source (p, 2, "s1");
orc_program_add_source (p, 2, "s2");
orc_program_add_source (p, 2, "s3");
orc_program_add_source (p, 2, "s4");
orc_program_add_source (p, 2, "s5");
orc_program_add_source (p, 2, "s6");
orc_program_add_constant (p, 4, 0x000000ff, "c1");
orc_program_add_constant (p, 4, 0x0000ff00, "c2");
orc_program_add_temporary (p, 4, "t1");
orc_program_add_temporary (p, 4, "t2");
orc_program_add_temporary (p, 2, "t3");
orc_program_add_temporary (p, 2, "t4");
orc_program_add_temporary (p, 2, "t5");
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T3, ORC_VAR_S2, ORC_VAR_S6,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_S1, ORC_VAR_S5,
ORC_VAR_D1);
orc_program_append_2 (p, "copyw", 0, ORC_VAR_T5, ORC_VAR_S4, ORC_VAR_D1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T5, ORC_VAR_T5, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "orw", 0, ORC_VAR_T4, ORC_VAR_T5, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T2, ORC_VAR_S3, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T1, ORC_VAR_T3, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "mergewl", 1, ORC_VAR_D1, ORC_VAR_T2, ORC_VAR_T1,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_S1] = (void *) s1;
ex->arrays[ORC_VAR_S2] = (void *) s2;
ex->arrays[ORC_VAR_S3] = (void *) s3;
ex->arrays[ORC_VAR_S4] = (void *) s4;
ex->arrays[ORC_VAR_S5] = (void *) s5;
ex->arrays[ORC_VAR_S6] = (void *) s6;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_merge_gr_bgra */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_gr_bgra (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
int i;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) d1;
ptr4 = (orc_union16 *) s1;
ptr5 = (orc_union16 *) s2;
ptr6 = (orc_union16 *) s3;
ptr7 = (orc_union16 *) s4;
ptr8 = (orc_union16 *) s5;
ptr9 = (orc_union16 *) s6;
/* 9: loadpw */
var42.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 11: loadpw */
var43.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 16: loadpb */
var45.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var45.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: loadw */
var38 = ptr8[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr5[i];
/* 4: loadw */
var40 = ptr9[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr6[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 14: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[0];
_dest.x2[1] = var53.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[1];
_dest.x2[1] = var53.x2[1];
var54.x2[1] = _dest.i;
}
/* 15: loadw */
var44 = ptr7[i];
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_merge_gr_bgra (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) ex->arrays[0];
ptr4 = (orc_union16 *) ex->arrays[4];
ptr5 = (orc_union16 *) ex->arrays[5];
ptr6 = (orc_union16 *) ex->arrays[6];
ptr7 = (orc_union16 *) ex->arrays[7];
ptr8 = (orc_union16 *) ex->arrays[8];
ptr9 = (orc_union16 *) ex->arrays[9];
/* 9: loadpw */
var42.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 11: loadpw */
var43.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 16: loadpb */
var45.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var45.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: loadw */
var38 = ptr8[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr5[i];
/* 4: loadw */
var40 = ptr9[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr6[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 14: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[0];
_dest.x2[1] = var53.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[1];
_dest.x2[1] = var53.x2[1];
var54.x2[1] = _dest.i;
}
/* 15: loadw */
var44 = ptr7[i];
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_gr_bgra (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_merge_gr_bgra");
orc_program_set_backup_function (p, _backup_bayer_orc_merge_gr_bgra);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 8, "d1");
orc_program_add_source (p, 2, "s1");
orc_program_add_source (p, 2, "s2");
orc_program_add_source (p, 2, "s3");
orc_program_add_source (p, 2, "s4");
orc_program_add_source (p, 2, "s5");
orc_program_add_source (p, 2, "s6");
orc_program_add_constant (p, 4, 0x0000ff00, "c1");
