Skip to content

F
FigureOut
Commit 9eb8ec43 by wangguotao
Options

更新到v2.1.1

parent 063d74b4
Showing with 1330 additions and 118 deletions
MomentaMedia/FigureOut.log
objct name changed "deviceOutputPage3"
objct name changed "deviceOutputPage4"
available device "DeckLink 8K Pro (1)"
available device "DeckLink 8K Pro (2)"
"2024-05-16 14:18:31.039" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
available device "DeckLink 8K Pro (3)"
"2024-05-16 14:18:31.059" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
available device "DeckLink 8K Pro (4)"
"2024-05-16 14:18:31.097" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:18:31.117" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:18:31.137" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:18:31.157" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:18:31.177" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:18:31.197" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:18:31.217" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:18:31.275" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:18:32.299" decklink input fps 50
"2024-05-16 14:18:33.319" decklink input fps 51
"2024-05-16 14:18:33.385" decklink output fps 51 , qsize 0
"2024-05-16 14:18:34.319" decklink input fps 50
"2024-05-16 14:18:34.405" decklink output fps 51 , qsize 0
"2024-05-16 14:18:35.339" decklink input fps 51
"2024-05-16 14:18:35.405" decklink output fps 50 , qsize 0
"2024-05-16 14:18:36.339" decklink input fps 50
"2024-05-16 14:18:36.404" decklink output fps 50 , qsize 0
MomentaMedia/include/Utils/FastMemcpy.h 0 → 100644
//=====================================================================
//
// FastMemcpy.c - skywind3000@163.com, 2015
//
// feature:
// 50% speed up in avg. vs standard memcpy (tested in vc2012/gcc5.1)
//
//=====================================================================
#ifndef __FAST_MEMCPY_H__
#define __FAST_MEMCPY_H__
#include <stddef.h>
#include <stdint.h>
#include <emmintrin.h>
//---------------------------------------------------------------------
// force inline for compilers
//---------------------------------------------------------------------
#ifndef INLINE
#ifdef __GNUC__
#if (__GNUC__ > 3) || ((__GNUC__ == 3) && (__GNUC_MINOR__ >= 1))
#define INLINE __inline__ __attribute__((always_inline))
#else
#define INLINE __inline__
#endif
#elif defined(_MSC_VER)
#define INLINE __forceinline
#elif (defined(__BORLANDC__) || defined(__WATCOMC__))
#define INLINE __inline
#else
#define INLINE
#endif
#endif
//---------------------------------------------------------------------
// fast copy for different sizes
//---------------------------------------------------------------------
static INLINE void memcpy_sse2_16(void *dst, const void *src) {
__m128i m0 = _mm_loadu_si128(((const __m128i*)src) + 0);
_mm_storeu_si128(((__m128i*)dst) + 0, m0);
}
static INLINE void memcpy_sse2_32(void *dst, const void *src) {
__m128i m0 = _mm_loadu_si128(((const __m128i*)src) + 0);
__m128i m1 = _mm_loadu_si128(((const __m128i*)src) + 1);
_mm_storeu_si128(((__m128i*)dst) + 0, m0);
_mm_storeu_si128(((__m128i*)dst) + 1, m1);
}
static INLINE void memcpy_sse2_64(void *dst, const void *src) {
__m128i m0 = _mm_loadu_si128(((const __m128i*)src) + 0);
__m128i m1 = _mm_loadu_si128(((const __m128i*)src) + 1);
__m128i m2 = _mm_loadu_si128(((const __m128i*)src) + 2);
__m128i m3 = _mm_loadu_si128(((const __m128i*)src) + 3);
_mm_storeu_si128(((__m128i*)dst) + 0, m0);
_mm_storeu_si128(((__m128i*)dst) + 1, m1);
_mm_storeu_si128(((__m128i*)dst) + 2, m2);
_mm_storeu_si128(((__m128i*)dst) + 3, m3);
}
static INLINE void memcpy_sse2_128(void *dst, const void *src) {
__m128i m0 = _mm_loadu_si128(((const __m128i*)src) + 0);
__m128i m1 = _mm_loadu_si128(((const __m128i*)src) + 1);
__m128i m2 = _mm_loadu_si128(((const __m128i*)src) + 2);
__m128i m3 = _mm_loadu_si128(((const __m128i*)src) + 3);
__m128i m4 = _mm_loadu_si128(((const __m128i*)src) + 4);
__m128i m5 = _mm_loadu_si128(((const __m128i*)src) + 5);
__m128i m6 = _mm_loadu_si128(((const __m128i*)src) + 6);
__m128i m7 = _mm_loadu_si128(((const __m128i*)src) + 7);
_mm_storeu_si128(((__m128i*)dst) + 0, m0);
_mm_storeu_si128(((__m128i*)dst) + 1, m1);
_mm_storeu_si128(((__m128i*)dst) + 2, m2);
_mm_storeu_si128(((__m128i*)dst) + 3, m3);
_mm_storeu_si128(((__m128i*)dst) + 4, m4);
_mm_storeu_si128(((__m128i*)dst) + 5, m5);
_mm_storeu_si128(((__m128i*)dst) + 6, m6);
_mm_storeu_si128(((__m128i*)dst) + 7, m7);
}
//---------------------------------------------------------------------
// tiny memory copy with jump table optimized
//---------------------------------------------------------------------
static INLINE void *memcpy_tiny(void *dst, const void *src, size_t size) {
unsigned char *dd = ((unsigned char*)dst) + size;
const unsigned char *ss = ((const unsigned char*)src) + size;
switch (size) {
case 64:
memcpy_sse2_64(dd - 64, ss - 64);
case 0:
break;
case 65:
memcpy_sse2_64(dd - 65, ss - 65);
case 1:
dd[-1] = ss[-1];
break;
case 66:
memcpy_sse2_64(dd - 66, ss - 66);
case 2:
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 67:
memcpy_sse2_64(dd - 67, ss - 67);
case 3:
*((uint16_t*)(dd - 3)) = *((uint16_t*)(ss - 3));
dd[-1] = ss[-1];
break;
case 68:
memcpy_sse2_64(dd - 68, ss - 68);
case 4:
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 69:
memcpy_sse2_64(dd - 69, ss - 69);
case 5:
*((uint32_t*)(dd - 5)) = *((uint32_t*)(ss - 5));
dd[-1] = ss[-1];
break;
case 70:
memcpy_sse2_64(dd - 70, ss - 70);
case 6:
*((uint32_t*)(dd - 6)) = *((uint32_t*)(ss - 6));
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 71:
memcpy_sse2_64(dd - 71, ss - 71);
case 7:
*((uint32_t*)(dd - 7)) = *((uint32_t*)(ss - 7));
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 72:
memcpy_sse2_64(dd - 72, ss - 72);
case 8:
*((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8));
break;
case 73:
memcpy_sse2_64(dd - 73, ss - 73);
case 9:
*((uint64_t*)(dd - 9)) = *((uint64_t*)(ss - 9));
dd[-1] = ss[-1];
break;
case 74:
memcpy_sse2_64(dd - 74, ss - 74);
case 10:
*((uint64_t*)(dd - 10)) = *((uint64_t*)(ss - 10));
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 75:
memcpy_sse2_64(dd - 75, ss - 75);
case 11:
*((uint64_t*)(dd - 11)) = *((uint64_t*)(ss - 11));
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 76:
memcpy_sse2_64(dd - 76, ss - 76);
case 12:
*((uint64_t*)(dd - 12)) = *((uint64_t*)(ss - 12));
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 77:
memcpy_sse2_64(dd - 77, ss - 77);
case 13:
*((uint64_t*)(dd - 13)) = *((uint64_t*)(ss - 13));
*((uint32_t*)(dd - 5)) = *((uint32_t*)(ss - 5));
dd[-1] = ss[-1];
break;
case 78:
memcpy_sse2_64(dd - 78, ss - 78);
case 14:
*((uint64_t*)(dd - 14)) = *((uint64_t*)(ss - 14));
*((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8));
break;
case 79:
memcpy_sse2_64(dd - 79, ss - 79);
case 15:
*((uint64_t*)(dd - 15)) = *((uint64_t*)(ss - 15));
*((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8));
break;
case 80:
memcpy_sse2_64(dd - 80, ss - 80);
case 16:
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 81:
memcpy_sse2_64(dd - 81, ss - 81);
case 17:
memcpy_sse2_16(dd - 17, ss - 17);
dd[-1] = ss[-1];
break;
case 82:
memcpy_sse2_64(dd - 82, ss - 82);
case 18:
memcpy_sse2_16(dd - 18, ss - 18);
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 83:
memcpy_sse2_64(dd - 83, ss - 83);
case 19:
memcpy_sse2_16(dd - 19, ss - 19);
*((uint16_t*)(dd - 3)) = *((uint16_t*)(ss - 3));
dd[-1] = ss[-1];
break;
case 84:
memcpy_sse2_64(dd - 84, ss - 84);
case 20:
memcpy_sse2_16(dd - 20, ss - 20);
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 85:
memcpy_sse2_64(dd - 85, ss - 85);
case 21:
memcpy_sse2_16(dd - 21, ss - 21);
*((uint32_t*)(dd - 5)) = *((uint32_t*)(ss - 5));
dd[-1] = ss[-1];
break;
case 86:
memcpy_sse2_64(dd - 86, ss - 86);
case 22:
memcpy_sse2_16(dd - 22, ss - 22);
*((uint32_t*)(dd - 6)) = *((uint32_t*)(ss - 6));
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 87:
memcpy_sse2_64(dd - 87, ss - 87);
case 23:
memcpy_sse2_16(dd - 23, ss - 23);
*((uint32_t*)(dd - 7)) = *((uint32_t*)(ss - 7));
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 88:
memcpy_sse2_64(dd - 88, ss - 88);
case 24:
memcpy_sse2_16(dd - 24, ss - 24);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 89:
memcpy_sse2_64(dd - 89, ss - 89);
case 25:
memcpy_sse2_16(dd - 25, ss - 25);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 90:
memcpy_sse2_64(dd - 90, ss - 90);
case 26:
memcpy_sse2_16(dd - 26, ss - 26);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 91:
memcpy_sse2_64(dd - 91, ss - 91);
case 27:
memcpy_sse2_16(dd - 27, ss - 27);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 92:
memcpy_sse2_64(dd - 92, ss - 92);
case 28:
memcpy_sse2_16(dd - 28, ss - 28);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 93:
