This commit is contained in:
powturbo
2015-05-28 16:49:30 +02:00
parent 3b40f005f2
commit b7ea12fc03

227
transpose.c Normal file
View File

@ -0,0 +1,227 @@
/**
Copyright (C) powturbo 2013-2015
GPL v2 License
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
- homepage : https://sites.google.com/site/powturbo/
- github : https://github.com/powturbo
- twitter : https://twitter.com/powturbo
- email : powturbo [_AT_] gmail [_DOT_] com
**/
// transpose.c - byte transpose
#ifndef ESIZE
#include "transpose.h"
#define ESIZE 2
#include __FILE__
#undef ESIZE
#define ESIZE 3
#include __FILE__
#undef ESIZE
#define ESIZE 4
#include __FILE__
#undef ESIZE
#define ESIZE 8
#include __FILE__
#undef ESIZE
#define ESIZE 16
#include __FILE__
#undef ESIZE
#ifdef __SSSE3__
#include <tmmintrin.h>
#include <stdio.h>
void transposev4(unsigned char *in, unsigned n, unsigned char *out) {
unsigned bsize = n/4,i;
static unsigned char s[16] = { 0, 4, 8, 12, 1, 5, 9, 13, 2, 6, 10, 14, 3, 7, 11, 15 };
unsigned char *op,*ip,*e;
__m128i sv = _mm_loadu_si128((__m128i *)s);
for(ip = in,op = out; ip != in+(n&~(64-1)); op += 16) {
__m128i iv[4],ov[4];
iv[0] = _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)ip), sv); ip += 16;
iv[1] = _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)ip), sv); ip += 16;
iv[2] = _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)ip), sv); ip += 16;
iv[3] = _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)ip), sv); ip += 16;
ov[0] = _mm_unpacklo_epi32(iv[0], iv[1]);
ov[1] = _mm_unpackhi_epi32(iv[0], iv[1]);
ov[2] = _mm_unpacklo_epi32(iv[2], iv[3]);
ov[3] = _mm_unpackhi_epi32(iv[2], iv[3]);
iv[0] = _mm_unpacklo_epi64(ov[0], ov[2]);
iv[1] = _mm_unpackhi_epi64(ov[0], ov[2]);
iv[2] = _mm_unpacklo_epi64(ov[1], ov[3]);
iv[3] = _mm_unpackhi_epi64(ov[1], ov[3]);
_mm_storeu_si128((__m128i *)op, iv[0]);
_mm_storeu_si128((__m128i *)(op+(i =bsize)), iv[1]);
_mm_storeu_si128((__m128i *)(op+(i+=bsize)), iv[2]);
_mm_storeu_si128((__m128i *)(op+(i+=bsize)), iv[3]);
}
transpose4(ip, (in+n)-ip, op);
}
void untransposev4(unsigned char *in, unsigned n, unsigned char *out) {
unsigned bsize = n/4,i;
unsigned char *op,*ip;
for(op = out,ip = in; op != out+(n&~(64-1)); ip += 16) {
__m128i iv[4], ov[4];
iv[0] = _mm_loadu_si128((__m128i *) ip) ;
iv[1] = _mm_loadu_si128((__m128i *)(ip+(i =bsize)));
iv[2] = _mm_loadu_si128((__m128i *)(ip+(i+=bsize)));
iv[3] = _mm_loadu_si128((__m128i *)(ip+(i+=bsize)));
ov[0] = _mm_unpacklo_epi8(iv[0], iv[1]);
ov[1] = _mm_unpackhi_epi8(iv[0], iv[1]);
ov[2] = _mm_unpacklo_epi8(iv[2], iv[3]);
ov[3] = _mm_unpackhi_epi8(iv[2], iv[3]);
_mm_storeu_si128((__m128i *)op, _mm_unpacklo_epi16(ov[0], ov[2])); op += 16;
_mm_storeu_si128((__m128i *)op, _mm_unpackhi_epi16(ov[0], ov[2])); op += 16;
_mm_storeu_si128((__m128i *)op, _mm_unpacklo_epi16(ov[1], ov[3])); op += 16;
_mm_storeu_si128((__m128i *)op, _mm_unpackhi_epi16(ov[1], ov[3])); op += 16;
}
untranspose4(ip, (out+n)-op, op);
}
#endif
void transpose(unsigned char *in, unsigned n, unsigned char *out, unsigned esize) {
switch(esize) {
case 2: transpose2 (in,n,out); break;
case 3: transpose3 (in,n,out); break;
case 4: transpose4 (in,n,out); break;
case 8: transpose8 (in,n,out); break;
case 16: transpose16(in,n,out); break;
default: {
unsigned bsize = n/esize,i;
unsigned char *op,*ip;
for(ip = in,op = out; ip < in+bsize*esize; op++)
for(i = 0; i < esize; i++)
op[i*bsize] = *ip++;
for(op += esize*bsize; ip < in+n;)
*op++ = *ip++;
}
}
}
void untranspose(unsigned char *in, unsigned n, unsigned char *out, unsigned esize) {
switch(esize) {
case 2: untranspose2 (in,n,out); break;
case 3: untranspose3 (in,n,out); break;
case 4: untranspose4 (in,n,out); break;
case 8: untranspose8 (in,n,out); break;
case 16: untranspose16(in,n,out); break;
default: {
unsigned bsize = n/esize,i;
unsigned char *op,*ip;
for(op = out,ip = in; op < out+bsize*esize; ip++)
for(i = 0; i < esize; i++)
*op++ = ip[i*bsize];
for(ip += esize*bsize; op < out+n;)
*op++ = *ip++;
}
}
}
#else
#include "conf.h"
#define powof2(n) !((n)&((n)-1))
void TEMPLATE2(transpose, ESIZE)(unsigned char *in, unsigned n, unsigned char *out) {
unsigned bsize = n/ESIZE,i;
unsigned char *op,*ip,*e;
#if powof2(ESIZE)
e = in+(n&~(ESIZE-1));
#else
e = in+bsize*ESIZE;
#endif
for(ip = in,op = out; ip < e; op++) {
op[0] = *ip++;
op[i =bsize] = *ip++;
#if ESIZE > 2
op[i+=bsize] = *ip++;
#if ESIZE > 3
op[i+=bsize] = *ip++;
#if ESIZE > 4
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
#if ESIZE > 8
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
op[i+=bsize] = *ip++;
#endif
#endif
#endif
#endif
}
op = out+bsize*ESIZE; while(ip < in+n) *op++ = *ip++;
}
void TEMPLATE2(untranspose, ESIZE)(unsigned char *in, unsigned n, unsigned char *out) {
unsigned bsize = n/ESIZE,i;
unsigned char *op,*ip,*e;
#if powof2(ESIZE)
e = out+(n&~(ESIZE-1));
#else
e = out+bsize*ESIZE;
#endif
for(op = out,ip = in; op < e; ip++) {
*op++ = ip[0];
*op++ = ip[i =bsize];
#if ESIZE > 2
*op++ = ip[i+=bsize];
#if ESIZE > 3
*op++ = ip[i+=bsize];
#if ESIZE > 4
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
#if ESIZE > 8
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
*op++ = ip[i+=bsize];
#endif
#endif
#endif
#endif
}
ip = in+bsize*ESIZE;
while(op < out+n)
*op++ = *ip++;
}
#endif