orc_program_add_constant (p, 4, 0x000000ff, "c2");
orc_program_add_temporary (p, 4, "t1");
orc_program_add_temporary (p, 4, "t2");
orc_program_add_temporary (p, 2, "t3");
orc_program_add_temporary (p, 2, "t4");
orc_program_add_temporary (p, 2, "t5");
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T3, ORC_VAR_S1, ORC_VAR_S5,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_S2, ORC_VAR_S6,
ORC_VAR_D1);
orc_program_append_2 (p, "copyw", 0, ORC_VAR_T5, ORC_VAR_S3, ORC_VAR_D1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T5, ORC_VAR_T5, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "orw", 0, ORC_VAR_T4, ORC_VAR_T5, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T2, ORC_VAR_T3, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T1, ORC_VAR_S4, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "mergewl", 1, ORC_VAR_D1, ORC_VAR_T2, ORC_VAR_T1,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_S1] = (void *) s1;
ex->arrays[ORC_VAR_S2] = (void *) s2;
ex->arrays[ORC_VAR_S3] = (void *) s3;
ex->arrays[ORC_VAR_S4] = (void *) s4;
ex->arrays[ORC_VAR_S5] = (void *) s5;
ex->arrays[ORC_VAR_S6] = (void *) s6;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_merge_bg_abgr */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_bg_abgr (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
int i;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) d1;
ptr4 = (orc_union16 *) s1;
ptr5 = (orc_union16 *) s2;
ptr6 = (orc_union16 *) s3;
ptr7 = (orc_union16 *) s4;
ptr8 = (orc_union16 *) s5;
ptr9 = (orc_union16 *) s6;
/* 9: loadpw */
var42.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 11: loadpw */
var43.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 14: loadpb */
var44.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var44.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr5[i];
/* 1: loadw */
var38 = ptr9[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr4[i];
/* 4: loadw */
var40 = ptr8[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr7[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 15: loadw */
var45 = ptr6[i];
/* 16: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var45.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var45.x2[1];
var54.x2[1] = _dest.i;
}
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[0];
_dest.x2[1] = var47.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[1];
_dest.x2[1] = var47.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_merge_bg_abgr (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) ex->arrays[0];
ptr4 = (orc_union16 *) ex->arrays[4];
ptr5 = (orc_union16 *) ex->arrays[5];
ptr6 = (orc_union16 *) ex->arrays[6];
ptr7 = (orc_union16 *) ex->arrays[7];
ptr8 = (orc_union16 *) ex->arrays[8];
ptr9 = (orc_union16 *) ex->arrays[9];
/* 9: loadpw */
var42.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 11: loadpw */
var43.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 14: loadpb */
var44.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var44.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr5[i];
/* 1: loadw */
var38 = ptr9[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr4[i];
/* 4: loadw */
var40 = ptr8[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr7[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 15: loadw */
var45 = ptr6[i];
/* 16: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var45.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var45.x2[1];
var54.x2[1] = _dest.i;
}
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[0];
_dest.x2[1] = var47.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[1];
_dest.x2[1] = var47.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_bg_abgr (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_merge_bg_abgr");
orc_program_set_backup_function (p, _backup_bayer_orc_merge_bg_abgr);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 8, "d1");
orc_program_add_source (p, 2, "s1");
orc_program_add_source (p, 2, "s2");
orc_program_add_source (p, 2, "s3");
orc_program_add_source (p, 2, "s4");
orc_program_add_source (p, 2, "s5");
orc_program_add_source (p, 2, "s6");