memcpy_sse2_64(dd - 93, ss - 93);
case 29:
memcpy_sse2_16(dd - 29, ss - 29);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 94:
memcpy_sse2_64(dd - 94, ss - 94);
case 30:
memcpy_sse2_16(dd - 30, ss - 30);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 95:
memcpy_sse2_64(dd - 95, ss - 95);
case 31:
memcpy_sse2_16(dd - 31, ss - 31);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 96:
memcpy_sse2_64(dd - 96, ss - 96);
case 32:
memcpy_sse2_32(dd - 32, ss - 32);
break;
case 97:
memcpy_sse2_64(dd - 97, ss - 97);
case 33:
memcpy_sse2_32(dd - 33, ss - 33);
dd[-1] = ss[-1];
break;
case 98:
memcpy_sse2_64(dd - 98, ss - 98);
case 34:
memcpy_sse2_32(dd - 34, ss - 34);
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 99:
memcpy_sse2_64(dd - 99, ss - 99);
case 35:
memcpy_sse2_32(dd - 35, ss - 35);
*((uint16_t*)(dd - 3)) = *((uint16_t*)(ss - 3));
dd[-1] = ss[-1];
break;
case 100:
memcpy_sse2_64(dd - 100, ss - 100);
case 36:
memcpy_sse2_32(dd - 36, ss - 36);
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 101:
memcpy_sse2_64(dd - 101, ss - 101);
case 37:
memcpy_sse2_32(dd - 37, ss - 37);
*((uint32_t*)(dd - 5)) = *((uint32_t*)(ss - 5));
dd[-1] = ss[-1];
break;
case 102:
memcpy_sse2_64(dd - 102, ss - 102);
case 38:
memcpy_sse2_32(dd - 38, ss - 38);
*((uint32_t*)(dd - 6)) = *((uint32_t*)(ss - 6));
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 103:
memcpy_sse2_64(dd - 103, ss - 103);
case 39:
memcpy_sse2_32(dd - 39, ss - 39);
*((uint32_t*)(dd - 7)) = *((uint32_t*)(ss - 7));
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 104:
memcpy_sse2_64(dd - 104, ss - 104);
case 40:
memcpy_sse2_32(dd - 40, ss - 40);
*((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8));
break;
case 105:
memcpy_sse2_64(dd - 105, ss - 105);
case 41:
memcpy_sse2_32(dd - 41, ss - 41);
*((uint64_t*)(dd - 9)) = *((uint64_t*)(ss - 9));
dd[-1] = ss[-1];
break;
case 106:
memcpy_sse2_64(dd - 106, ss - 106);
case 42:
memcpy_sse2_32(dd - 42, ss - 42);
*((uint64_t*)(dd - 10)) = *((uint64_t*)(ss - 10));
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 107:
memcpy_sse2_64(dd - 107, ss - 107);
case 43:
memcpy_sse2_32(dd - 43, ss - 43);
*((uint64_t*)(dd - 11)) = *((uint64_t*)(ss - 11));
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 108:
memcpy_sse2_64(dd - 108, ss - 108);
case 44:
memcpy_sse2_32(dd - 44, ss - 44);
*((uint64_t*)(dd - 12)) = *((uint64_t*)(ss - 12));
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 109:
memcpy_sse2_64(dd - 109, ss - 109);
case 45:
memcpy_sse2_32(dd - 45, ss - 45);
*((uint64_t*)(dd - 13)) = *((uint64_t*)(ss - 13));
*((uint32_t*)(dd - 5)) = *((uint32_t*)(ss - 5));
dd[-1] = ss[-1];
break;
case 110:
memcpy_sse2_64(dd - 110, ss - 110);
case 46:
memcpy_sse2_32(dd - 46, ss - 46);
*((uint64_t*)(dd - 14)) = *((uint64_t*)(ss - 14));
*((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8));
break;
case 111:
memcpy_sse2_64(dd - 111, ss - 111);
case 47:
memcpy_sse2_32(dd - 47, ss - 47);
*((uint64_t*)(dd - 15)) = *((uint64_t*)(ss - 15));
*((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8));
break;
case 112:
memcpy_sse2_64(dd - 112, ss - 112);
case 48:
memcpy_sse2_32(dd - 48, ss - 48);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 113:
memcpy_sse2_64(dd - 113, ss - 113);
case 49:
memcpy_sse2_32(dd - 49, ss - 49);
memcpy_sse2_16(dd - 17, ss - 17);
dd[-1] = ss[-1];
break;
case 114:
memcpy_sse2_64(dd - 114, ss - 114);
case 50:
memcpy_sse2_32(dd - 50, ss - 50);
memcpy_sse2_16(dd - 18, ss - 18);
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 115:
memcpy_sse2_64(dd - 115, ss - 115);
case 51:
memcpy_sse2_32(dd - 51, ss - 51);
memcpy_sse2_16(dd - 19, ss - 19);
*((uint16_t*)(dd - 3)) = *((uint16_t*)(ss - 3));
dd[-1] = ss[-1];
break;
case 116:
memcpy_sse2_64(dd - 116, ss - 116);
case 52:
memcpy_sse2_32(dd - 52, ss - 52);
memcpy_sse2_16(dd - 20, ss - 20);
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 117:
memcpy_sse2_64(dd - 117, ss - 117);
case 53:
memcpy_sse2_32(dd - 53, ss - 53);
memcpy_sse2_16(dd - 21, ss - 21);
*((uint32_t*)(dd - 5)) = *((uint32_t*)(ss - 5));
dd[-1] = ss[-1];
break;
case 118:
memcpy_sse2_64(dd - 118, ss - 118);
case 54:
memcpy_sse2_32(dd - 54, ss - 54);
memcpy_sse2_16(dd - 22, ss - 22);
*((uint32_t*)(dd - 6)) = *((uint32_t*)(ss - 6));
*((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2));
break;
case 119:
memcpy_sse2_64(dd - 119, ss - 119);
case 55:
memcpy_sse2_32(dd - 55, ss - 55);
memcpy_sse2_16(dd - 23, ss - 23);
*((uint32_t*)(dd - 7)) = *((uint32_t*)(ss - 7));
*((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4));
break;
case 120:
memcpy_sse2_64(dd - 120, ss - 120);
case 56:
memcpy_sse2_32(dd - 56, ss - 56);
memcpy_sse2_16(dd - 24, ss - 24);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 121:
memcpy_sse2_64(dd - 121, ss - 121);
case 57:
memcpy_sse2_32(dd - 57, ss - 57);
memcpy_sse2_16(dd - 25, ss - 25);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 122:
memcpy_sse2_64(dd - 122, ss - 122);
case 58:
memcpy_sse2_32(dd - 58, ss - 58);
memcpy_sse2_16(dd - 26, ss - 26);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 123:
memcpy_sse2_64(dd - 123, ss - 123);
case 59:
memcpy_sse2_32(dd - 59, ss - 59);
memcpy_sse2_16(dd - 27, ss - 27);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 124:
memcpy_sse2_64(dd - 124, ss - 124);
case 60:
memcpy_sse2_32(dd - 60, ss - 60);
memcpy_sse2_16(dd - 28, ss - 28);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 125:
memcpy_sse2_64(dd - 125, ss - 125);
case 61:
memcpy_sse2_32(dd - 61, ss - 61);
memcpy_sse2_16(dd - 29, ss - 29);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 126:
memcpy_sse2_64(dd - 126, ss - 126);
case 62:
memcpy_sse2_32(dd - 62, ss - 62);
memcpy_sse2_16(dd - 30, ss - 30);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 127:
memcpy_sse2_64(dd - 127, ss - 127);
case 63:
memcpy_sse2_32(dd - 63, ss - 63);
memcpy_sse2_16(dd - 31, ss - 31);
memcpy_sse2_16(dd - 16, ss - 16);
break;
case 128:
memcpy_sse2_128(dd - 128, ss - 128);
break;
}
return dst;
}
//---------------------------------------------------------------------
// main routine
//---------------------------------------------------------------------
static void* memcpy_fast(void *destination, const void *source, size_t size)
{
unsigned char *dst = (unsigned char*)destination;
const unsigned char *src = (const unsigned char*)source;
static size_t cachesize = 0x200000; // L2-cache size
size_t padding;
// small memory copy
if (size <= 128) {
return memcpy_tiny(dst, src, size);
}
// align destination to 16 bytes boundary
padding = (16 - (((size_t)dst) & 15)) & 15;
if (padding > 0) {
__m128i head = _mm_loadu_si128((const __m128i*)src);
_mm_storeu_si128((__m128i*)dst, head);
dst += padding;
src += padding;
size -= padding;
}
// medium size copy
if (size <= cachesize) {
__m128i c0, c1, c2, c3, c4, c5, c6, c7;
for (; size >= 128; size -= 128) {
c0 = _mm_loadu_si128(((const __m128i*)src) + 0);
c1 = _mm_loadu_si128(((const __m128i*)src) + 1);
c2 = _mm_loadu_si128(((const __m128i*)src) + 2);
c3 = _mm_loadu_si128(((const __m128i*)src) + 3);
c4 = _mm_loadu_si128(((const __m128i*)src) + 4);
c5 = _mm_loadu_si128(((const __m128i*)src) + 5);
c6 = _mm_loadu_si128(((const __m128i*)src) + 6);
c7 = _mm_loadu_si128(((const __m128i*)src) + 7);
_mm_prefetch((const char*)(src + 256), _MM_HINT_NTA);
src += 128;
_mm_store_si128((((__m128i*)dst) + 0), c0);
_mm_store_si128((((__m128i*)dst) + 1), c1);
_mm_store_si128((((__m128i*)dst) + 2), c2);
_mm_store_si128((((__m128i*)dst) + 3), c3);
_mm_store_si128((((__m128i*)dst) + 4), c4);
_mm_store_si128((((__m128i*)dst) + 5), c5);
_mm_store_si128((((__m128i*)dst) + 6), c6);
_mm_store_si128((((__m128i*)dst) + 7), c7);
dst += 128;
}
}
else { // big memory copy
__m128i c0, c1, c2, c3, c4, c5, c6, c7;
_mm_prefetch((const char*)(src), _MM_HINT_NTA);
if ((((size_t)src) & 15) == 0) { // source aligned
for (; size >= 128; size -= 128) {
c0 = _mm_load_si128(((const __m128i*)src) + 0);
c1 = _mm_load_si128(((const __m128i*)src) + 1);
c2 = _mm_load_si128(((const __m128i*)src) + 2);
c3 = _mm_load_si128(((const __m128i*)src) + 3);
c4 = _mm_load_si128(((const __m128i*)src) + 4);
c5 = _mm_load_si128(((const __m128i*)src) + 5);
c6 = _mm_load_si128(((const __m128i*)src) + 6);
c7 = _mm_load_si128(((const __m128i*)src) + 7);
_mm_prefetch((const char*)(src + 256), _MM_HINT_NTA);
src += 128;
_mm_stream_si128((((__m128i*)dst) + 0), c0);
_mm_stream_si128((((__m128i*)dst) + 1), c1);
_mm_stream_si128((((__m128i*)dst) + 2), c2);
_mm_stream_si128((((__m128i*)dst) + 3), c3);
_mm_stream_si128((((__m128i*)dst) + 4), c4);
_mm_stream_si128((((__m128i*)dst) + 5), c5);
_mm_stream_si128((((__m128i*)dst) + 6), c6);
_mm_stream_si128((((__m128i*)dst) + 7), c7);
dst += 128;
}
}
else { // source unaligned
for (; size >= 128; size -= 128) {
