summaryrefslogtreecommitdiff
path: root/src/lib/libcrypto/sha/sha512_amd64_generic.S
diff options
context:
space:
mode:
authorjsing <>2026-05-07 15:38:03 +0000
committerjsing <>2026-05-07 15:38:03 +0000
commitccc4eae3de33ef320762657c4be6c9dfd9c2de82 (patch)
tree81f773dc013844248de1103391a3e49d252c1587 /src/lib/libcrypto/sha/sha512_amd64_generic.S
parent97930339bcae324ce6c463f52bdcdf98c4be2941 (diff)
downloadopenbsd-ccc4eae3de33ef320762657c4be6c9dfd9c2de82.tar.gz
openbsd-ccc4eae3de33ef320762657c4be6c9dfd9c2de82.tar.bz2
openbsd-ccc4eae3de33ef320762657c4be6c9dfd9c2de82.zip
Use a define based instruction separator in SHA assembly.
Unfortunately, not all assemblers use the same instruction separator. In particular, LLVM on macOS uses %% as an instruction separator, while most other assemblers use a semi-colon. ok kenjiro@ tb@
Diffstat (limited to 'src/lib/libcrypto/sha/sha512_amd64_generic.S')
-rw-r--r--src/lib/libcrypto/sha/sha512_amd64_generic.S104
1 files changed, 53 insertions, 51 deletions
diff --git a/src/lib/libcrypto/sha/sha512_amd64_generic.S b/src/lib/libcrypto/sha/sha512_amd64_generic.S
index de759875f4..fac9d95655 100644
--- a/src/lib/libcrypto/sha/sha512_amd64_generic.S
+++ b/src/lib/libcrypto/sha/sha512_amd64_generic.S
@@ -1,4 +1,4 @@
1/* $OpenBSD: sha512_amd64_generic.S,v 1.4 2026/03/28 13:11:28 jsing Exp $ */ 1/* $OpenBSD: sha512_amd64_generic.S,v 1.5 2026/05/07 15:38:03 jsing Exp $ */
2/* 2/*
3 * Copyright (c) 2024 Joel Sing <jsing@openbsd.org> 3 * Copyright (c) 2024 Joel Sing <jsing@openbsd.org>
4 * 4 *
@@ -17,6 +17,8 @@
17 17
18#include "crypto_assembly.h" 18#include "crypto_assembly.h"
19 19
20#define _SEP CRYPTO_ASSEMBLY_SEPARATOR
21
20#define ctx %rdi 22#define ctx %rdi
21#define in %rsi 23#define in %rsi
22#define num %rdx 24#define num %rdx
@@ -45,8 +47,8 @@
45 * Wt = Mt 47 * Wt = Mt
46 */ 48 */
47#define sha512_message_schedule_load(idx, m, w, wt) \ 49#define sha512_message_schedule_load(idx, m, w, wt) \
48 movq (m, round, 8), wt; \ 50 movq (m, round, 8), wt _SEP \
49 bswapq wt; \ 51 bswapq wt _SEP \
50 movq wt, ((idx&0xf)*8)(w) 52 movq wt, ((idx&0xf)*8)(w)
51 53
52/* 54/*
@@ -59,25 +61,25 @@
59 * 61 *
60 */ 62 */
61#define sha512_message_schedule_update(idx, w, wt) \ 63#define sha512_message_schedule_update(idx, w, wt) \
62 movq (((idx-2)&0xf)*8)(w), wt; /* sigma1 */ \ 64 movq (((idx-2)&0xf)*8)(w), wt /* sigma1 */ _SEP \