orc_program_add_constant (p, 4, 0x000000ff, "c1");
orc_program_add_constant (p, 4, 0x0000ff00, "c2");
orc_program_add_temporary (p, 4, "t1");
orc_program_add_temporary (p, 4, "t2");
orc_program_add_temporary (p, 2, "t3");
orc_program_add_temporary (p, 2, "t4");
orc_program_add_temporary (p, 2, "t5");
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T3, ORC_VAR_S2, ORC_VAR_S6,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_S1, ORC_VAR_S5,
ORC_VAR_D1);
orc_program_append_2 (p, "copyw", 0, ORC_VAR_T5, ORC_VAR_S4, ORC_VAR_D1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T5, ORC_VAR_T5, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "orw", 0, ORC_VAR_T4, ORC_VAR_T5, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T1, ORC_VAR_C1, ORC_VAR_S3,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T2, ORC_VAR_T4, ORC_VAR_T3,
ORC_VAR_D1);
orc_program_append_2 (p, "mergewl", 1, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_T2,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_S1] = (void *) s1;
ex->arrays[ORC_VAR_S2] = (void *) s2;
ex->arrays[ORC_VAR_S3] = (void *) s3;
ex->arrays[ORC_VAR_S4] = (void *) s4;
ex->arrays[ORC_VAR_S5] = (void *) s5;
ex->arrays[ORC_VAR_S6] = (void *) s6;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_merge_gr_abgr */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_gr_abgr (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
int i;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) d1;
ptr4 = (orc_union16 *) s1;
ptr5 = (orc_union16 *) s2;
ptr6 = (orc_union16 *) s3;
ptr7 = (orc_union16 *) s4;
ptr8 = (orc_union16 *) s5;
ptr9 = (orc_union16 *) s6;
/* 9: loadpw */
var42.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 11: loadpw */
var43.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 14: loadpb */
var44.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var44.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: loadw */
var38 = ptr8[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr5[i];
/* 4: loadw */
var40 = ptr9[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr6[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 15: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var47.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var47.x2[1];
var54.x2[1] = _dest.i;
}
/* 16: loadw */
var45 = ptr7[i];
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_merge_gr_abgr (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) ex->arrays[0];
ptr4 = (orc_union16 *) ex->arrays[4];
ptr5 = (orc_union16 *) ex->arrays[5];
ptr6 = (orc_union16 *) ex->arrays[6];
ptr7 = (orc_union16 *) ex->arrays[7];
ptr8 = (orc_union16 *) ex->arrays[8];
ptr9 = (orc_union16 *) ex->arrays[9];
/* 9: loadpw */
var42.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 11: loadpw */
var43.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 14: loadpb */
var44.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var44.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: loadw */
var38 = ptr8[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr5[i];
/* 4: loadw */
var40 = ptr9[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr6[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 15: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var47.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var47.x2[1];
var54.x2[1] = _dest.i;
}
/* 16: loadw */
var45 = ptr7[i];
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_gr_abgr (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_merge_gr_abgr");
orc_program_set_backup_function (p, _backup_bayer_orc_merge_gr_abgr);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 8, "d1");
orc_program_add_source (p, 2, "s1");
orc_program_add_source (p, 2, "s2");
orc_program_add_source (p, 2, "s3");
orc_program_add_source (p, 2, "s4");
orc_program_add_source (p, 2, "s5");
orc_program_add_source (p, 2, "s6");
orc_program_add_constant (p, 4, 0x0000ff00, "c1");
orc_program_add_constant (p, 4, 0x000000ff, "c2");
orc_program_add_temporary (p, 4, "t1");
orc_program_add_temporary (p, 4, "t2");
orc_program_add_temporary (p, 2, "t3");
orc_program_add_temporary (p, 2, "t4");
orc_program_add_temporary (p, 2, "t5");
