1/*
2   BLAKE2 reference source code package - optimized C implementations
3
4   Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
5
6   To the extent possible under law, the author(s) have dedicated all copyright
7   and related and neighboring rights to this software to the public domain
8   worldwide. This software is distributed without any warranty.
9
10   You should have received a copy of the CC0 Public Domain Dedication along with
11   this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
12*/
13#pragma once
14#ifndef __BLAKE2S_LOAD_SSE41_H__
15#define __BLAKE2S_LOAD_SSE41_H__
16
17#define LOAD_MSG_0_1(buf) \
18buf = TOI(_mm_shuffle_ps(TOF(m0), TOF(m1), _MM_SHUFFLE(2,0,2,0)));
19
20#define LOAD_MSG_0_2(buf) \
21buf = TOI(_mm_shuffle_ps(TOF(m0), TOF(m1), _MM_SHUFFLE(3,1,3,1)));
22
23#define LOAD_MSG_0_3(buf) \
24buf = TOI(_mm_shuffle_ps(TOF(m2), TOF(m3), _MM_SHUFFLE(2,0,2,0)));
25
26#define LOAD_MSG_0_4(buf) \
27buf = TOI(_mm_shuffle_ps(TOF(m2), TOF(m3), _MM_SHUFFLE(3,1,3,1)));
28
29#define LOAD_MSG_1_1(buf) \
30t0 = _mm_blend_epi16(m1, m2, 0x0C); \
31t1 = _mm_slli_si128(m3, 4); \
32t2 = _mm_blend_epi16(t0, t1, 0xF0); \
33buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,1,0,3));
34
35#define LOAD_MSG_1_2(buf) \
36t0 = _mm_shuffle_epi32(m2,_MM_SHUFFLE(0,0,2,0)); \
37t1 = _mm_blend_epi16(m1,m3,0xC0); \
38t2 = _mm_blend_epi16(t0, t1, 0xF0); \
39buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,3,0,1));
40
41#define LOAD_MSG_1_3(buf) \
42t0 = _mm_slli_si128(m1, 4); \
43t1 = _mm_blend_epi16(m2, t0, 0x30); \
44t2 = _mm_blend_epi16(m0, t1, 0xF0); \
45buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,3,0,1));
46
47#define LOAD_MSG_1_4(buf) \
48t0 = _mm_unpackhi_epi32(m0,m1); \
49t1 = _mm_slli_si128(m3, 4); \
50t2 = _mm_blend_epi16(t0, t1, 0x0C); \
51buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,3,0,1));
52
53#define LOAD_MSG_2_1(buf) \
54t0 = _mm_unpackhi_epi32(m2,m3); \
55t1 = _mm_blend_epi16(m3,m1,0x0C); \
56t2 = _mm_blend_epi16(t0, t1, 0x0F); \
57buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(3,1,0,2));
58
59#define LOAD_MSG_2_2(buf) \
60t0 = _mm_unpacklo_epi32(m2,m0); \
61t1 = _mm_blend_epi16(t0, m0, 0xF0); \
62t2 = _mm_slli_si128(m3, 8); \
63buf = _mm_blend_epi16(t1, t2, 0xC0);
64
65#define LOAD_MSG_2_3(buf) \
66t0 = _mm_blend_epi16(m0, m2, 0x3C); \
67t1 = _mm_srli_si128(m1, 12); \
68t2 = _mm_blend_epi16(t0,t1,0x03); \
69buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,0,3,2));
70
71#define LOAD_MSG_2_4(buf) \
72t0 = _mm_slli_si128(m3, 4); \
73t1 = _mm_blend_epi16(m0, m1, 0x33); \
