modes/asm/ghashv8-armx.pl: modulo-schedule loop.
[openssl.git] / crypto / cast / cast_lcl.h
1 /*
2  * Copyright 1995-2017 The OpenSSL Project Authors. All Rights Reserved.
3  *
4  * Licensed under the OpenSSL license (the "License").  You may not use
5  * this file except in compliance with the License.  You can obtain a copy
6  * in the file LICENSE in the source distribution or at
7  * https://www.openssl.org/source/license.html
8  */
9
10 #ifdef OPENSSL_SYS_WIN32
11 # include <stdlib.h>
12 #endif
13
14 #undef c2l
15 #define c2l(c,l)        (l =((unsigned long)(*((c)++)))    , \
16                          l|=((unsigned long)(*((c)++)))<< 8L, \
17                          l|=((unsigned long)(*((c)++)))<<16L, \
18                          l|=((unsigned long)(*((c)++)))<<24L)
19
20 /* NOTE - c is not incremented as per c2l */
21 #undef c2ln
22 #define c2ln(c,l1,l2,n) { \
23                         c+=n; \
24                         l1=l2=0; \
25                         switch (n) { \
26                         case 8: l2 =((unsigned long)(*(--(c))))<<24L; \
27                         case 7: l2|=((unsigned long)(*(--(c))))<<16L; \
28                         case 6: l2|=((unsigned long)(*(--(c))))<< 8L; \
29                         case 5: l2|=((unsigned long)(*(--(c))));     \
30                         case 4: l1 =((unsigned long)(*(--(c))))<<24L; \
31                         case 3: l1|=((unsigned long)(*(--(c))))<<16L; \
32                         case 2: l1|=((unsigned long)(*(--(c))))<< 8L; \
33                         case 1: l1|=((unsigned long)(*(--(c))));     \
34                                 } \
35                         }
36
37 #undef l2c
38 #define l2c(l,c)        (*((c)++)=(unsigned char)(((l)     )&0xff), \
39                          *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
40                          *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
41                          *((c)++)=(unsigned char)(((l)>>24L)&0xff))
42
43 /* NOTE - c is not incremented as per l2c */
44 #undef l2cn
45 #define l2cn(l1,l2,c,n) { \
46                         c+=n; \
47                         switch (n) { \
48                         case 8: *(--(c))=(unsigned char)(((l2)>>24L)&0xff); \
49                         case 7: *(--(c))=(unsigned char)(((l2)>>16L)&0xff); \
50                         case 6: *(--(c))=(unsigned char)(((l2)>> 8L)&0xff); \
51                         case 5: *(--(c))=(unsigned char)(((l2)     )&0xff); \
52                         case 4: *(--(c))=(unsigned char)(((l1)>>24L)&0xff); \
53                         case 3: *(--(c))=(unsigned char)(((l1)>>16L)&0xff); \
54                         case 2: *(--(c))=(unsigned char)(((l1)>> 8L)&0xff); \
55                         case 1: *(--(c))=(unsigned char)(((l1)     )&0xff); \
56                                 } \
57                         }
58
59 /* NOTE - c is not incremented as per n2l */
60 #define n2ln(c,l1,l2,n) { \
61                         c+=n; \
62                         l1=l2=0; \
63                         switch (n) { \
64                         case 8: l2 =((unsigned long)(*(--(c))))    ; \
65                         /* fall thru */                              \
66                         case 7: l2|=((unsigned long)(*(--(c))))<< 8; \
67                         /* fall thru */                              \
68                         case 6: l2|=((unsigned long)(*(--(c))))<<16; \
69                         /* fall thru */                              \
70                         case 5: l2|=((unsigned long)(*(--(c))))<<24; \
71                         /* fall thru */                              \
72                         case 4: l1 =((unsigned long)(*(--(c))))    ; \
73                         /* fall thru */                              \
74                         case 3: l1|=((unsigned long)(*(--(c))))<< 8; \
75                         /* fall thru */                              \
76                         case 2: l1|=((unsigned long)(*(--(c))))<<16; \
77                         /* fall thru */                              \
78                         case 1: l1|=((unsigned long)(*(--(c))))<<24; \
79                                 } \
80                         }
81
82 /* NOTE - c is not incremented as per l2n */
83 #define l2nn(l1,l2,c,n) { \
84                         c+=n; \
85                         switch (n) { \
86                         case 8: *(--(c))=(unsigned char)(((l2)    )&0xff); \
87                         /* fall thru */                                    \