63 movq wt, tmp1; /* sigma1 */ \ 65 movq wt, tmp1 /* sigma1 */ _SEP \
64 rorq $(61-19), tmp1; /* sigma1 */ \ 66 rorq $(61-19), tmp1 /* sigma1 */ _SEP \
65 xorq wt, tmp1; /* sigma1 */ \ 67 xorq wt, tmp1 /* sigma1 */ _SEP \
66 rorq $19, tmp1; /* sigma1 */ \ 68 rorq $19, tmp1 /* sigma1 */ _SEP \
67 shrq $6, wt; /* sigma1 */ \ 69 shrq $6, wt /* sigma1 */ _SEP \
68 xorq tmp1, wt; /* sigma1 */ \ 70 xorq tmp1, wt /* sigma1 */ _SEP \
69 \ 71 \
70 addq (((idx-7)&0xf)*8)(w), wt; /* Wt-7 */ \ 72 addq (((idx-7)&0xf)*8)(w), wt /* Wt-7 */ _SEP \
71 addq (((idx-16)&0xf)*8)(w), wt; /* Wt-16 */ \ 73 addq (((idx-16)&0xf)*8)(w), wt /* Wt-16 */ _SEP \
72 \ 74 \
73 movq (((idx-15)&0xf)*8)(w), tmp2; /* sigma0 */ \ 75 movq (((idx-15)&0xf)*8)(w), tmp2 /* sigma0 */ _SEP \
74 movq tmp2, tmp3; /* sigma0 */ \ 76 movq tmp2, tmp3 /* sigma0 */ _SEP \
75 rorq $(8-1), tmp2; /* sigma0 */ \ 77 rorq $(8-1), tmp2 /* sigma0 */ _SEP \
76 xorq tmp3, tmp2; /* sigma0 */ \ 78 xorq tmp3, tmp2 /* sigma0 */ _SEP \
77 rorq $1, tmp2; /* sigma0 */ \ 79 rorq $1, tmp2 /* sigma0 */ _SEP \
78 shrq $7, tmp3; /* sigma0 */ \ 80 shrq $7, tmp3 /* sigma0 */ _SEP \
79 xorq tmp3, tmp2; /* sigma0 */ \ 81 xorq tmp3, tmp2 /* sigma0 */ _SEP \
80 addq tmp2, wt; /* sigma0 */ \ 82 addq tmp2, wt /* sigma0 */ _SEP \
81 \ 83 \
82 movq wt, ((idx&0xf)*8)(w) 84 movq wt, ((idx&0xf)*8)(w)
83 85
@@ -95,49 +97,49 @@
95 * Upon completion d = d + T1, h = T1 + T2, pending rotation. 97 * Upon completion d = d + T1, h = T1 + T2, pending rotation.
96 */ 98 */
97#define sha512_round(idx, a, b, c, d, e, f, g, h, k, w, wt) \ 99#define sha512_round(idx, a, b, c, d, e, f, g, h, k, w, wt) \
98 addq wt, h; /* T1 Wt */ \ 100 addq wt, h /* T1 Wt */ _SEP \
99 addq (k512, round, 8), h; /* T1 Kt */ \ 101 addq (k512, round, 8), h /* T1 Kt */ _SEP \
100 \ 102 \
101 movq e, tmp1; /* T1 Sigma1 */ \ 103 movq e, tmp1 /* T1 Sigma1 */ _SEP \
102 rorq $(41-18), tmp1; /* T1 Sigma1 */ \ 104 rorq $(41-18), tmp1 /* T1 Sigma1 */ _SEP \
103 xorq e, tmp1; /* T1 Sigma1 */ \ 105 xorq e, tmp1 /* T1 Sigma1 */ _SEP \
104 rorq $(18-14), tmp1; /* T1 Sigma1 */ \ 106 rorq $(18-14), tmp1 /* T1 Sigma1 */ _SEP \
105 xorq e, tmp1; /* T1 Sigma1 */ \ 107 xorq e, tmp1 /* T1 Sigma1 */ _SEP \
106 rorq $14, tmp1; /* T1 Sigma1 */ \ 108 rorq $14, tmp1 /* T1 Sigma1 */ _SEP \
107 addq tmp1, h; /* T1 Sigma1 */ \ 109 addq tmp1, h /* T1 Sigma1 */ _SEP \
108 \ 110 \