c0 = _mm_loadu_si128(((const __m128i*)src) + 0);
c1 = _mm_loadu_si128(((const __m128i*)src) + 1);
c2 = _mm_loadu_si128(((const __m128i*)src) + 2);
c3 = _mm_loadu_si128(((const __m128i*)src) + 3);
c4 = _mm_loadu_si128(((const __m128i*)src) + 4);
c5 = _mm_loadu_si128(((const __m128i*)src) + 5);
c6 = _mm_loadu_si128(((const __m128i*)src) + 6);
c7 = _mm_loadu_si128(((const __m128i*)src) + 7);
_mm_prefetch((const char*)(src + 256), _MM_HINT_NTA);
src += 128;
_mm_stream_si128((((__m128i*)dst) + 0), c0);
_mm_stream_si128((((__m128i*)dst) + 1), c1);
_mm_stream_si128((((__m128i*)dst) + 2), c2);
_mm_stream_si128((((__m128i*)dst) + 3), c3);
_mm_stream_si128((((__m128i*)dst) + 4), c4);
_mm_stream_si128((((__m128i*)dst) + 5), c5);
_mm_stream_si128((((__m128i*)dst) + 6), c6);
_mm_stream_si128((((__m128i*)dst) + 7), c7);
dst += 128;
}
}
_mm_sfence();
}
memcpy_tiny(dst, src, size);
return destination;
}
#endif
MomentaMedia/include/Utils/FastMemcpy_Avx.h 0 → 100644
//=====================================================================
//
// FastMemcpy.c - skywind3000@163.com, 2015
//
// feature:
// 50% speed up in avg. vs standard memcpy (tested in vc2012/gcc5.1)
//
//=====================================================================
#ifndef __FAST_MEMCPY_H__
#define __FAST_MEMCPY_H__
#include <stddef.h>
#include <stdint.h>
#include <immintrin.h>
//---------------------------------------------------------------------
// force inline for compilers
//---------------------------------------------------------------------
#ifndef INLINE
#ifdef __GNUC__
#if (__GNUC__ > 3) || ((__GNUC__ == 3) && (__GNUC_MINOR__ >= 1))
#define INLINE __inline__ __attribute__((always_inline))
#else
#define INLINE __inline__
#endif
#elif defined(_MSC_VER)
#define INLINE __forceinline
#elif (defined(__BORLANDC__) || defined(__WATCOMC__))
#define INLINE __inline
#else
#define INLINE
#endif
#endif
//---------------------------------------------------------------------
// fast copy for different sizes
//---------------------------------------------------------------------
static INLINE void memcpy_avx_16(void *dst, const void *src) {
#if 1
__m128i m0 = _mm_loadu_si128(((const __m128i*)src) + 0);
_mm_storeu_si128(((__m128i*)dst) + 0, m0);
#else
*((uint64_t*)((char*)dst + 0)) = *((uint64_t*)((const char*)src + 0));
*((uint64_t*)((char*)dst + 8)) = *((uint64_t*)((const char*)src + 8));
#endif
}
static INLINE void memcpy_avx_32(void *dst, const void *src) {
__m256i m0 = _mm256_loadu_si256(((const __m256i*)src) + 0);
_mm256_storeu_si256(((__m256i*)dst) + 0, m0);
}
static INLINE void memcpy_avx_64(void *dst, const void *src) {
__m256i m0 = _mm256_loadu_si256(((const __m256i*)src) + 0);
__m256i m1 = _mm256_loadu_si256(((const __m256i*)src) + 1);
_mm256_storeu_si256(((__m256i*)dst) + 0, m0);
_mm256_storeu_si256(((__m256i*)dst) + 1, m1);
}
static INLINE void memcpy_avx_128(void *dst, const void *src) {
__m256i m0 = _mm256_loadu_si256(((const __m256i*)src) + 0);
__m256i m1 = _mm256_loadu_si256(((const __m256i*)src) + 1);
__m256i m2 = _mm256_loadu_si256(((const __m256i*)src) + 2);
__m256i m3 = _mm256_loadu_si256(((const __m256i*)src) + 3);
_mm256_storeu_si256(((__m256i*)dst) + 0, m0);
_mm256_storeu_si256(((__m256i*)dst) + 1, m1);
_mm256_storeu_si256(((__m256i*)dst) + 2, m2);
_mm256_storeu_si256(((__m256i*)dst) + 3, m3);
}
static INLINE void memcpy_avx_256(void *dst, const void *src) {
__m256i m0 = _mm256_loadu_si256(((const __m256i*)src) + 0);
__m256i m1 = _mm256_loadu_si256(((const __m256i*)src) + 1);
__m256i m2 = _mm256_loadu_si256(((const __m256i*)src) + 2);
__m256i m3 = _mm256_loadu_si256(((const __m256i*)src) + 3);
__m256i m4 = _mm256_loadu_si256(((const __m256i*)src) + 4);
__m256i m5 = _mm256_loadu_si256(((const __m256i*)src) + 5);
__m256i m6 = _mm256_loadu_si256(((const __m256i*)src) + 6);
__m256i m7 = _mm256_loadu_si256(((const __m256i*)src) + 7);
_mm256_storeu_si256(((__m256i*)dst) + 0, m0);
_mm256_storeu_si256(((__m256i*)dst) + 1, m1);
_mm256_storeu_si256(((__m256i*)dst) + 2, m2);
_mm256_storeu_si256(((__m256i*)dst) + 3, m3);
_mm256_storeu_si256(((__m256i*)dst) + 4, m4);
_mm256_storeu_si256(((__m256i*)dst) + 5, m5);
_mm256_storeu_si256(((__m256i*)dst) + 6, m6);
_mm256_storeu_si256(((__m256i*)dst) + 7, m7);
}
//---------------------------------------------------------------------
// tiny memory copy with jump table optimized
//---------------------------------------------------------------------
static INLINE void *memcpy_tiny(void *dst, const void *src, size_t size) {
unsigned char *dd = ((unsigned char*)dst) + size;
const unsigned char *ss = ((const unsigned char*)src) + size;
switch (size) {
case 128: memcpy_avx_128(dd - 128, ss - 128);
case 0: break;
case 129: memcpy_avx_128(dd - 129, ss - 129);
case 1: dd[-1] = ss[-1]; break;
case 130: memcpy_avx_128(dd - 130, ss - 130);
case 2: *((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2)); break;
case 131: memcpy_avx_128(dd - 131, ss - 131);
case 3: *((uint16_t*)(dd - 3)) = *((uint16_t*)(ss - 3)); dd[-1] = ss[-1]; break;
case 132: memcpy_avx_128(dd - 132, ss - 132);
case 4: *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 133: memcpy_avx_128(dd - 133, ss - 133);
case 5: *((uint32_t*)(dd - 5)) = *((uint32_t*)(ss - 5)); dd[-1] = ss[-1]; break;
case 134: memcpy_avx_128(dd - 134, ss - 134);
case 6: *((uint32_t*)(dd - 6)) = *((uint32_t*)(ss - 6)); *((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2)); break;
case 135: memcpy_avx_128(dd - 135, ss - 135);
case 7: *((uint32_t*)(dd - 7)) = *((uint32_t*)(ss - 7)); *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 136: memcpy_avx_128(dd - 136, ss - 136);
case 8: *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 137: memcpy_avx_128(dd - 137, ss - 137);
case 9: *((uint64_t*)(dd - 9)) = *((uint64_t*)(ss - 9)); dd[-1] = ss[-1]; break;
case 138: memcpy_avx_128(dd - 138, ss - 138);
case 10: *((uint64_t*)(dd - 10)) = *((uint64_t*)(ss - 10)); *((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2)); break;
case 139: memcpy_avx_128(dd - 139, ss - 139);
case 11: *((uint64_t*)(dd - 11)) = *((uint64_t*)(ss - 11)); *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 140: memcpy_avx_128(dd - 140, ss - 140);
case 12: *((uint64_t*)(dd - 12)) = *((uint64_t*)(ss - 12)); *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 141: memcpy_avx_128(dd - 141, ss - 141);
case 13: *((uint64_t*)(dd - 13)) = *((uint64_t*)(ss - 13)); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 142: memcpy_avx_128(dd - 142, ss - 142);
case 14: *((uint64_t*)(dd - 14)) = *((uint64_t*)(ss - 14)); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 143: memcpy_avx_128(dd - 143, ss - 143);
case 15: *((uint64_t*)(dd - 15)) = *((uint64_t*)(ss - 15)); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 144: memcpy_avx_128(dd - 144, ss - 144);
case 16: memcpy_avx_16(dd - 16, ss - 16); break;
case 145: memcpy_avx_128(dd - 145, ss - 145);
case 17: memcpy_avx_16(dd - 17, ss - 17); dd[-1] = ss[-1]; break;
case 146: memcpy_avx_128(dd - 146, ss - 146);
case 18: memcpy_avx_16(dd - 18, ss - 18); *((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2)); break;
case 147: memcpy_avx_128(dd - 147, ss - 147);
case 19: memcpy_avx_16(dd - 19, ss - 19); *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 148: memcpy_avx_128(dd - 148, ss - 148);
case 20: memcpy_avx_16(dd - 20, ss - 20); *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 149: memcpy_avx_128(dd - 149, ss - 149);
case 21: memcpy_avx_16(dd - 21, ss - 21); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 150: memcpy_avx_128(dd - 150, ss - 150);
case 22: memcpy_avx_16(dd - 22, ss - 22); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 151: memcpy_avx_128(dd - 151, ss - 151);
case 23: memcpy_avx_16(dd - 23, ss - 23); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 152: memcpy_avx_128(dd - 152, ss - 152);
case 24: memcpy_avx_16(dd - 24, ss - 24); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 153: memcpy_avx_128(dd - 153, ss - 153);
case 25: memcpy_avx_16(dd - 25, ss - 25); memcpy_avx_16(dd - 16, ss - 16); break;
case 154: memcpy_avx_128(dd - 154, ss - 154);
case 26: memcpy_avx_16(dd - 26, ss - 26); memcpy_avx_16(dd - 16, ss - 16); break;
case 155: memcpy_avx_128(dd - 155, ss - 155);
case 27: memcpy_avx_16(dd - 27, ss - 27); memcpy_avx_16(dd - 16, ss - 16); break;
case 156: memcpy_avx_128(dd - 156, ss - 156);
case 28: memcpy_avx_16(dd - 28, ss - 28); memcpy_avx_16(dd - 16, ss - 16); break;
case 157: memcpy_avx_128(dd - 157, ss - 157);
case 29: memcpy_avx_16(dd - 29, ss - 29); memcpy_avx_16(dd - 16, ss - 16); break;
case 158: memcpy_avx_128(dd - 158, ss - 158);
case 30: memcpy_avx_16(dd - 30, ss - 30); memcpy_avx_16(dd - 16, ss - 16); break;
case 159: memcpy_avx_128(dd - 159, ss - 159);
case 31: memcpy_avx_16(dd - 31, ss - 31); memcpy_avx_16(dd - 16, ss - 16); break;