88                         case 7: *(--(c))=(unsigned char)(((l2)>> 8)&0xff); \
89                         /* fall thru */                                    \
90                         case 6: *(--(c))=(unsigned char)(((l2)>>16)&0xff); \
91                         /* fall thru */                                    \
92                         case 5: *(--(c))=(unsigned char)(((l2)>>24)&0xff); \
93                         /* fall thru */                                    \
94                         case 4: *(--(c))=(unsigned char)(((l1)    )&0xff); \
95                         /* fall thru */                                    \
96                         case 3: *(--(c))=(unsigned char)(((l1)>> 8)&0xff); \
97                         /* fall thru */                                    \
98                         case 2: *(--(c))=(unsigned char)(((l1)>>16)&0xff); \
99                         /* fall thru */                                    \
100                         case 1: *(--(c))=(unsigned char)(((l1)>>24)&0xff); \
101                                 } \
102                         }
103
104 #undef n2l
105 #define n2l(c,l)        (l =((unsigned long)(*((c)++)))<<24L, \
106                          l|=((unsigned long)(*((c)++)))<<16L, \
107                          l|=((unsigned long)(*((c)++)))<< 8L, \
108                          l|=((unsigned long)(*((c)++))))
109
110 #undef l2n
111 #define l2n(l,c)        (*((c)++)=(unsigned char)(((l)>>24L)&0xff), \
112                          *((c)++)=(unsigned char)(((l)>>16L)&0xff), \
113                          *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \
114                          *((c)++)=(unsigned char)(((l)     )&0xff))
115
116 #if defined(OPENSSL_SYS_WIN32) && defined(_MSC_VER)
117 # define ROTL(a,n)     (_lrotl(a,n))
118 #else
119 # define ROTL(a,n)     ((((a)<<(n))&0xffffffffL)|((a)>>((32-(n))&31)))
120 #endif
121
122 #define C_M    0x3fc
123 #define C_0    22L
124 #define C_1    14L
125 #define C_2     6L
126 #define C_3     2L              /* left shift */
127
128 /* The rotate has an extra 16 added to it to help the x86 asm */
129 #if defined(CAST_PTR)
130 # define E_CAST(n,key,L,R,OP1,OP2,OP3) \
131         { \
132         int i; \
133         t=(key[n*2] OP1 R)&0xffffffffL; \
134         i=key[n*2+1]; \
135         t=ROTL(t,i); \
136         L^= (((((*(CAST_LONG *)((unsigned char *) \
137                         CAST_S_table0+((t>>C_2)&C_M)) OP2 \
138                 *(CAST_LONG *)((unsigned char *) \
139                         CAST_S_table1+((t<<C_3)&C_M)))&0xffffffffL) OP3 \
140                 *(CAST_LONG *)((unsigned char *) \
141                         CAST_S_table2+((t>>C_0)&C_M)))&0xffffffffL) OP1 \
142                 *(CAST_LONG *)((unsigned char *) \
143                         CAST_S_table3+((t>>C_1)&C_M)))&0xffffffffL; \
144         }
145 #elif defined(CAST_PTR2)
146 # define E_CAST(n,key,L,R,OP1,OP2,OP3) \
147         { \
148         int i; \
149         CAST_LONG u,v,w; \
150         w=(key[n*2] OP1 R)&0xffffffffL; \
151         i=key[n*2+1]; \
152         w=ROTL(w,i); \
153         u=w>>C_2; \
154         v=w<<C_3; \
155         u&=C_M; \
156         v&=C_M; \
157         t= *(CAST_LONG *)((unsigned char *)CAST_S_table0+u); \
158         u=w>>C_0; \
159         t=(t OP2 *(CAST_LONG *)((unsigned char *)CAST_S_table1+v))&0xffffffffL;\
160         v=w>>C_1; \
161         u&=C_M; \
162         v&=C_M; \
163         t=(t OP3 *(CAST_LONG *)((unsigned char *)CAST_S_table2+u)&0xffffffffL);\
164         t=(t OP1 *(CAST_LONG *)((unsigned char *)CAST_S_table3+v)&0xffffffffL);\
165         L^=(t&0xffffffff); \
166         }
167 #else
168 # define E_CAST(n,key,L,R,OP1,OP2,OP3) \
169         { \
170         CAST_LONG a,b,c,d; \
171         t=(key[n*2] OP1 R)&0xffffffff; \
172         t=ROTL(t,(key[n*2+1])); \
173         a=CAST_S_table0[(t>> 8)&0xff]; \
174         b=CAST_S_table1[(t    )&0xff]; \
175         c=CAST_S_table2[(t>>24)&0xff]; \
176         d=CAST_S_table3[(t>>16)&0xff]; \
177         L^=(((((a OP2 b)&0xffffffffL) OP3 c)&0xffffffffL) OP1 d)&0xffffffffL; \
178         }
179 #endif
180
181 extern const CAST_LONG CAST_S_table0[256];
182 extern const CAST_LONG CAST_S_table1[256];
183 extern const CAST_LONG CAST_S_table2[256];
184 extern const CAST_LONG CAST_S_table3[256];
185 extern const CAST_LONG CAST_S_table4[256];
186 extern const CAST_LONG CAST_S_table5[256];
187 extern const CAST_LONG CAST_S_table6[256];
188 extern const CAST_LONG CAST_S_table7[256];