74t2 = _mm_blend_epi16(t1, t0, 0xC0); \
75buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(0,1,2,3));
76
77#define LOAD_MSG_3_1(buf) \
78t0 = _mm_unpackhi_epi32(m0,m1); \
79t1 = _mm_unpackhi_epi32(t0, m2); \
80t2 = _mm_blend_epi16(t1, m3, 0x0C); \
81buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(3,1,0,2));
82
83#define LOAD_MSG_3_2(buf) \
84t0 = _mm_slli_si128(m2, 8); \
85t1 = _mm_blend_epi16(m3,m0,0x0C); \
86t2 = _mm_blend_epi16(t1, t0, 0xC0); \
87buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,0,1,3));
88
89#define LOAD_MSG_3_3(buf) \
90t0 = _mm_blend_epi16(m0,m1,0x0F); \
91t1 = _mm_blend_epi16(t0, m3, 0xC0); \
92buf = _mm_shuffle_epi32(t1, _MM_SHUFFLE(3,0,1,2));
93
94#define LOAD_MSG_3_4(buf) \
95t0 = _mm_unpacklo_epi32(m0,m2); \
96t1 = _mm_unpackhi_epi32(m1,m2); \
97buf = _mm_unpacklo_epi64(t1,t0);
98
99#define LOAD_MSG_4_1(buf) \
100t0 = _mm_unpacklo_epi64(m1,m2); \
101t1 = _mm_unpackhi_epi64(m0,m2); \
102t2 = _mm_blend_epi16(t0,t1,0x33); \
103buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,0,1,3));
104
105#define LOAD_MSG_4_2(buf) \
106t0 = _mm_unpackhi_epi64(m1,m3); \
107t1 = _mm_unpacklo_epi64(m0,m1); \
108buf = _mm_blend_epi16(t0,t1,0x33);
109
110#define LOAD_MSG_4_3(buf) \
111t0 = _mm_unpackhi_epi64(m3,m1); \
112t1 = _mm_unpackhi_epi64(m2,m0); \
113buf = _mm_blend_epi16(t1,t0,0x33);
114
115#define LOAD_MSG_4_4(buf) \
116t0 = _mm_blend_epi16(m0,m2,0x03); \
117t1 = _mm_slli_si128(t0, 8); \
118t2 = _mm_blend_epi16(t1,m3,0x0F); \
119buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,2,0,3));
120
121#define LOAD_MSG_5_1(buf) \
122t0 = _mm_unpackhi_epi32(m0,m1); \
123t1 = _mm_unpacklo_epi32(m0,m2); \
124buf = _mm_unpacklo_epi64(t0,t1);
125
126#define LOAD_MSG_5_2(buf) \
127t0 = _mm_srli_si128(m2, 4); \
128t1 = _mm_blend_epi16(m0,m3,0x03); \
129buf = _mm_blend_epi16(t1,t0,0x3C);
130
131#define LOAD_MSG_5_3(buf) \
132t0 = _mm_blend_epi16(m1,m0,0x0C); \
133t1 = _mm_srli_si128(m3, 4); \
134t2 = _mm_blend_epi16(t0,t1,0x30); \
135buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,2,3,0));
136
137#define LOAD_MSG_5_4(buf) \
138t0 = _mm_unpacklo_epi64(m1,m2); \
139t1= _mm_shuffle_epi32(m3, _MM_SHUFFLE(0,2,0,1)); \
140buf = _mm_blend_epi16(t0,t1,0x33);
141
142#define LOAD_MSG_6_1(buf) \
143t0 = _mm_slli_si128(m1, 12); \
144t1 = _mm_blend_epi16(m0,m3,0x33); \
145buf = _mm_blend_epi16(t1,t0,0xC0);
146
147#define LOAD_MSG_6_2(buf) \
148t0 = _mm_blend_epi16(m3,m2,0x30); \
149t1 = _mm_srli_si128(m1, 4); \
150t2 = _mm_blend_epi16(t0,t1,0x03); \
151buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,1,3,0));