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T3, ORC_VAR_S1, ORC_VAR_S5,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_S2, ORC_VAR_S6,
ORC_VAR_D1);
orc_program_append_2 (p, "copyw", 0, ORC_VAR_T5, ORC_VAR_S3, ORC_VAR_D1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T5, ORC_VAR_T5, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "orw", 0, ORC_VAR_T4, ORC_VAR_T5, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T1, ORC_VAR_C2, ORC_VAR_T3,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T2, ORC_VAR_T4, ORC_VAR_S4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergewl", 1, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_T2,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_S1] = (void *) s1;
ex->arrays[ORC_VAR_S2] = (void *) s2;
ex->arrays[ORC_VAR_S3] = (void *) s3;
ex->arrays[ORC_VAR_S4] = (void *) s4;
ex->arrays[ORC_VAR_S5] = (void *) s5;
ex->arrays[ORC_VAR_S6] = (void *) s6;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_merge_bg_rgba */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_bg_rgba (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
int i;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) d1;
ptr4 = (orc_union16 *) s1;
ptr5 = (orc_union16 *) s2;
ptr6 = (orc_union16 *) s3;
ptr7 = (orc_union16 *) s4;
ptr8 = (orc_union16 *) s5;
ptr9 = (orc_union16 *) s6;
/* 9: loadpw */
var42.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 11: loadpw */
var43.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 16: loadpb */
var45.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var45.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr5[i];
/* 1: loadw */
var38 = ptr9[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr4[i];
/* 4: loadw */
var40 = ptr8[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr7[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 14: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[0];
_dest.x2[1] = var53.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[1];
_dest.x2[1] = var53.x2[1];
var54.x2[1] = _dest.i;
}
/* 15: loadw */
var44 = ptr6[i];
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_merge_bg_rgba (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) ex->arrays[0];
ptr4 = (orc_union16 *) ex->arrays[4];
ptr5 = (orc_union16 *) ex->arrays[5];
ptr6 = (orc_union16 *) ex->arrays[6];
ptr7 = (orc_union16 *) ex->arrays[7];
ptr8 = (orc_union16 *) ex->arrays[8];
ptr9 = (orc_union16 *) ex->arrays[9];
/* 9: loadpw */
var42.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 11: loadpw */
var43.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 16: loadpb */
var45.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var45.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr5[i];
/* 1: loadw */
var38 = ptr9[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr4[i];
/* 4: loadw */
var40 = ptr8[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr7[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 14: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[0];
_dest.x2[1] = var53.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[1];
_dest.x2[1] = var53.x2[1];
var54.x2[1] = _dest.i;
}
/* 15: loadw */
var44 = ptr6[i];
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_bg_rgba (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_merge_bg_rgba");
orc_program_set_backup_function (p, _backup_bayer_orc_merge_bg_rgba);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 8, "d1");
orc_program_add_source (p, 2, "s1");
orc_program_add_source (p, 2, "s2");
orc_program_add_source (p, 2, "s3");
orc_program_add_source (p, 2, "s4");
orc_program_add_source (p, 2, "s5");
orc_program_add_source (p, 2, "s6");
orc_program_add_constant (p, 4, 0x000000ff, "c1");
orc_program_add_constant (p, 4, 0x0000ff00, "c2");
orc_program_add_temporary (p, 4, "t1");
orc_program_add_temporary (p, 4, "t2");
orc_program_add_temporary (p, 2, "t3");
orc_program_add_temporary (p, 2, "t4");
orc_program_add_temporary (p, 2, "t5");
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T3, ORC_VAR_S2, ORC_VAR_S6,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_S1, ORC_VAR_S5,
ORC_VAR_D1);
