annotate src/SFMT/SFMT-sse2.h @ 280:d5e5c73af7e6

reorginazed save corruption code
author Robert McIntyre <rlm@mit.edu>
date Tue, 27 Mar 2012 21:08:44 -0500
parents f9f4f1b99eed
children
rev   line source
rlm@1 1 /**
rlm@1 2 * @file SFMT-sse2.h
rlm@1 3 * @brief SIMD oriented Fast Mersenne Twister(SFMT) for Intel SSE2
rlm@1 4 *
rlm@1 5 * @author Mutsuo Saito (Hiroshima University)
rlm@1 6 * @author Makoto Matsumoto (Hiroshima University)
rlm@1 7 *
rlm@1 8 * @note We assume LITTLE ENDIAN in this file
rlm@1 9 *
rlm@1 10 * Copyright (C) 2006, 2007 Mutsuo Saito, Makoto Matsumoto and Hiroshima
rlm@1 11 * University. All rights reserved.
rlm@1 12 *
rlm@1 13 * The new BSD License is applied to this software, see LICENSE.txt
rlm@1 14 */
rlm@1 15
rlm@1 16 #ifndef SFMT_SSE2_H
rlm@1 17 #define SFMT_SSE2_H
rlm@1 18
rlm@1 19 PRE_ALWAYS static __m128i mm_recursion(__m128i *a, __m128i *b, __m128i c,
rlm@1 20 __m128i d, __m128i mask) ALWAYSINLINE;
rlm@1 21
rlm@1 22 /**
rlm@1 23 * This function represents the recursion formula.
rlm@1 24 * @param a a 128-bit part of the interal state array
rlm@1 25 * @param b a 128-bit part of the interal state array
rlm@1 26 * @param c a 128-bit part of the interal state array
rlm@1 27 * @param d a 128-bit part of the interal state array
rlm@1 28 * @param mask 128-bit mask
rlm@1 29 * @return output
rlm@1 30 */
rlm@1 31 PRE_ALWAYS static __m128i mm_recursion(__m128i *a, __m128i *b,
rlm@1 32 __m128i c, __m128i d, __m128i mask) {
rlm@1 33 __m128i v, x, y, z;
rlm@1 34
rlm@1 35 x = _mm_load_si128(a);
rlm@1 36 y = _mm_srli_epi32(*b, SR1);
rlm@1 37 z = _mm_srli_si128(c, SR2);
rlm@1 38 v = _mm_slli_epi32(d, SL1);
rlm@1 39 z = _mm_xor_si128(z, x);
rlm@1 40 z = _mm_xor_si128(z, v);
rlm@1 41 x = _mm_slli_si128(x, SL2);
rlm@1 42 y = _mm_and_si128(y, mask);
rlm@1 43 z = _mm_xor_si128(z, x);
rlm@1 44 z = _mm_xor_si128(z, y);
rlm@1 45 return z;
rlm@1 46 }
rlm@1 47
rlm@1 48 /**
rlm@1 49 * This function fills the internal state array with pseudorandom
rlm@1 50 * integers.
rlm@1 51 */
rlm@1 52 inline static void gen_rand_all(void) {
rlm@1 53 int i;
rlm@1 54 __m128i r, r1, r2, mask;
rlm@1 55 mask = _mm_set_epi32(MSK4, MSK3, MSK2, MSK1);
rlm@1 56
rlm@1 57 r1 = _mm_load_si128(&sfmt[N - 2].si);
rlm@1 58 r2 = _mm_load_si128(&sfmt[N - 1].si);
rlm@1 59 for (i = 0; i < N - POS1; i++) {
rlm@1 60 r = mm_recursion(&sfmt[i].si, &sfmt[i + POS1].si, r1, r2, mask);
rlm@1 61 _mm_store_si128(&sfmt[i].si, r);
rlm@1 62 r1 = r2;
rlm@1 63 r2 = r;
rlm@1 64 }
rlm@1 65 for (; i < N; i++) {
rlm@1 66 r = mm_recursion(&sfmt[i].si, &sfmt[i + POS1 - N].si, r1, r2, mask);
rlm@1 67 _mm_store_si128(&sfmt[i].si, r);
rlm@1 68 r1 = r2;
rlm@1 69 r2 = r;
rlm@1 70 }
rlm@1 71 }
rlm@1 72
rlm@1 73 /**
rlm@1 74 * This function fills the user-specified array with pseudorandom
rlm@1 75 * integers.
rlm@1 76 *
rlm@1 77 * @param array an 128-bit array to be filled by pseudorandom numbers.
rlm@1 78 * @param size number of 128-bit pesudorandom numbers to be generated.
rlm@1 79 */
rlm@1 80 inline static void gen_rand_array(w128_t *array, int size) {
rlm@1 81 int i, j;
rlm@1 82 __m128i r, r1, r2, mask;
rlm@1 83 mask = _mm_set_epi32(MSK4, MSK3, MSK2, MSK1);
rlm@1 84
rlm@1 85 r1 = _mm_load_si128(&sfmt[N - 2].si);
rlm@1 86 r2 = _mm_load_si128(&sfmt[N - 1].si);
rlm@1 87 for (i = 0; i < N - POS1; i++) {
rlm@1 88 r = mm_recursion(&sfmt[i].si, &sfmt[i + POS1].si, r1, r2, mask);
rlm@1 89 _mm_store_si128(&array[i].si, r);
rlm@1 90 r1 = r2;
rlm@1 91 r2 = r;
rlm@1 92 }
rlm@1 93 for (; i < N; i++) {
rlm@1 94 r = mm_recursion(&sfmt[i].si, &array[i + POS1 - N].si, r1, r2, mask);
rlm@1 95 _mm_store_si128(&array[i].si, r);
rlm@1 96 r1 = r2;
rlm@1 97 r2 = r;
rlm@1 98 }
rlm@1 99 /* main loop */
rlm@1 100 for (; i < size - N; i++) {
rlm@1 101 r = mm_recursion(&array[i - N].si, &array[i + POS1 - N].si, r1, r2,
rlm@1 102 mask);
rlm@1 103 _mm_store_si128(&array[i].si, r);
rlm@1 104 r1 = r2;
rlm@1 105 r2 = r;
rlm@1 106 }
rlm@1 107 for (j = 0; j < 2 * N - size; j++) {
rlm@1 108 r = _mm_load_si128(&array[j + size - N].si);
rlm@1 109 _mm_store_si128(&sfmt[j].si, r);
rlm@1 110 }
rlm@1 111 for (; i < size; i++) {
rlm@1 112 r = mm_recursion(&array[i - N].si, &array[i + POS1 - N].si, r1, r2,
rlm@1 113 mask);
rlm@1 114 _mm_store_si128(&array[i].si, r);
rlm@1 115 _mm_store_si128(&sfmt[j++].si, r);
rlm@1 116 r1 = r2;
rlm@1 117 r2 = r;
rlm@1 118 }
rlm@1 119 }
rlm@1 120
rlm@1 121 #endif