case 160: memcpy_avx_128(dd - 160, ss - 160);
case 32: memcpy_avx_32(dd - 32, ss - 32); break;
case 161: memcpy_avx_128(dd - 161, ss - 161);
case 33: memcpy_avx_32(dd - 33, ss - 33); dd[-1] = ss[-1]; break;
case 162: memcpy_avx_128(dd - 162, ss - 162);
case 34: memcpy_avx_32(dd - 34, ss - 34); *((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2)); break;
case 163: memcpy_avx_128(dd - 163, ss - 163);
case 35: memcpy_avx_32(dd - 35, ss - 35); *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 164: memcpy_avx_128(dd - 164, ss - 164);
case 36: memcpy_avx_32(dd - 36, ss - 36); *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 165: memcpy_avx_128(dd - 165, ss - 165);
case 37: memcpy_avx_32(dd - 37, ss - 37); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 166: memcpy_avx_128(dd - 166, ss - 166);
case 38: memcpy_avx_32(dd - 38, ss - 38); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 167: memcpy_avx_128(dd - 167, ss - 167);
case 39: memcpy_avx_32(dd - 39, ss - 39); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 168: memcpy_avx_128(dd - 168, ss - 168);
case 40: memcpy_avx_32(dd - 40, ss - 40); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 169: memcpy_avx_128(dd - 169, ss - 169);
case 41: memcpy_avx_32(dd - 41, ss - 41); memcpy_avx_16(dd - 16, ss - 16); break;
case 170: memcpy_avx_128(dd - 170, ss - 170);
case 42: memcpy_avx_32(dd - 42, ss - 42); memcpy_avx_16(dd - 16, ss - 16); break;
case 171: memcpy_avx_128(dd - 171, ss - 171);
case 43: memcpy_avx_32(dd - 43, ss - 43); memcpy_avx_16(dd - 16, ss - 16); break;
case 172: memcpy_avx_128(dd - 172, ss - 172);
case 44: memcpy_avx_32(dd - 44, ss - 44); memcpy_avx_16(dd - 16, ss - 16); break;
case 173: memcpy_avx_128(dd - 173, ss - 173);
case 45: memcpy_avx_32(dd - 45, ss - 45); memcpy_avx_16(dd - 16, ss - 16); break;
case 174: memcpy_avx_128(dd - 174, ss - 174);
case 46: memcpy_avx_32(dd - 46, ss - 46); memcpy_avx_16(dd - 16, ss - 16); break;
case 175: memcpy_avx_128(dd - 175, ss - 175);
case 47: memcpy_avx_32(dd - 47, ss - 47); memcpy_avx_16(dd - 16, ss - 16); break;
case 176: memcpy_avx_128(dd - 176, ss - 176);
case 48: memcpy_avx_32(dd - 48, ss - 48); memcpy_avx_16(dd - 16, ss - 16); break;
case 177: memcpy_avx_128(dd - 177, ss - 177);
case 49: memcpy_avx_32(dd - 49, ss - 49); memcpy_avx_32(dd - 32, ss - 32); break;
case 178: memcpy_avx_128(dd - 178, ss - 178);
case 50: memcpy_avx_32(dd - 50, ss - 50); memcpy_avx_32(dd - 32, ss - 32); break;
case 179: memcpy_avx_128(dd - 179, ss - 179);
case 51: memcpy_avx_32(dd - 51, ss - 51); memcpy_avx_32(dd - 32, ss - 32); break;
case 180: memcpy_avx_128(dd - 180, ss - 180);
case 52: memcpy_avx_32(dd - 52, ss - 52); memcpy_avx_32(dd - 32, ss - 32); break;
case 181: memcpy_avx_128(dd - 181, ss - 181);
case 53: memcpy_avx_32(dd - 53, ss - 53); memcpy_avx_32(dd - 32, ss - 32); break;
case 182: memcpy_avx_128(dd - 182, ss - 182);
case 54: memcpy_avx_32(dd - 54, ss - 54); memcpy_avx_32(dd - 32, ss - 32); break;
case 183: memcpy_avx_128(dd - 183, ss - 183);
case 55: memcpy_avx_32(dd - 55, ss - 55); memcpy_avx_32(dd - 32, ss - 32); break;
case 184: memcpy_avx_128(dd - 184, ss - 184);
case 56: memcpy_avx_32(dd - 56, ss - 56); memcpy_avx_32(dd - 32, ss - 32); break;
case 185: memcpy_avx_128(dd - 185, ss - 185);
case 57: memcpy_avx_32(dd - 57, ss - 57); memcpy_avx_32(dd - 32, ss - 32); break;
case 186: memcpy_avx_128(dd - 186, ss - 186);
case 58: memcpy_avx_32(dd - 58, ss - 58); memcpy_avx_32(dd - 32, ss - 32); break;
case 187: memcpy_avx_128(dd - 187, ss - 187);
case 59: memcpy_avx_32(dd - 59, ss - 59); memcpy_avx_32(dd - 32, ss - 32); break;
case 188: memcpy_avx_128(dd - 188, ss - 188);
case 60: memcpy_avx_32(dd - 60, ss - 60); memcpy_avx_32(dd - 32, ss - 32); break;
case 189: memcpy_avx_128(dd - 189, ss - 189);
case 61: memcpy_avx_32(dd - 61, ss - 61); memcpy_avx_32(dd - 32, ss - 32); break;
case 190: memcpy_avx_128(dd - 190, ss - 190);
case 62: memcpy_avx_32(dd - 62, ss - 62); memcpy_avx_32(dd - 32, ss - 32); break;
case 191: memcpy_avx_128(dd - 191, ss - 191);
case 63: memcpy_avx_32(dd - 63, ss - 63); memcpy_avx_32(dd - 32, ss - 32); break;
case 192: memcpy_avx_128(dd - 192, ss - 192);
case 64: memcpy_avx_64(dd - 64, ss - 64); break;
case 193: memcpy_avx_128(dd - 193, ss - 193);
case 65: memcpy_avx_64(dd - 65, ss - 65); dd[-1] = ss[-1]; break;
case 194: memcpy_avx_128(dd - 194, ss - 194);
case 66: memcpy_avx_64(dd - 66, ss - 66); *((uint16_t*)(dd - 2)) = *((uint16_t*)(ss - 2)); break;
case 195: memcpy_avx_128(dd - 195, ss - 195);
case 67: memcpy_avx_64(dd - 67, ss - 67); *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 196: memcpy_avx_128(dd - 196, ss - 196);
case 68: memcpy_avx_64(dd - 68, ss - 68); *((uint32_t*)(dd - 4)) = *((uint32_t*)(ss - 4)); break;
case 197: memcpy_avx_128(dd - 197, ss - 197);
case 69: memcpy_avx_64(dd - 69, ss - 69); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 198: memcpy_avx_128(dd - 198, ss - 198);
case 70: memcpy_avx_64(dd - 70, ss - 70); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 199: memcpy_avx_128(dd - 199, ss - 199);
case 71: memcpy_avx_64(dd - 71, ss - 71); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 200: memcpy_avx_128(dd - 200, ss - 200);
case 72: memcpy_avx_64(dd - 72, ss - 72); *((uint64_t*)(dd - 8)) = *((uint64_t*)(ss - 8)); break;
case 201: memcpy_avx_128(dd - 201, ss - 201);
case 73: memcpy_avx_64(dd - 73, ss - 73); memcpy_avx_16(dd - 16, ss - 16); break;
case 202: memcpy_avx_128(dd - 202, ss - 202);
case 74: memcpy_avx_64(dd - 74, ss - 74); memcpy_avx_16(dd - 16, ss - 16); break;
case 203: memcpy_avx_128(dd - 203, ss - 203);
case 75: memcpy_avx_64(dd - 75, ss - 75); memcpy_avx_16(dd - 16, ss - 16); break;
case 204: memcpy_avx_128(dd - 204, ss - 204);
case 76: memcpy_avx_64(dd - 76, ss - 76); memcpy_avx_16(dd - 16, ss - 16); break;
case 205: memcpy_avx_128(dd - 205, ss - 205);
case 77: memcpy_avx_64(dd - 77, ss - 77); memcpy_avx_16(dd - 16, ss - 16); break;
case 206: memcpy_avx_128(dd - 206, ss - 206);
case 78: memcpy_avx_64(dd - 78, ss - 78); memcpy_avx_16(dd - 16, ss - 16); break;
case 207: memcpy_avx_128(dd - 207, ss - 207);
case 79: memcpy_avx_64(dd - 79, ss - 79); memcpy_avx_16(dd - 16, ss - 16); break;
case 208: memcpy_avx_128(dd - 208, ss - 208);
case 80: memcpy_avx_64(dd - 80, ss - 80); memcpy_avx_16(dd - 16, ss - 16); break;
case 209: memcpy_avx_128(dd - 209, ss - 209);
case 81: memcpy_avx_64(dd - 81, ss - 81); memcpy_avx_32(dd - 32, ss - 32); break;
case 210: memcpy_avx_128(dd - 210, ss - 210);
case 82: memcpy_avx_64(dd - 82, ss - 82); memcpy_avx_32(dd - 32, ss - 32); break;
case 211: memcpy_avx_128(dd - 211, ss - 211);
case 83: memcpy_avx_64(dd - 83, ss - 83); memcpy_avx_32(dd - 32, ss - 32); break;
case 212: memcpy_avx_128(dd - 212, ss - 212);
case 84: memcpy_avx_64(dd - 84, ss - 84); memcpy_avx_32(dd - 32, ss - 32); break;
case 213: memcpy_avx_128(dd - 213, ss - 213);
case 85: memcpy_avx_64(dd - 85, ss - 85); memcpy_avx_32(dd - 32, ss - 32); break;
case 214: memcpy_avx_128(dd - 214, ss - 214);
case 86: memcpy_avx_64(dd - 86, ss - 86); memcpy_avx_32(dd - 32, ss - 32); break;
case 215: memcpy_avx_128(dd - 215, ss - 215);
case 87: memcpy_avx_64(dd - 87, ss - 87); memcpy_avx_32(dd - 32, ss - 32); break;
case 216: memcpy_avx_128(dd - 216, ss - 216);
case 88: memcpy_avx_64(dd - 88, ss - 88); memcpy_avx_32(dd - 32, ss - 32); break;
case 217: memcpy_avx_128(dd - 217, ss - 217);
case 89: memcpy_avx_64(dd - 89, ss - 89); memcpy_avx_32(dd - 32, ss - 32); break;
case 218: memcpy_avx_128(dd - 218, ss - 218);
case 90: memcpy_avx_64(dd - 90, ss - 90); memcpy_avx_32(dd - 32, ss - 32); break;
case 219: memcpy_avx_128(dd - 219, ss - 219);
case 91: memcpy_avx_64(dd - 91, ss - 91); memcpy_avx_32(dd - 32, ss - 32); break;
case 220: memcpy_avx_128(dd - 220, ss - 220);
case 92: memcpy_avx_64(dd - 92, ss - 92); memcpy_avx_32(dd - 32, ss - 32); break;
case 221: memcpy_avx_128(dd - 221, ss - 221);
case 93: memcpy_avx_64(dd - 93, ss - 93); memcpy_avx_32(dd - 32, ss - 32); break;
case 222: memcpy_avx_128(dd - 222, ss - 222);
case 94: memcpy_avx_64(dd - 94, ss - 94); memcpy_avx_32(dd - 32, ss - 32); break;
case 223: memcpy_avx_128(dd - 223, ss - 223);
case 95: memcpy_avx_64(dd - 95, ss - 95); memcpy_avx_32(dd - 32, ss - 32); break;
case 224: memcpy_avx_128(dd - 224, ss - 224);
case 96: memcpy_avx_64(dd - 96, ss - 96); memcpy_avx_32(dd - 32, ss - 32); break;
case 225: memcpy_avx_128(dd - 225, ss - 225);
case 97: memcpy_avx_64(dd - 97, ss - 97); memcpy_avx_64(dd - 64, ss - 64); break;
case 226: memcpy_avx_128(dd - 226, ss - 226);
case 98: memcpy_avx_64(dd - 98, ss - 98); memcpy_avx_64(dd - 64, ss - 64); break;
case 227: memcpy_avx_128(dd - 227, ss - 227);
case 99: memcpy_avx_64(dd - 99, ss - 99); memcpy_avx_64(dd - 64, ss - 64); break;
case 228: memcpy_avx_128(dd - 228, ss - 228);
