annotate src/madplug/SFMT-sse2.c @ 2284:d19b53359b24

cleaned up the sndfile wav plugin, currently limiting it ONLY TO WAV PLAYBACK. if somebody is more experienced with it and wants to restore the other formats, go ahead (maybe change the name of the plugin too?).
author mf0102 <0102@gmx.at>
date Wed, 09 Jan 2008 15:41:22 +0100
parents 7e14701aef54
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
922
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
1 /**
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
2 * @file SFMT-sse2.c
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
3 * @brief SIMD oriented Fast Mersenne Twister(SFMT) for intel SSE2
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
4 *
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
5 * @author Mutsuo Saito (Hiroshima University)
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
6 * @author Makoto Matsumoto (Hiroshima University)
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
7 *
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
8 * @note We assume LITTLE ENDIAN in this file
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
9 *
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
10 * Copyright (C) 2006, 2007 Mutsuo Saito, Makoto Matsumoto and Hiroshima
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
11 * University. All rights reserved.
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
12 *
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
13 * The new BSD License is applied to this software, see LICENSE.txt
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
14 */
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
15
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
16 /**
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
17 * This function represents the recursion formula.
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
18 * @param a a 128-bit part of the interal state array
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
19 * @param b a 128-bit part of the interal state array
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
20 * @param c a 128-bit part of the interal state array
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
21 * @param d a 128-bit part of the interal state array
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
22 * @param mask 128-bit mask
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
23 * @return output
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
24 */
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
25 inline static __m128i mm_recursion(__m128i *a, __m128i *b,
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
26 __m128i c, __m128i d, __m128i mask) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
27 __m128i v, x, y, z;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
28
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
29 x = _mm_load_si128(a);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
30 y = _mm_srli_epi32(*b, SR1);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
31 z = _mm_srli_si128(c, SR2);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
32 v = _mm_slli_epi32(d, SL1);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
33 z = _mm_xor_si128(z, x);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
34 z = _mm_xor_si128(z, v);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
35 x = _mm_slli_si128(x, SL2);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
36 y = _mm_and_si128(y, mask);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
37 z = _mm_xor_si128(z, x);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
38 z = _mm_xor_si128(z, y);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
39 return z;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
40 }
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
41
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
42 /**
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
43 * This function fills the internal state array with psedorandom
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
44 * integers.
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
45 */
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
46 inline void gen_rand_all(void) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
47 int i;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
48 __m128i r, r1, r2, mask;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
49 mask = _mm_set_epi32(MSK4, MSK3, MSK2, MSK1);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
50
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
51 r1 = _mm_load_si128(&sfmt[N - 2].si);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
52 r2 = _mm_load_si128(&sfmt[N - 1].si);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
53 for (i = 0; i < N - POS1; i++) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
54 r = mm_recursion(&sfmt[i].si, &sfmt[i + POS1].si, r1, r2, mask);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
55 _mm_store_si128(&sfmt[i].si, r);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
56 r1 = r2;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
57 r2 = r;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
58 }
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
59 for (; i < N; i++) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
60 r = mm_recursion(&sfmt[i].si, &sfmt[i + POS1 - N].si, r1, r2, mask);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
61 _mm_store_si128(&sfmt[i].si, r);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
62 r1 = r2;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
63 r2 = r;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
64 }
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
65 }
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
66
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
67 /**
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
68 * This function fills the user-specified array with psedorandom
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
69 * integers.
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
70 *
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
71 * @param array an 128-bit array to be filled by pseudorandom numbers.
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
72 * @param size number of 128-bit pesudorandom numbers to be generated.
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
73 */
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
74 inline static void gen_rand_array(w128_t array[], int size) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
75 int i, j;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
76 __m128i r, r1, r2, mask;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
77 mask = _mm_set_epi32(MSK4, MSK3, MSK2, MSK1);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
78
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
79 r1 = _mm_load_si128(&sfmt[N - 2].si);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
80 r2 = _mm_load_si128(&sfmt[N - 1].si);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
81 for (i = 0; i < N - POS1; i++) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
82 r = mm_recursion(&sfmt[i].si, &sfmt[i + POS1].si, r1, r2, mask);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
83 _mm_store_si128(&array[i].si, r);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
84 r1 = r2;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
85 r2 = r;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
86 }
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
87 for (; i < N; i++) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
88 r = mm_recursion(&sfmt[i].si, &array[i + POS1 - N].si, r1, r2, mask);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
89 _mm_store_si128(&array[i].si, r);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
90 r1 = r2;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
91 r2 = r;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
92 }
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
93 /* main loop */
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
94 for (; i < size - N; i++) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
95 r = mm_recursion(&array[i - N].si, &array[i + POS1 - N].si, r1, r2,
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
96 mask);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
97 _mm_store_si128(&array[i].si, r);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
98 r1 = r2;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
99 r2 = r;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
100 }
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
101 for (j = 0; j < 2 * N - size; j++) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
102 r = _mm_load_si128(&array[j + size - N].si);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
103 _mm_store_si128(&sfmt[j].si, r);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
104 }
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
105 for (; i < size; i++) {
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
106 r = mm_recursion(&array[i - N].si, &array[i + POS1 - N].si, r1, r2,
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
107 mask);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
108 _mm_store_si128(&array[i].si, r);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
109 _mm_store_si128(&sfmt[j++].si, r);
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
110 r1 = r2;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
111 r2 = r;
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
112 }
7e14701aef54 [svn] - replace random number generator in dithering code with SIMD-oriented Fast Mersenne Twister (SFMT). it reduces CPU load on SSE2 or AltiVec capable platform.
yaz
parents:
diff changeset
113 }