109 movq f, tmp2; /* T1 Ch */ \ 111 movq f, tmp2 /* T1 Ch */ _SEP \
110 xorq g, tmp2; /* T1 Ch */ \ 112 xorq g, tmp2 /* T1 Ch */ _SEP \
111 andq e, tmp2; /* T1 Ch */ \ 113 andq e, tmp2 /* T1 Ch */ _SEP \
112 xorq g, tmp2; /* T1 Ch */ \ 114 xorq g, tmp2 /* T1 Ch */ _SEP \
113 addq tmp2, h; /* T1 Ch */ \ 115 addq tmp2, h /* T1 Ch */ _SEP \
114 \ 116 \
115 addq h, d; /* d += T1 */ \ 117 addq h, d /* d += T1 */ _SEP \
116 \ 118 \
117 movq a, tmp1; /* T2 Sigma0 */ \ 119 movq a, tmp1 /* T2 Sigma0 */ _SEP \
118 rorq $(39-34), tmp1; /* T2 Sigma0 */ \ 120 rorq $(39-34), tmp1 /* T2 Sigma0 */ _SEP \
119 xorq a, tmp1; /* T2 Sigma0 */ \ 121 xorq a, tmp1 /* T2 Sigma0 */ _SEP \
120 rorq $(34-28), tmp1; /* T2 Sigma0 */ \ 122 rorq $(34-28), tmp1 /* T2 Sigma0 */ _SEP \
121 xorq a, tmp1; /* T2 Sigma0 */ \ 123 xorq a, tmp1 /* T2 Sigma0 */ _SEP \
122 rorq $28, tmp1; /* T2 Sigma0 */ \ 124 rorq $28, tmp1 /* T2 Sigma0 */ _SEP \
123 addq tmp1, h; /* T2 Sigma0 */ \ 125 addq tmp1, h /* T2 Sigma0 */ _SEP \
124 \ 126 \
125 movq b, tmp2; /* T2 Maj */ \ 127 movq b, tmp2 /* T2 Maj */ _SEP \
126 xorq c, tmp2; /* T2 Maj */ \ 128 xorq c, tmp2 /* T2 Maj */ _SEP \
127 andq a, tmp2; /* T2 Maj */ \ 129 andq a, tmp2 /* T2 Maj */ _SEP \
128 movq b, tmp3; /* T2 Maj */ \ 130 movq b, tmp3 /* T2 Maj */ _SEP \
129 andq c, tmp3; /* T2 Maj */ \ 131 andq c, tmp3 /* T2 Maj */ _SEP \
130 xorq tmp2, tmp3; /* T2 Maj */ \ 132 xorq tmp2, tmp3 /* T2 Maj */ _SEP \
131 addq tmp3, h; /* T2 Maj */ \ 133 addq tmp3, h /* T2 Maj */ _SEP \
132 \ 134 \
133 addq $1, round 135 addq $1, round
134 136
135#define sha512_round_load(idx, a, b, c, d, e, f, g, h) \ 137#define sha512_round_load(idx, a, b, c, d, e, f, g, h) \
136 sha512_message_schedule_load(idx, in, %rsp, tmp0); \ 138 sha512_message_schedule_load(idx, in, %rsp, tmp0) _SEP \
137 sha512_round(idx, a, b, c, d, e, f, g, h, k512, %rsp, tmp0) 139 sha512_round(idx, a, b, c, d, e, f, g, h, k512, %rsp, tmp0)
138 140
139#define sha512_round_update(idx, a, b, c, d, e, f, g, h) \ 141#define sha512_round_update(idx, a, b, c, d, e, f, g, h) \
140 sha512_message_schedule_update(idx, %rsp, tmp0); \ 142 sha512_message_schedule_update(idx, %rsp, tmp0) _SEP \
141 sha512_round(idx, a, b, c, d, e, f, g, h, k512, %rsp, tmp0) 143 sha512_round(idx, a, b, c, d, e, f, g, h, k512, %rsp, tmp0)
142 144
143.section .text 145.section .text
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2026-09-08 13:21:22 +0000