152
153#define LOAD_MSG_6_3(buf) \
154t0 = _mm_unpacklo_epi64(m0,m2); \
155t1 = _mm_srli_si128(m1, 4); \
156buf = _mm_shuffle_epi32(_mm_blend_epi16(t0,t1,0x0C), _MM_SHUFFLE(2,3,1,0));
157
158#define LOAD_MSG_6_4(buf) \
159t0 = _mm_unpackhi_epi32(m1,m2); \
160t1 = _mm_unpackhi_epi64(m0,t0); \
161buf = _mm_shuffle_epi32(t1, _MM_SHUFFLE(3,0,1,2));
162
163#define LOAD_MSG_7_1(buf) \
164t0 = _mm_unpackhi_epi32(m0,m1); \
165t1 = _mm_blend_epi16(t0,m3,0x0F); \
166buf = _mm_shuffle_epi32(t1,_MM_SHUFFLE(2,0,3,1));
167
168#define LOAD_MSG_7_2(buf) \
169t0 = _mm_blend_epi16(m2,m3,0x30); \
170t1 = _mm_srli_si128(m0,4); \
171t2 = _mm_blend_epi16(t0,t1,0x03); \
172buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,0,2,3));
173
174#define LOAD_MSG_7_3(buf) \
175t0 = _mm_unpackhi_epi64(m0,m3); \
176t1 = _mm_unpacklo_epi64(m1,m2); \
177t2 = _mm_blend_epi16(t0,t1,0x3C); \
178buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(0,2,3,1));
179
180#define LOAD_MSG_7_4(buf) \
181t0 = _mm_unpacklo_epi32(m0,m1); \
182t1 = _mm_unpackhi_epi32(m1,m2); \
183buf = _mm_unpacklo_epi64(t0,t1);
184
185#define LOAD_MSG_8_1(buf) \
186t0 = _mm_unpackhi_epi32(m1,m3); \
187t1 = _mm_unpacklo_epi64(t0,m0); \
188t2 = _mm_blend_epi16(t1,m2,0xC0); \
189buf = _mm_shufflehi_epi16(t2,_MM_SHUFFLE(1,0,3,2));
190
191#define LOAD_MSG_8_2(buf) \
192t0 = _mm_unpackhi_epi32(m0,m3); \
193t1 = _mm_blend_epi16(m2,t0,0xF0); \
194buf = _mm_shuffle_epi32(t1,_MM_SHUFFLE(0,2,1,3));
195
196#define LOAD_MSG_8_3(buf) \
197t0 = _mm_blend_epi16(m2,m0,0x0C); \
198t1 = _mm_slli_si128(t0,4); \
199buf = _mm_blend_epi16(t1,m3,0x0F);
200
201#define LOAD_MSG_8_4(buf) \
202t0 = _mm_blend_epi16(m1,m0,0x30); \
203buf = _mm_shuffle_epi32(t0,_MM_SHUFFLE(1,0,3,2));
204
205#define LOAD_MSG_9_1(buf) \
206t0 = _mm_blend_epi16(m0,m2,0x03); \
207t1 = _mm_blend_epi16(m1,m2,0x30); \
208t2 = _mm_blend_epi16(t1,t0,0x0F); \
209buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(1,3,0,2));
210
211#define LOAD_MSG_9_2(buf) \
212t0 = _mm_slli_si128(m0,4); \
213t1 = _mm_blend_epi16(m1,t0,0xC0); \
214buf = _mm_shuffle_epi32(t1,_MM_SHUFFLE(1,2,0,3));
215
216#define LOAD_MSG_9_3(buf) \
217t0 = _mm_unpackhi_epi32(m0,m3); \
218t1 = _mm_unpacklo_epi32(m2,m3); \
219t2 = _mm_unpackhi_epi64(t0,t1); \
220buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(3,0,2,1));
221
222#define LOAD_MSG_9_4(buf) \
223t0 = _mm_blend_epi16(m3,m2,0xC0); \
224t1 = _mm_unpacklo_epi32(m0,m3); \
225t2 = _mm_blend_epi16(t0,t1,0x0F); \
226buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(0,1,2,3));
227
228#endif
229
230