case 100: memcpy_avx_64(dd - 100, ss - 100); memcpy_avx_64(dd - 64, ss - 64); break;
case 229: memcpy_avx_128(dd - 229, ss - 229);
case 101: memcpy_avx_64(dd - 101, ss - 101); memcpy_avx_64(dd - 64, ss - 64); break;
case 230: memcpy_avx_128(dd - 230, ss - 230);
case 102: memcpy_avx_64(dd - 102, ss - 102); memcpy_avx_64(dd - 64, ss - 64); break;
case 231: memcpy_avx_128(dd - 231, ss - 231);
case 103: memcpy_avx_64(dd - 103, ss - 103); memcpy_avx_64(dd - 64, ss - 64); break;
case 232: memcpy_avx_128(dd - 232, ss - 232);
case 104: memcpy_avx_64(dd - 104, ss - 104); memcpy_avx_64(dd - 64, ss - 64); break;
case 233: memcpy_avx_128(dd - 233, ss - 233);
case 105: memcpy_avx_64(dd - 105, ss - 105); memcpy_avx_64(dd - 64, ss - 64); break;
case 234: memcpy_avx_128(dd - 234, ss - 234);
case 106: memcpy_avx_64(dd - 106, ss - 106); memcpy_avx_64(dd - 64, ss - 64); break;
case 235: memcpy_avx_128(dd - 235, ss - 235);
case 107: memcpy_avx_64(dd - 107, ss - 107); memcpy_avx_64(dd - 64, ss - 64); break;
case 236: memcpy_avx_128(dd - 236, ss - 236);
case 108: memcpy_avx_64(dd - 108, ss - 108); memcpy_avx_64(dd - 64, ss - 64); break;
case 237: memcpy_avx_128(dd - 237, ss - 237);
case 109: memcpy_avx_64(dd - 109, ss - 109); memcpy_avx_64(dd - 64, ss - 64); break;
case 238: memcpy_avx_128(dd - 238, ss - 238);
case 110: memcpy_avx_64(dd - 110, ss - 110); memcpy_avx_64(dd - 64, ss - 64); break;
case 239: memcpy_avx_128(dd - 239, ss - 239);
case 111: memcpy_avx_64(dd - 111, ss - 111); memcpy_avx_64(dd - 64, ss - 64); break;
case 240: memcpy_avx_128(dd - 240, ss - 240);
case 112: memcpy_avx_64(dd - 112, ss - 112); memcpy_avx_64(dd - 64, ss - 64); break;
case 241: memcpy_avx_128(dd - 241, ss - 241);
case 113: memcpy_avx_64(dd - 113, ss - 113); memcpy_avx_64(dd - 64, ss - 64); break;
case 242: memcpy_avx_128(dd - 242, ss - 242);
case 114: memcpy_avx_64(dd - 114, ss - 114); memcpy_avx_64(dd - 64, ss - 64); break;
case 243: memcpy_avx_128(dd - 243, ss - 243);
case 115: memcpy_avx_64(dd - 115, ss - 115); memcpy_avx_64(dd - 64, ss - 64); break;
case 244: memcpy_avx_128(dd - 244, ss - 244);
case 116: memcpy_avx_64(dd - 116, ss - 116); memcpy_avx_64(dd - 64, ss - 64); break;
case 245: memcpy_avx_128(dd - 245, ss - 245);
case 117: memcpy_avx_64(dd - 117, ss - 117); memcpy_avx_64(dd - 64, ss - 64); break;
case 246: memcpy_avx_128(dd - 246, ss - 246);
case 118: memcpy_avx_64(dd - 118, ss - 118); memcpy_avx_64(dd - 64, ss - 64); break;
case 247: memcpy_avx_128(dd - 247, ss - 247);
case 119: memcpy_avx_64(dd - 119, ss - 119); memcpy_avx_64(dd - 64, ss - 64); break;
case 248: memcpy_avx_128(dd - 248, ss - 248);
case 120: memcpy_avx_64(dd - 120, ss - 120); memcpy_avx_64(dd - 64, ss - 64); break;
case 249: memcpy_avx_128(dd - 249, ss - 249);
case 121: memcpy_avx_64(dd - 121, ss - 121); memcpy_avx_64(dd - 64, ss - 64); break;
case 250: memcpy_avx_128(dd - 250, ss - 250);
case 122: memcpy_avx_64(dd - 122, ss - 122); memcpy_avx_64(dd - 64, ss - 64); break;
case 251: memcpy_avx_128(dd - 251, ss - 251);
case 123: memcpy_avx_64(dd - 123, ss - 123); memcpy_avx_64(dd - 64, ss - 64); break;
case 252: memcpy_avx_128(dd - 252, ss - 252);
case 124: memcpy_avx_64(dd - 124, ss - 124); memcpy_avx_64(dd - 64, ss - 64); break;
case 253: memcpy_avx_128(dd - 253, ss - 253);
case 125: memcpy_avx_64(dd - 125, ss - 125); memcpy_avx_64(dd - 64, ss - 64); break;
case 254: memcpy_avx_128(dd - 254, ss - 254);
case 126: memcpy_avx_64(dd - 126, ss - 126); memcpy_avx_64(dd - 64, ss - 64); break;
case 255: memcpy_avx_128(dd - 255, ss - 255);
case 127: memcpy_avx_64(dd - 127, ss - 127); memcpy_avx_64(dd - 64, ss - 64); break;
case 256: memcpy_avx_256(dd - 256, ss - 256); break;
}
return dst;
}
//---------------------------------------------------------------------
// main routine
//---------------------------------------------------------------------
static void* memcpy_fast(void *destination, const void *source, size_t size)
{
unsigned char *dst = (unsigned char*)destination;
const unsigned char *src = (const unsigned char*)source;
static size_t cachesize = 0x200000; // L3-cache size
size_t padding;
// small memory copy
if (size <= 256) {
memcpy_tiny(dst, src, size);
_mm256_zeroupper();
return destination;
}
// align destination to 16 bytes boundary
padding = (32 - (((size_t)dst) & 31)) & 31;
#if 0
if (padding > 0) {
__m256i head = _mm256_loadu_si256((const __m256i*)src);
_mm256_storeu_si256((__m256i*)dst, head);
dst += padding;
src += padding;
size -= padding;
}
#else
__m256i head = _mm256_loadu_si256((const __m256i*)src);
_mm256_storeu_si256((__m256i*)dst, head);
dst += padding;
src += padding;
size -= padding;
#endif
// medium size copy
if (size <= cachesize) {
__m256i c0, c1, c2, c3, c4, c5, c6, c7;
for (; size >= 256; size -= 256) {
c0 = _mm256_loadu_si256(((const __m256i*)src) + 0);
c1 = _mm256_loadu_si256(((const __m256i*)src) + 1);
c2 = _mm256_loadu_si256(((const __m256i*)src) + 2);
c3 = _mm256_loadu_si256(((const __m256i*)src) + 3);
c4 = _mm256_loadu_si256(((const __m256i*)src) + 4);
c5 = _mm256_loadu_si256(((const __m256i*)src) + 5);
c6 = _mm256_loadu_si256(((const __m256i*)src) + 6);
c7 = _mm256_loadu_si256(((const __m256i*)src) + 7);
_mm_prefetch((const char*)(src + 512), _MM_HINT_NTA);
src += 256;
_mm256_storeu_si256((((__m256i*)dst) + 0), c0);
_mm256_storeu_si256((((__m256i*)dst) + 1), c1);
_mm256_storeu_si256((((__m256i*)dst) + 2), c2);
_mm256_storeu_si256((((__m256i*)dst) + 3), c3);
_mm256_storeu_si256((((__m256i*)dst) + 4), c4);
_mm256_storeu_si256((((__m256i*)dst) + 5), c5);
_mm256_storeu_si256((((__m256i*)dst) + 6), c6);
_mm256_storeu_si256((((__m256i*)dst) + 7), c7);
dst += 256;
}
}
else { // big memory copy
__m256i c0, c1, c2, c3, c4, c5, c6, c7;
/* __m256i c0, c1, c2, c3, c4, c5, c6, c7; */
_mm_prefetch((const char*)(src), _MM_HINT_NTA);
if ((((size_t)src) & 31) == 0) { // source aligned
for (; size >= 256; size -= 256) {
c0 = _mm256_load_si256(((const __m256i*)src) + 0);
c1 = _mm256_load_si256(((const __m256i*)src) + 1);
c2 = _mm256_load_si256(((const __m256i*)src) + 2);
c3 = _mm256_load_si256(((const __m256i*)src) + 3);
c4 = _mm256_load_si256(((const __m256i*)src) + 4);
c5 = _mm256_load_si256(((const __m256i*)src) + 5);
c6 = _mm256_load_si256(((const __m256i*)src) + 6);
c7 = _mm256_load_si256(((const __m256i*)src) + 7);
_mm_prefetch((const char*)(src + 512), _MM_HINT_NTA);
src += 256;
_mm256_stream_si256((((__m256i*)dst) + 0), c0);
_mm256_stream_si256((((__m256i*)dst) + 1), c1);
_mm256_stream_si256((((__m256i*)dst) + 2), c2);
_mm256_stream_si256((((__m256i*)dst) + 3), c3);
_mm256_stream_si256((((__m256i*)dst) + 4), c4);
_mm256_stream_si256((((__m256i*)dst) + 5), c5);
_mm256_stream_si256((((__m256i*)dst) + 6), c6);
_mm256_stream_si256((((__m256i*)dst) + 7), c7);
dst += 256;
}
}
else { // source unaligned
for (; size >= 256; size -= 256) {
c0 = _mm256_loadu_si256(((const __m256i*)src) + 0);
c1 = _mm256_loadu_si256(((const __m256i*)src) + 1);
c2 = _mm256_loadu_si256(((const __m256i*)src) + 2);
c3 = _mm256_loadu_si256(((const __m256i*)src) + 3);
c4 = _mm256_loadu_si256(((const __m256i*)src) + 4);
c5 = _mm256_loadu_si256(((const __m256i*)src) + 5);
c6 = _mm256_loadu_si256(((const __m256i*)src) + 6);
c7 = _mm256_loadu_si256(((const __m256i*)src) + 7);
_mm_prefetch((const char*)(src + 512), _MM_HINT_NTA);
src += 256;
_mm256_stream_si256((((__m256i*)dst) + 0), c0);
_mm256_stream_si256((((__m256i*)dst) + 1), c1);
_mm256_stream_si256((((__m256i*)dst) + 2), c2);
_mm256_stream_si256((((__m256i*)dst) + 3), c3);
_mm256_stream_si256((((__m256i*)dst) + 4), c4);
_mm256_stream_si256((((__m256i*)dst) + 5), c5);
_mm256_stream_si256((((__m256i*)dst) + 6), c6);
_mm256_stream_si256((((__m256i*)dst) + 7), c7);
dst += 256;
}
}
_mm_sfence();
}
memcpy_tiny(dst, src, size);
_mm256_zeroupper();
return destination;
}
#endif
MomentaMedia/include/Utils/Settings.h
...@@ -18,8 +18,8 @@ public: ...@@ -18,8 +18,8 @@ public:
static int32_t RecordStoreDuration ; static int32_t RecordStoreDuration ;
static int32_t RecordFlag; static int32_t RecordFlag;
static int32_t OpenOMP ; static int32_t OpenOMP ;
static int32_t TimeoutFrames ; static int32_t TimeoutFrames ; //mask 贴图的超时时间 针对如果未收到算法给的结束帧情况使用
static int32_t SdiOutWaitNums; static int32_t SecondSdiOutWaitNums; //sdi输出的二次确认功能保留贴图开始的前多少帧
static bool HaveBlackDataFlag; static bool HaveBlackDataFlag;
static int32_t DrawFlag; static int32_t DrawFlag;
...@@ -40,7 +40,7 @@ public: ...@@ -40,7 +40,7 @@ public:
static uint32_t ReplayForward; static uint32_t ReplayForward;
static uint32_t ReplayDeley; static uint32_t ReplayDeley;
static uint32_t SDIOutputWaitNums; static uint32_t RTSDIOutputWaitNums; //实时输出的缓存帧大小 一般5帧
static int32_t CropFlag; static int32_t CropFlag;
static int32_t CropX; static int32_t CropX;
static int32_t CropY; static int32_t CropY;
......