orc_program_append_2 (p, "copyw", 0, ORC_VAR_T5, ORC_VAR_S4, ORC_VAR_D1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T5, ORC_VAR_T5, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "orw", 0, ORC_VAR_T4, ORC_VAR_T5, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T1, ORC_VAR_T3, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T2, ORC_VAR_S3, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "mergewl", 1, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_T2,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_S1] = (void *) s1;
ex->arrays[ORC_VAR_S2] = (void *) s2;
ex->arrays[ORC_VAR_S3] = (void *) s3;
ex->arrays[ORC_VAR_S4] = (void *) s4;
ex->arrays[ORC_VAR_S5] = (void *) s5;
ex->arrays[ORC_VAR_S6] = (void *) s6;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_merge_gr_rgba */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_gr_rgba (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
int i;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) d1;
ptr4 = (orc_union16 *) s1;
ptr5 = (orc_union16 *) s2;
ptr6 = (orc_union16 *) s3;
ptr7 = (orc_union16 *) s4;
ptr8 = (orc_union16 *) s5;
ptr9 = (orc_union16 *) s6;
/* 9: loadpw */
var42.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 11: loadpw */
var43.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 16: loadpb */
var45.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var45.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: loadw */
var38 = ptr8[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr5[i];
/* 4: loadw */
var40 = ptr9[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr6[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 14: loadw */
var44 = ptr7[i];
/* 15: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var53.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var53.x2[1];
var54.x2[1] = _dest.i;
}
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_merge_gr_rgba (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) ex->arrays[0];
ptr4 = (orc_union16 *) ex->arrays[4];
ptr5 = (orc_union16 *) ex->arrays[5];
ptr6 = (orc_union16 *) ex->arrays[6];
ptr7 = (orc_union16 *) ex->arrays[7];
ptr8 = (orc_union16 *) ex->arrays[8];
ptr9 = (orc_union16 *) ex->arrays[9];
/* 9: loadpw */
var42.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 11: loadpw */
var43.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 16: loadpb */
var45.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var45.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: loadw */
var38 = ptr8[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr5[i];
/* 4: loadw */
var40 = ptr9[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr6[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 14: loadw */
var44 = ptr7[i];
/* 15: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var53.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var53.x2[1];
var54.x2[1] = _dest.i;
}
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var47.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_gr_rgba (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_merge_gr_rgba");
orc_program_set_backup_function (p, _backup_bayer_orc_merge_gr_rgba);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 8, "d1");
orc_program_add_source (p, 2, "s1");
orc_program_add_source (p, 2, "s2");
orc_program_add_source (p, 2, "s3");
orc_program_add_source (p, 2, "s4");
orc_program_add_source (p, 2, "s5");
orc_program_add_source (p, 2, "s6");
orc_program_add_constant (p, 4, 0x0000ff00, "c1");
orc_program_add_constant (p, 4, 0x000000ff, "c2");
orc_program_add_temporary (p, 4, "t1");
orc_program_add_temporary (p, 4, "t2");
orc_program_add_temporary (p, 2, "t3");
orc_program_add_temporary (p, 2, "t4");
orc_program_add_temporary (p, 2, "t5");
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T3, ORC_VAR_S1, ORC_VAR_S5,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_S2, ORC_VAR_S6,
ORC_VAR_D1);