MomentaMedia/src/BlackMagicDesign/DeckLinkInputPage.cpp
...@@ -423,7 +423,7 @@ void DeckLinkInputPage::ObjectNameChanged(const QString& newName) ...@@ -423,7 +423,7 @@ void DeckLinkInputPage::ObjectNameChanged(const QString& newName)
//connect(Capture.get(), SIGNAL(PushFrame(std::shared_ptr<videoFrameData>)), NDIOutput.get(), SLOT(AddFrame(std::shared_ptr<videoFrameData>))); //connect(Capture.get(), SIGNAL(PushFrame(std::shared_ptr<videoFrameData>)), NDIOutput.get(), SLOT(AddFrame(std::shared_ptr<videoFrameData>)));
connect(Replay.get(), SIGNAL(PushFrame(std::shared_ptr<videoFrameData>)), NDIOutput.get(), SLOT(AddFrame(std::shared_ptr<videoFrameData>)), Qt::DirectConnection); connect(Replay.get(), SIGNAL(PushFrame(std::shared_ptr<videoFrameData>)), NDIOutput.get(), SLOT(AddFrame(std::shared_ptr<videoFrameData>)), Qt::DirectConnection);
connect(Capture.get(), SIGNAL(PushFrameToReplay(std::shared_ptr<videoFrameData>)), Replay.get(), SLOT(addFrame(std::shared_ptr<videoFrameData>))/*, Qt::DirectConnection*/); connect(Capture.get(), SIGNAL(PushFrameToReplay(std::shared_ptr<videoFrameData>)), Replay.get(), SLOT(addFrame(std::shared_ptr<videoFrameData>)), Qt::DirectConnection);
connect(udpServer.get(), SIGNAL(SendMsg(QByteArray)),this, SLOT(RecvMsg(QByteArray))); connect(udpServer.get(), SIGNAL(SendMsg(QByteArray)),this, SLOT(RecvMsg(QByteArray)));
//NDIOutput->moveToThread(Capture.get()); //NDIOutput->moveToThread(Capture.get());
//connect(SelectedDevice.Get(), SIGNAL(ArrivedFrame(ComPtr<IDeckLinkVideoInputFrame>)), NDIOutput.get(), SLOT(AddFrame(ComPtr<IDeckLinkVideoInputFrame>))); //connect(SelectedDevice.Get(), SIGNAL(ArrivedFrame(ComPtr<IDeckLinkVideoInputFrame>)), NDIOutput.get(), SLOT(AddFrame(ComPtr<IDeckLinkVideoInputFrame>)));
......
MomentaMedia/src/BlackMagicDesign/DeckLinkOutputDevice.cpp
...@@ -245,13 +245,13 @@ bool DeckLinkOutputDevice::StartPlayback(BMDDisplayMode displayMode, bool enable ...@@ -245,13 +245,13 @@ bool DeckLinkOutputDevice::StartPlayback(BMDDisplayMode displayMode, bool enable
/*scheduleVideoFramesThread = std::thread(&DeckLinkOutputDevice::scheduleVideoFramesFunc, this); /*scheduleVideoFramesThread = std::thread(&DeckLinkOutputDevice::scheduleVideoFramesFunc, this);
scheduleAudioFramesThread = std::thread(&DeckLinkOutputDevice::scheduleAudioFramesFuncDeley, this);*/ scheduleAudioFramesThread = std::thread(&DeckLinkOutputDevice::scheduleAudioFramesFuncDeley, this);*/
if (Settings::SdiOutWaitNums) if (Settings::SecondSdiOutWaitNums)
{ {
scheduleVideoFramesThread = std::thread(&DeckLinkOutputDevice::scheduleVideoFramesWaitFunc, this); scheduleVideoFramesThread = std::thread(&DeckLinkOutputDevice::scheduleVideoFramesWaitFunc, this);
} }
else else
{ {
if(Settings::SDIOutputWaitNums) scheduleVideoFramesThread = std::thread(&DeckLinkOutputDevice::scheduleVideoFramesRealTimeFunc, this); if(Settings::RTSDIOutputWaitNums) scheduleVideoFramesThread = std::thread(&DeckLinkOutputDevice::scheduleVideoFramesRealTimeFunc, this);
else scheduleVideoFramesThread = std::thread(&DeckLinkOutputDevice::scheduleVideoFramesNoDeleyFunc, this); else scheduleVideoFramesThread = std::thread(&DeckLinkOutputDevice::scheduleVideoFramesNoDeleyFunc, this);
} }
...@@ -503,10 +503,10 @@ bool DeckLinkOutputDevice::getReferenceSignalMode(BMDDisplayMode* mode) ...@@ -503,10 +503,10 @@ bool DeckLinkOutputDevice::getReferenceSignalMode(BMDDisplayMode* mode)
while (true) while (true)
{ {
if (state == PlaybackState::Stopping) break; if (state == PlaybackState::Stopping) break;
std::shared_ptr<VideoFrameWithMask> outputFrame; std::shared_ptr<VideoFrameWithMask> outputFrame = nullptr;
std::vector<std::shared_ptr<VideoFrameWithMask>> frames; std::vector<std::shared_ptr<VideoFrameWithMask>> frames;
if (outputMaskVideoFrameQueue.WaitFor(Settings::SdiOutWaitNums)) if (outputMaskVideoFrameQueue.WaitFor(Settings::SecondSdiOutWaitNums))
{ {
outputMaskVideoFrameQueue.Pop(outputFrame); outputMaskVideoFrameQueue.Pop(outputFrame);
outputMaskVideoFrameDeque.Pop(); outputMaskVideoFrameDeque.Pop();
...@@ -839,7 +839,7 @@ bool DeckLinkOutputDevice::getReferenceSignalMode(BMDDisplayMode* mode) ...@@ -839,7 +839,7 @@ bool DeckLinkOutputDevice::getReferenceSignalMode(BMDDisplayMode* mode)
//bool send_zoom_frame_flag = false; //bool send_zoom_frame_flag = false;
std::shared_ptr<VideoFrameWithMask> wait_zoom_frame = nullptr; std::shared_ptr<VideoFrameWithMask> wait_zoom_frame = nullptr;
HRESULT ret = S_OK; HRESULT ret = S_OK;
int wait_size = Settings::SDIOutputWaitNums; int wait_size = Settings::RTSDIOutputWaitNums;
while (true) while (true)
{ {
...@@ -1544,7 +1544,7 @@ void DeckLinkOutputDevice::AddZoomFrame(std::shared_ptr<VideoFrameWithMask> fram ...@@ -1544,7 +1544,7 @@ void DeckLinkOutputDevice::AddZoomFrame(std::shared_ptr<VideoFrameWithMask> fram
{ {
if (frame && (frame->data_ || (frame->pImage && (frame->pImage->data || frame->pImage->uyvy_data)))) if (frame && (frame->data_ || (frame->pImage && (frame->pImage->data || frame->pImage->uyvy_data))))
{ {
if(Settings::SdiOutWaitNums) if(Settings::SecondSdiOutWaitNums)
{ {
auto start_time = frame->start_tm_; auto start_time = frame->start_tm_;
//std::lock_guard<std::mutex> locker(sdi_clear_mutex); //std::lock_guard<std::mutex> locker(sdi_clear_mutex);
...@@ -1572,7 +1572,7 @@ void DeckLinkOutputDevice::AddZoomFrame(std::shared_ptr<VideoFrameWithMask> fram ...@@ -1572,7 +1572,7 @@ void DeckLinkOutputDevice::AddZoomFrame(std::shared_ptr<VideoFrameWithMask> fram
void DeckLinkOutputDevice::AddVideoFrameMask(std::shared_ptr<VideoFrameWithMask> frame) void DeckLinkOutputDevice::AddVideoFrameMask(std::shared_ptr<VideoFrameWithMask> frame)
{ {
//auto t1 = TimeMilliSecond(); //auto t1 = TimeMilliSecond();
if (outputMaskVideoFrameQueue.Size() > (50 * 3 + Settings::SdiOutWaitNums) || output_video_frame_map.Size() > (50 * 3 + Settings::SdiOutWaitNums)) if (outputMaskVideoFrameQueue.Size() > (50 * 3 + Settings::SecondSdiOutWaitNums) || output_video_frame_map.Size() > (50 * 3 + Settings::SecondSdiOutWaitNums))
{ {
if(outputMaskVideoFrameQueue.Size()) outputMaskVideoFrameQueue.Reset(); if(outputMaskVideoFrameQueue.Size()) outputMaskVideoFrameQueue.Reset();
if (output_video_frame_map.Size()) output_video_frame_map.Clear(); if (output_video_frame_map.Size()) output_video_frame_map.Clear();
...@@ -1586,78 +1586,125 @@ void DeckLinkOutputDevice::AddVideoFrameMask(std::shared_ptr<VideoFrameWithMask> ...@@ -1586,78 +1586,125 @@ void DeckLinkOutputDevice::AddVideoFrameMask(std::shared_ptr<VideoFrameWithMask>
//qDebug() << "AddVideoFrameMask:" << "timestamp:" << frame->sequenceNum_ << GetCurrDateTimeStr() << "\n"; //qDebug() << "AddVideoFrameMask:" << "timestamp:" << frame->sequenceNum_ << GetCurrDateTimeStr() << "\n";
} }
if (Settings::SDIOutputWaitNums)
if (frame->fmt_ == bmdFormat8BitYUV)
{ {
if (frame->fmt_ == bmdFormat8BitYUV) if (Settings::SecondSdiOutWaitNums)
{ {
output_video_frame_map.Insert(frame->timestamp_, frame); outputMaskVideoFrameQueue.Push(frame);
outputMaskVideoFrameDeque.PushBack(frame);
} }
else if (frame->fmt_ == bmdFormat8BitBGRA) else
{ {
if (p_dispatch_queue) if(Settings::RTSDIOutputWaitNums) output_video_frame_map.Insert(frame->timestamp_, frame);
{ else outputMaskVideoFrameQueue.Push(frame);
DispatchQueue* dispath = (DispatchQueue*)p_dispatch_queue;
dispath->dispatch(std::bind(&DeckLinkOutputDevice::BGRAToUYVY, this, std::placeholders::_1), frame);
}
} }
} }
else else if (frame->fmt_ == bmdFormat8BitBGRA)
{ {
if (frame->fmt_ == bmdFormat8BitYUV) if (Settings::SecondSdiOutWaitNums)
{ {
outputMaskVideoFrameQueue.Push(frame); if (frame->flag_ == BS_START)
if (Settings::SdiOutWaitNums) outputMaskVideoFrameDeque.PushBack(frame);
}
else if (frame->fmt_ == bmdFormat8BitBGRA)
{
if (Settings::SdiOutWaitNums)
{ {
if (frame->flag_ == BS_START) if (masked_map.find(frame->timestamp_) != masked_map.end())return;
else
{ {
if (masked_map.find(frame->timestamp_) != masked_map.end())return; auto tmp_map = new std::map<qint64, std::shared_ptr<VideoFrameWithMask>>();
else int size = outputMaskVideoFrameDeque.Size();
int nums = (size > Settings::SecondSdiOutWaitNums ? Settings::SecondSdiOutWaitNums : size);
std::vector<std::shared_ptr<VideoFrameWithMask>> tmp_vec;
outputMaskVideoFrameDeque.Put(nums, tmp_vec);
for (auto data : tmp_vec)
{ {
auto tmp_map = new std::map<qint64, std::shared_ptr<VideoFrameWithMask>>();
/*auto tmp_tm = frame->timestamp_ - SdiOutWaitNums;
tmp_tm = (tmp_tm < 0 ? 0 : tmp_tm);*/
int size = outputMaskVideoFrameDeque.Size();
int nums = (size > Settings::SdiOutWaitNums ? Settings::SdiOutWaitNums : size);
std::vector<std::shared_ptr<VideoFrameWithMask>> tmp_vec;
outputMaskVideoFrameDeque.Put(nums, tmp_vec);
for (auto data : tmp_vec)
{
if (data) if (data)
{ {
data->start_tm_ = frame->timestamp_; data->start_tm_ = frame->timestamp_;
tmp_map->insert({ data->timestamp_,data }); tmp_map->insert({ data->timestamp_,data });
}
} }
masked_map.insert({ frame->timestamp_,tmp_map });
} }
masked_map.insert({ frame->timestamp_,tmp_map });
} }
} }
/*while (masked_map.size() > 2)
{
auto itor = masked_map.begin();
if (itor->second)
{
itor->second->clear();
delete itor->second;
}
masked_map.erase(itor->first);
}*/
if (p_dispatch_queue)
{
DispatchQueue* dispath = (DispatchQueue*)p_dispatch_queue;
dispath->dispatch(std::bind(&DeckLinkOutputDevice::BGRAToUYVY, this, std::placeholders::_1), frame);
}
} }
if (p_dispatch_queue)
{
DispatchQueue* dispath = (DispatchQueue*)p_dispatch_queue;
dispath->dispatch(std::bind(&DeckLinkOutputDevice::BGRAToUYVY, this, std::placeholders::_1), frame);
}
} }
//if (Settings::SDIOutputWaitNums)
//{
// if (frame->fmt_ == bmdFormat8BitYUV)
// {
// output_video_frame_map.Insert(frame->timestamp_, frame);
// }
// else if (frame->fmt_ == bmdFormat8BitBGRA)
// {
// if (p_dispatch_queue)
// {
// DispatchQueue* dispath = (DispatchQueue*)p_dispatch_queue;
// dispath->dispatch(std::bind(&DeckLinkOutputDevice::BGRAToUYVY, this, std::placeholders::_1), frame);
// }
// }
//}
//else
//{
// if (frame->fmt_ == bmdFormat8BitYUV)
// {
// outputMaskVideoFrameQueue.Push(frame);
// if (Settings::SdiOutWaitNums) outputMaskVideoFrameDeque.PushBack(frame);
// }
// else if (frame->fmt_ == bmdFormat8BitBGRA)
// {
// if (Settings::SdiOutWaitNums)
// {
// if (frame->flag_ == BS_START)
// {
// if (masked_map.find(frame->timestamp_) != masked_map.end())return;
// else
// {
// auto tmp_map = new std::map<qint64, std::shared_ptr<VideoFrameWithMask>>();
// /*auto tmp_tm = frame->timestamp_ - SdiOutWaitNums;
// tmp_tm = (tmp_tm < 0 ? 0 : tmp_tm);*/
// int size = outputMaskVideoFrameDeque.Size();
// int nums = (size > Settings::SdiOutWaitNums ? Settings::SdiOutWaitNums : size);
// std::vector<std::shared_ptr<VideoFrameWithMask>> tmp_vec;
// outputMaskVideoFrameDeque.Put(nums, tmp_vec);
// for (auto data : tmp_vec)
// {
// if (data)
// {
// data->start_tm_ = frame->timestamp_;
// tmp_map->insert({ data->timestamp_,data });
// }
// }
// masked_map.insert({ frame->timestamp_,tmp_map });
// }
// }
// }
// /*while (masked_map.size() > 2)
// {
// auto itor = masked_map.begin();
// if (itor->second)
// {
// itor->second->clear();
// delete itor->second;
// }
// masked_map.erase(itor->first);
// }*/
// if (p_dispatch_queue)
// {
// DispatchQueue* dispath = (DispatchQueue*)p_dispatch_queue;
// dispath->dispatch(std::bind(&DeckLinkOutputDevice::BGRAToUYVY, this, std::placeholders::_1), frame);
// }
// }
//}
} }
#else #else
if (frame && (frame->data_ || (frame->pImage && frame->pImage->data))) if (frame && (frame->data_ || (frame->pImage && frame->pImage->data)))
...@@ -1774,7 +1821,7 @@ void DeckLinkOutputDevice::BGRAToUYVY(const std::shared_ptr<VideoFrameWithMask>& ...@@ -1774,7 +1821,7 @@ void DeckLinkOutputDevice::BGRAToUYVY(const std::shared_ptr<VideoFrameWithMask>&
std::make_shared<VideoFrameWithMask>(src_w,src_h,tm,image->sequenceNum_,uyvy,bmdFormat8BitYUV,image->flag_,image->meta_, start_time); std::make_shared<VideoFrameWithMask>(src_w,src_h,tm,image->sequenceNum_,uyvy,bmdFormat8BitYUV,image->flag_,image->meta_, start_time);
//sort_map_.insert({ image->timestamp_ ,video_frame }); //sort_map_.insert({ image->timestamp_ ,video_frame });
if(Settings::SdiOutWaitNums) if(Settings::SecondSdiOutWaitNums)
{ {
std::lock_guard<std::mutex> locker(sdi_clear_mutex); std::lock_guard<std::mutex> locker(sdi_clear_mutex);
auto itor = masked_map.find(start_time); auto itor = masked_map.find(start_time);
...@@ -1786,7 +1833,7 @@ void DeckLinkOutputDevice::BGRAToUYVY(const std::shared_ptr<VideoFrameWithMask>& ...@@ -1786,7 +1833,7 @@ void DeckLinkOutputDevice::BGRAToUYVY(const std::shared_ptr<VideoFrameWithMask>&
} }
else else
{ {
if (Settings::SDIOutputWaitNums) if (Settings::RTSDIOutputWaitNums)
{ {
output_video_frame_map.Insert(video_frame->timestamp_, video_frame); output_video_frame_map.Insert(video_frame->timestamp_, video_frame);
} }
......
MomentaMedia/src/NDI/NDIOutputThread.cpp
...@@ -233,7 +233,7 @@ void NDIOutputThread::run() ...@@ -233,7 +233,7 @@ void NDIOutputThread::run()
} }
else if (timePlusFlag) else if (timePlusFlag)
{ {
if (VideoMaskQueue.WaitFor(Settings::SDIOutputWaitNums)) if (VideoMaskQueue.WaitFor(Settings::RTSDIOutputWaitNums))
{ {
//if (!frame_mask ||(frame_mask->mask_flag && !frame_mask->data_)) continue; //if (!frame_mask ||(frame_mask->mask_flag && !frame_mask->data_)) continue;
VideoMaskQueue.Pop(frame_mask); VideoMaskQueue.Pop(frame_mask);
......
MomentaMedia/src/Threads/ProcessMaskThread.cpp
...@@ -117,7 +117,8 @@ void ProcessMaskThread::process() ...@@ -117,7 +117,8 @@ void ProcessMaskThread::process()
std::shared_ptr<videoFrameData> image = nullptr; std::shared_ptr<videoFrameData> image = nullptr;
if (taskImageQueue.WaitFor(min_size - 1) /*&& taskImageQueue.Size() >= min_size*/)//有横屏数据了 if (taskImageQueue.WaitFor(min_size - 1) /*&& taskImageQueue.Size() >= min_size*/)//有横屏数据了
{ {
if (taskImageQueue.Front(image)) /*auto ret = taskImageQueue.Front(image);
if (ret)
{ {
taskImageQueue.Pop(); taskImageQueue.Pop();
if (!image || !image->data) if (!image || !image->data)
...@@ -125,6 +126,10 @@ void ProcessMaskThread::process() ...@@ -125,6 +126,10 @@ void ProcessMaskThread::process()
continue; continue;
} }
} }
else continue;*/
taskImageQueue.Pop(image);
if (!image || !image->data) continue;
timestamp = (image->replaySeq ? image->replaySeq : image->timestamp); timestamp = (image->replaySeq ? image->replaySeq : image->timestamp);
//auto sequence = image->sequenceNum; //auto sequence = image->sequenceNum;
//seqnum = sequence; //seqnum = sequence;
...@@ -446,7 +451,7 @@ void ProcessMaskThread::workMaskModify(const std::shared_ptr<videoFrameData>& pI ...@@ -446,7 +451,7 @@ void ProcessMaskThread::workMaskModify(const std::shared_ptr<videoFrameData>& pI
{ {
status = BS_START; status = BS_START;
start_time = timestamp; start_time = timestamp;
QString msg = "{\"type\":\"MaskMsg\",\"data\":{\"status\":" + QString::number(status) + ",\"timestamp\":" + QString::number(timestamp) + "}}"; QString msg = "{\"signal\":\"MaskMsg\",\"data\":{\"status\":" + QString::number(status) + ",\"timestamp\":" + QString::number(timestamp) + "}}";
UdpSend::GetInstance().SendUdpMsg(msg, Settings::UIIpAddr, QString::number(Settings::UIUdpPort)); UdpSend::GetInstance().SendUdpMsg(msg, Settings::UIIpAddr, QString::number(Settings::UIUdpPort));
//qDebug() << "start copy crop pic.......\n"; //qDebug() << "start copy crop pic.......\n";
} }
...@@ -457,7 +462,7 @@ void ProcessMaskThread::workMaskModify(const std::shared_ptr<videoFrameData>& pI ...@@ -457,7 +462,7 @@ void ProcessMaskThread::workMaskModify(const std::shared_ptr<videoFrameData>& pI
if (status == BS_END) if (status == BS_END)
{ {
//send msg to ui with udp-protocol //send msg to ui with udp-protocol
QString msg = "{\"type\":\"MaskMsg\",\"data\":{\"status\":" + QString::number(status) + ",\"timestamp\":" + QString::number(timestamp) + "}}"; QString msg = "{\"signal\":\"MaskMsg\",\"data\":{\"status\":" + QString::number(status) + ",\"timestamp\":" + QString::number(timestamp) + "}}";
UdpSend::GetInstance().SendUdpMsg(msg, Settings::UIIpAddr, QString::number(Settings::UIUdpPort)); UdpSend::GetInstance().SendUdpMsg(msg, Settings::UIIpAddr, QString::number(Settings::UIUdpPort));
} }
} }
......