orc_program_append_2 (p, "copyw", 0, ORC_VAR_T5, ORC_VAR_S3, ORC_VAR_D1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T5, ORC_VAR_T5, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "orw", 0, ORC_VAR_T4, ORC_VAR_T5, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T1, ORC_VAR_S4, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T2, ORC_VAR_T3, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "mergewl", 1, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_T2,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_S1] = (void *) s1;
ex->arrays[ORC_VAR_S2] = (void *) s2;
ex->arrays[ORC_VAR_S3] = (void *) s3;
ex->arrays[ORC_VAR_S4] = (void *) s4;
ex->arrays[ORC_VAR_S5] = (void *) s5;
ex->arrays[ORC_VAR_S6] = (void *) s6;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_merge_bg_argb */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_bg_argb (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
int i;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) d1;
ptr4 = (orc_union16 *) s1;
ptr5 = (orc_union16 *) s2;
ptr6 = (orc_union16 *) s3;
ptr7 = (orc_union16 *) s4;
ptr8 = (orc_union16 *) s5;
ptr9 = (orc_union16 *) s6;
/* 9: loadpw */
var42.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 11: loadpw */
var43.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 14: loadpb */
var44.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var44.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr5[i];
/* 1: loadw */
var38 = ptr9[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr4[i];
/* 4: loadw */
var40 = ptr8[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr7[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 15: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var47.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var47.x2[1];
var54.x2[1] = _dest.i;
}
/* 16: loadw */
var45 = ptr6[i];
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_merge_bg_argb (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) ex->arrays[0];
ptr4 = (orc_union16 *) ex->arrays[4];
ptr5 = (orc_union16 *) ex->arrays[5];
ptr6 = (orc_union16 *) ex->arrays[6];
ptr7 = (orc_union16 *) ex->arrays[7];
ptr8 = (orc_union16 *) ex->arrays[8];
ptr9 = (orc_union16 *) ex->arrays[9];
/* 9: loadpw */
var42.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 11: loadpw */
var43.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 14: loadpb */
var44.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var44.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr5[i];
/* 1: loadw */
var38 = ptr9[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr4[i];
/* 4: loadw */
var40 = ptr8[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr7[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 15: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var47.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var47.x2[1];
var54.x2[1] = _dest.i;
}
/* 16: loadw */
var45 = ptr6[i];
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[0];
_dest.x2[1] = var45.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[1];
_dest.x2[1] = var45.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_bg_argb (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_merge_bg_argb");
orc_program_set_backup_function (p, _backup_bayer_orc_merge_bg_argb);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 8, "d1");
orc_program_add_source (p, 2, "s1");
orc_program_add_source (p, 2, "s2");
orc_program_add_source (p, 2, "s3");
orc_program_add_source (p, 2, "s4");
orc_program_add_source (p, 2, "s5");
orc_program_add_source (p, 2, "s6");
orc_program_add_constant (p, 4, 0x000000ff, "c1");
orc_program_add_constant (p, 4, 0x0000ff00, "c2");
orc_program_add_temporary (p, 4, "t1");
orc_program_add_temporary (p, 4, "t2");
orc_program_add_temporary (p, 2, "t3");
orc_program_add_temporary (p, 2, "t4");
orc_program_add_temporary (p, 2, "t5");
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T3, ORC_VAR_S2, ORC_VAR_S6,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_S1, ORC_VAR_S5,
ORC_VAR_D1);
orc_program_append_2 (p, "copyw", 0, ORC_VAR_T5, ORC_VAR_S4, ORC_VAR_D1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T5, ORC_VAR_T5, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "orw", 0, ORC_VAR_T4, ORC_VAR_T5, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T1, ORC_VAR_C1, ORC_VAR_T3,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T2, ORC_VAR_T4, ORC_VAR_S3,
ORC_VAR_D1);
orc_program_append_2 (p, "mergewl", 1, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_T2,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_S1] = (void *) s1;
ex->arrays[ORC_VAR_S2] = (void *) s2;
ex->arrays[ORC_VAR_S3] = (void *) s3;
ex->arrays[ORC_VAR_S4] = (void *) s4;
ex->arrays[ORC_VAR_S5] = (void *) s5;
ex->arrays[ORC_VAR_S6] = (void *) s6;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
/* bayer_orc_merge_gr_argb */
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
#ifdef DISABLE_ORC
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_gr_argb (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
int i;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) d1;
ptr4 = (orc_union16 *) s1;
ptr5 = (orc_union16 *) s2;
ptr6 = (orc_union16 *) s3;
ptr7 = (orc_union16 *) s4;
ptr8 = (orc_union16 *) s5;
ptr9 = (orc_union16 *) s6;
/* 9: loadpw */
var42.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 11: loadpw */
var43.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 14: loadpb */
var44.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var44.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: loadw */
var38 = ptr8[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr5[i];
/* 4: loadw */
var40 = ptr9[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr6[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 15: loadw */
var45 = ptr7[i];
/* 16: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var45.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var45.x2[1];
var54.x2[1] = _dest.i;
}
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[0];
_dest.x2[1] = var47.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[1];
_dest.x2[1] = var47.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
#else
static void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
_backup_bayer_orc_merge_gr_argb (OrcExecutor * ORC_RESTRICT ex)
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
{
int i;
int n = ex->n;
orc_union64 *ORC_RESTRICT ptr0;
const orc_union16 *ORC_RESTRICT ptr4;
const orc_union16 *ORC_RESTRICT ptr5;
const orc_union16 *ORC_RESTRICT ptr6;
const orc_union16 *ORC_RESTRICT ptr7;
const orc_union16 *ORC_RESTRICT ptr8;
const orc_union16 *ORC_RESTRICT ptr9;
orc_union16 var37;
orc_union16 var38;
orc_union16 var39;
orc_union16 var40;
orc_union16 var41;
orc_union16 var42;
orc_union16 var43;
orc_union16 var44;
orc_union16 var45;
orc_union64 var46;
orc_union16 var47;
orc_union16 var48;
orc_union16 var49;
orc_union16 var50;
orc_union16 var51;
orc_union16 var52;
orc_union16 var53;
orc_union32 var54;
orc_union32 var55;
ptr0 = (orc_union64 *) ex->arrays[0];
ptr4 = (orc_union16 *) ex->arrays[4];
ptr5 = (orc_union16 *) ex->arrays[5];
ptr6 = (orc_union16 *) ex->arrays[6];
ptr7 = (orc_union16 *) ex->arrays[7];
ptr8 = (orc_union16 *) ex->arrays[8];
ptr9 = (orc_union16 *) ex->arrays[9];
/* 9: loadpw */
var42.i = (int) 0x0000ff00; /* 65280 or 3.22526e-319f */
/* 11: loadpw */
var43.i = (int) 0x000000ff; /* 255 or 1.25987e-321f */
/* 14: loadpb */
var44.x2[0] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
var44.x2[1] = (int) 0x000000ff; /* 255 or 1.25987e-321f */
for (i = 0; i < n; i++) {
/* 0: loadw */
var37 = ptr4[i];
/* 1: loadw */
var38 = ptr8[i];
/* 2: avgub */
var47.x2[0] = ((orc_uint8) var37.x2[0] + (orc_uint8) var38.x2[0] + 1) >> 1;
var47.x2[1] = ((orc_uint8) var37.x2[1] + (orc_uint8) var38.x2[1] + 1) >> 1;
/* 3: loadw */
var39 = ptr5[i];
/* 4: loadw */
var40 = ptr9[i];
/* 5: avgub */
var48.x2[0] = ((orc_uint8) var39.x2[0] + (orc_uint8) var40.x2[0] + 1) >> 1;
var48.x2[1] = ((orc_uint8) var39.x2[1] + (orc_uint8) var40.x2[1] + 1) >> 1;
/* 6: loadw */
var41 = ptr6[i];
/* 7: copyw */
var49.i = var41.i;
/* 8: avgub */
var50.x2[0] = ((orc_uint8) var48.x2[0] + (orc_uint8) var49.x2[0] + 1) >> 1;
var50.x2[1] = ((orc_uint8) var48.x2[1] + (orc_uint8) var49.x2[1] + 1) >> 1;
/* 10: andw */
var51.i = var50.i & var42.i;
/* 12: andw */
var52.i = var49.i & var43.i;
/* 13: orw */
var53.i = var52.i | var51.i;
/* 15: loadw */
var45 = ptr7[i];
/* 16: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[0];
_dest.x2[1] = var45.x2[0];
var54.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var44.x2[1];
_dest.x2[1] = var45.x2[1];
var54.x2[1] = _dest.i;
}
/* 17: mergebw */
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[0];
_dest.x2[1] = var47.x2[0];
var55.x2[0] = _dest.i;
}
{
orc_union16 _dest;
_dest.x2[0] = var53.x2[1];
_dest.x2[1] = var47.x2[1];
var55.x2[1] = _dest.i;
}
/* 18: mergewl */
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[0];
_dest.x2[1] = var55.x2[0];
var46.x2[0] = _dest.i;
}
{
orc_union32 _dest;
_dest.x2[0] = var54.x2[1];
_dest.x2[1] = var55.x2[1];
var46.x2[1] = _dest.i;
}
/* 19: storeq */
ptr0[i] = var46;
}
}
void
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
bayer_orc_merge_gr_argb (guint8 * ORC_RESTRICT d1,
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
const guint8 * ORC_RESTRICT s1, const guint8 * ORC_RESTRICT s2,
const guint8 * ORC_RESTRICT s3, const guint8 * ORC_RESTRICT s4,
const guint8 * ORC_RESTRICT s5, const guint8 * ORC_RESTRICT s6, int n)
{
OrcExecutor _ex, *ex = &_ex;
Release 0.11.90
2012-04-12 08:50:10 +00:00
static volatile int p_inited = 0;
Release 0.11.92
2012-06-07 14:23:09 +00:00
static OrcCode *c = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
void (*func) (OrcExecutor *);
if (!p_inited) {
orc_once_mutex_lock ();
if (!p_inited) {
Release 0.11.92
2012-06-07 14:23:09 +00:00
OrcProgram *p;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
p = orc_program_new ();
bayer: prefix orc functions with bayer_orc Add bayer orc to the unit test
2012-07-24 08:32:51 +00:00
orc_program_set_name (p, "bayer_orc_merge_gr_argb");
orc_program_set_backup_function (p, _backup_bayer_orc_merge_gr_argb);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
orc_program_add_destination (p, 8, "d1");
orc_program_add_source (p, 2, "s1");
orc_program_add_source (p, 2, "s2");
orc_program_add_source (p, 2, "s3");
orc_program_add_source (p, 2, "s4");
orc_program_add_source (p, 2, "s5");
orc_program_add_source (p, 2, "s6");
orc_program_add_constant (p, 4, 0x0000ff00, "c1");
orc_program_add_constant (p, 4, 0x000000ff, "c2");
orc_program_add_temporary (p, 4, "t1");
orc_program_add_temporary (p, 4, "t2");
orc_program_add_temporary (p, 2, "t3");
orc_program_add_temporary (p, 2, "t4");
orc_program_add_temporary (p, 2, "t5");
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T3, ORC_VAR_S1, ORC_VAR_S5,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_S2, ORC_VAR_S6,
ORC_VAR_D1);
orc_program_append_2 (p, "copyw", 0, ORC_VAR_T5, ORC_VAR_S3, ORC_VAR_D1,
ORC_VAR_D1);
orc_program_append_2 (p, "avgub", 1, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_T5,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T4, ORC_VAR_T4, ORC_VAR_C1,
ORC_VAR_D1);
orc_program_append_2 (p, "andw", 0, ORC_VAR_T5, ORC_VAR_T5, ORC_VAR_C2,
ORC_VAR_D1);
orc_program_append_2 (p, "orw", 0, ORC_VAR_T4, ORC_VAR_T5, ORC_VAR_T4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T1, ORC_VAR_C2, ORC_VAR_S4,
ORC_VAR_D1);
orc_program_append_2 (p, "mergebw", 1, ORC_VAR_T2, ORC_VAR_T4, ORC_VAR_T3,
ORC_VAR_D1);
orc_program_append_2 (p, "mergewl", 1, ORC_VAR_D1, ORC_VAR_T1, ORC_VAR_T2,
ORC_VAR_D1);
orc_program_compile (p);
Release 0.11.92
2012-06-07 14:23:09 +00:00
c = orc_program_take_code (p);
orc_program_free (p);
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
}
p_inited = TRUE;
orc_once_mutex_unlock ();
}
Release 0.11.92
2012-06-07 14:23:09 +00:00
ex->arrays[ORC_VAR_A2] = c;
ex->program = 0;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
ex->n = n;
ex->arrays[ORC_VAR_D1] = d1;
ex->arrays[ORC_VAR_S1] = (void *) s1;
ex->arrays[ORC_VAR_S2] = (void *) s2;
ex->arrays[ORC_VAR_S3] = (void *) s3;
ex->arrays[ORC_VAR_S4] = (void *) s4;
ex->arrays[ORC_VAR_S5] = (void *) s5;
ex->arrays[ORC_VAR_S6] = (void *) s6;
Release 0.11.92
2012-06-07 14:23:09 +00:00
func = c->exec;
bayer2rgb: Convert to Orc Seriously faster. Algorithm is nearly the same as bilinear, which given the speed of this code, should be considered the baseline of quality. Speed appears to be limited by memory bandwidth, so I didn't bother trying to make it any faster.
2011-05-31 06:43:39 +00:00
func (ex);
}
#endif
Reference in a new issue Copy permalink