MomentaMedia/src/Threads/ReplayThread.cpp
...@@ -114,7 +114,7 @@ void ReplayThread::run() ...@@ -114,7 +114,7 @@ void ReplayThread::run()
qint32 errNo = (itor_end == storeVideoMap.end() ? 2 : (itor_begin == storeVideoMap.end() ? 1 : 3)); qint32 errNo = (itor_end == storeVideoMap.end() ? 2 : (itor_begin == storeVideoMap.end() ? 1 : 3));
QString msg = "{\"type\":\"ReplayResp\",\"data\":{\"inTime\":" + QString::number(replay_params.start_time) +",\"outTime\":" + QString::number(replay_params.end_time) + QString msg = "{\"signal\":\"ReplayResp\",\"data\":{\"inTime\":" + QString::number(replay_params.start_time) +",\"outTime\":" + QString::number(replay_params.end_time) +
",\"status\":0,\"errNo\":" + QString::number(errNo) +"}}"; ",\"status\":0,\"errNo\":" + QString::number(errNo) +"}}";
UdpSend::GetInstance().SendUdpMsg(msg, Settings::UIIpAddr, QString::number(Settings::UIUdpPort)); UdpSend::GetInstance().SendUdpMsg(msg, Settings::UIIpAddr, QString::number(Settings::UIUdpPort));
qDebug() << "replay fail,errno:"<<errNo << "\n"; qDebug() << "replay fail,errno:"<<errNo << "\n";
...@@ -165,7 +165,7 @@ void ReplayThread::run() ...@@ -165,7 +165,7 @@ void ReplayThread::run()
std::shared_ptr<videoFrameData> frame = nullptr; std::shared_ptr<videoFrameData> frame = nullptr;
frame = (resend_frame ? (resend_start ? resend_frame : replayVideoVec[replay_position] ) : replayVideoVec[replay_position]); frame = (resend_frame ? (resend_start ? resend_frame : replayVideoVec[replay_position] ) : replayVideoVec[replay_position]);
frame->replaySeq = current_seq; frame->replaySeq = current_seq;
emit PushFrame(frame); if(frame) emit PushFrame(frame);
if ((!resend_start || !resend_frame) && replay_position < (replayVideoVec.size() - 1)) replay_position++; if ((!resend_start || !resend_frame) && replay_position < (replayVideoVec.size() - 1)) replay_position++;
if (resend_frame) if (resend_frame)
......
MomentaMedia/src/TimePlus.cpp
...@@ -21,7 +21,7 @@ std::map<qint32, qint32> map_scale_mode; ...@@ -21,7 +21,7 @@ std::map<qint32, qint32> map_scale_mode;
int main_ver = 2; int main_ver = 2;
int mid_ver = 1; int mid_ver = 1;
int small_ver = 0; int small_ver = 1;
//int ReplayStoreTime = 10000;//单位ms //int ReplayStoreTime = 10000;//单位ms
//int FrontDeleyTime = 3000;//单位ms //int FrontDeleyTime = 3000;//单位ms
...@@ -778,12 +778,12 @@ void MomentaMedia::ReadSettings() ...@@ -778,12 +778,12 @@ void MomentaMedia::ReadSettings()
Settings::TimeoutFrames = settings.value("TIMEOUTFRAMES").toInt(); Settings::TimeoutFrames = settings.value("TIMEOUTFRAMES").toInt();
if (Settings::TimeoutFrames <= 0) Settings::TimeoutFrames = 50; if (Settings::TimeoutFrames <= 0) Settings::TimeoutFrames = 50;
Settings::SdiOutWaitNums = settings.value("DELEYSDINUMS",100).toInt(); Settings::SecondSdiOutWaitNums = settings.value("SECONDDELEYSDINUMS",100).toInt();
Settings::ReplayForward = settings.value("REPLAY_START_FORWARD", 5).toInt(); Settings::ReplayForward = settings.value("REPLAY_START_FORWARD", 5).toInt();
Settings::ReplayDeley = settings.value("REPLAY_END_DELEY", 5).toInt(); Settings::ReplayDeley = settings.value("REPLAY_END_DELEY", 5).toInt();
Settings::SDIOutputWaitNums = settings.value("SDIOUTPUTWAITNUMS", 5).toInt(); Settings::RTSDIOutputWaitNums = settings.value("RTSDIOUTPUTWAITNUMS", 5).toInt();
Settings::CropFlag = settings.value("CROPRECORD", 0).toInt(); Settings::CropFlag = settings.value("CROPRECORD", 0).toInt();
Settings::CropX = settings.value("CROP_X", 0).toInt(); Settings::CropX = settings.value("CROP_X", 0).toInt();
......
MomentaMedia/src/Utils/Settings.cpp
...@@ -15,7 +15,7 @@ int32_t Settings::RecordStoreDuration = 10000; ...@@ -15,7 +15,7 @@ int32_t Settings::RecordStoreDuration = 10000;
int32_t Settings::RecordFlag = 0; int32_t Settings::RecordFlag = 0;
int32_t Settings::OpenOMP = 1; int32_t Settings::OpenOMP = 1;
int32_t Settings::TimeoutFrames = 50; int32_t Settings::TimeoutFrames = 50;
int32_t Settings::SdiOutWaitNums = 100; int32_t Settings::SecondSdiOutWaitNums = 100;
bool Settings::HaveBlackDataFlag = false; bool Settings::HaveBlackDataFlag = false;
int32_t Settings::DrawFlag = 0; int32_t Settings::DrawFlag = 0;
...@@ -37,7 +37,7 @@ uint32_t Settings::ZoomOutWaitCnt = 0; ...@@ -37,7 +37,7 @@ uint32_t Settings::ZoomOutWaitCnt = 0;
uint32_t Settings::ReplayForward = 0; uint32_t Settings::ReplayForward = 0;
uint32_t Settings::ReplayDeley = 0; uint32_t Settings::ReplayDeley = 0;
uint32_t Settings::SDIOutputWaitNums = 0; uint32_t Settings::RTSDIOutputWaitNums = 0;
int32_t Settings::CropFlag = 0; int32_t Settings::CropFlag = 0;
int32_t Settings::CropX = 0; int32_t Settings::CropX = 0;
......
x64/Debug/FigureOut.exe
No preview for this file type
x64/Debug/FigureOut.ilk
No preview for this file type
x64/Debug/FigureOut.log
...@@ -4,13 +4,14 @@ available device "DeckLink 8K Pro (1)" ...@@ -4,13 +4,14 @@ available device "DeckLink 8K Pro (1)"
available device "DeckLink 8K Pro (2)" available device "DeckLink 8K Pro (2)"
available device "DeckLink 8K Pro (3)" available device "DeckLink 8K Pro (3)"
available device "DeckLink 8K Pro (4)" available device "DeckLink 8K Pro (4)"
"2024-05-16 14:30:21.752" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------ "2024-05-17 19:02:03.216" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:30:21.811" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------ "2024-05-17 19:02:03.236" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:30:21.831" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------ "2024-05-17 19:02:03.256" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:30:21.851" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------ "2024-05-17 19:02:03.276" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:30:21.871" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------ "2024-05-17 19:02:03.296" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:30:21.891" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------ "2024-05-17 19:02:03.316" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:30:21.911" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------ "2024-05-17 19:02:03.336" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:30:21.931" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------ "2024-05-17 19:02:03.394" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------
"2024-05-16 14:30:21.989" index: 0 DeckLinkInputDevice get video frame No input source 80000000 ------------ "2024-05-17 19:02:04.438" decklink input fps 51
"2024-05-16 14:30:23.034" decklink input fps 51 "2024-05-17 19:02:05.439" decklink input fps 50
"2024-05-17 19:02:06.439" decklink input fps 50
x64/Debug/settings.ini
...@@ -42,13 +42,13 @@ SDIONEFRAMEDURATION=50 ...@@ -42,13 +42,13 @@ SDIONEFRAMEDURATION=50
RECORDFLAG=1 RECORDFLAG=1
OPENOMP=1 OPENOMP=1
TIMEOUTFRAMES=250 TIMEOUTFRAMES=250
DELEYSDINUMS=0 SECONDDELEYSDINUMS=100
REPLAY_START_FORWARD=50 REPLAY_START_FORWARD=50
REPLAY_END_DELEY=50 REPLAY_END_DELEY=50
SDIOUTPUTWAITNUMS=5 RTSDIOUTPUTWAITNUMS=5
CROPRECORD=0c CROPRECORD=0
CROP_X=400 CROP_X=400
CROP_Y=1800 CROP_Y=1800
CROP_DIRECTION=3 CROP_DIRECTION=3
UI_UDP_PORT=8100 UI_UDP_PORT=8100
UI_IP_ADDR=192.168.31.83 UI_IP_ADDR=127.0.0.1
x64/Release/settings.ini
...@@ -42,10 +42,10 @@ SDIONEFRAMEDURATION=50 ...@@ -42,10 +42,10 @@ SDIONEFRAMEDURATION=50
RECORDFLAG=1 RECORDFLAG=1
OPENOMP=1 OPENOMP=1
TIMEOUTFRAMES=250 TIMEOUTFRAMES=250
DELEYSDINUMS=0 SECONDDELEYSDINUMS=100
REPLAY_START_FORWARD=50 REPLAY_START_FORWARD=50
REPLAY_END_DELEY=50 REPLAY_END_DELEY=50
SDIOUTPUTWAITNUMS=5 RTSDIOUTPUTWAITNUMS=5
CROPRECORD=0 CROPRECORD=0
CROP_X=400 CROP_X=400
CROP_Y=1800 CROP_Y=1800
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment