1// This file is generated from a similarly-named Perl script in the BoringSSL
2// source tree. Do not edit by hand.
3
4#if defined(__has_feature)
5#if __has_feature(memory_sanitizer) && !defined(OPENSSL_NO_ASM)
6#define OPENSSL_NO_ASM
7#endif
8#endif
9
10#if !defined(OPENSSL_NO_ASM)
11#if defined(__aarch64__)
12#if defined(BORINGSSL_PREFIX)
13#include <boringssl_prefix_symbols_asm.h>
14#endif
15#include <openssl/arm_arch.h>
16
17.text
18.arch	armv8-a+crypto
19.globl	gcm_init_v8
20.hidden	gcm_init_v8
21.type	gcm_init_v8,%function
22.align	4
23gcm_init_v8:
24	ld1	{v17.2d},[x1]		//load input H
25	movi	v19.16b,#0xe1
26	shl	v19.2d,v19.2d,#57		//0xc2.0
27	ext	v3.16b,v17.16b,v17.16b,#8
28	ushr	v18.2d,v19.2d,#63
29	dup	v17.4s,v17.s[1]
30	ext	v16.16b,v18.16b,v19.16b,#8		//t0=0xc2....01
31	ushr	v18.2d,v3.2d,#63
32	sshr	v17.4s,v17.4s,#31		//broadcast carry bit
33	and	v18.16b,v18.16b,v16.16b
34	shl	v3.2d,v3.2d,#1
35	ext	v18.16b,v18.16b,v18.16b,#8
36	and	v16.16b,v16.16b,v17.16b
37	orr	v3.16b,v3.16b,v18.16b		//H<<<=1
38	eor	v20.16b,v3.16b,v16.16b		//twisted H
39	st1	{v20.2d},[x0],#16		//store Htable[0]
40
41	//calculate H^2
42	ext	v16.16b,v20.16b,v20.16b,#8		//Karatsuba pre-processing
43	pmull	v0.1q,v20.1d,v20.1d
44	eor	v16.16b,v16.16b,v20.16b
45	pmull2	v2.1q,v20.2d,v20.2d
46	pmull	v1.1q,v16.1d,v16.1d
47
48	ext	v17.16b,v0.16b,v2.16b,#8		//Karatsuba post-processing
49	eor	v18.16b,v0.16b,v2.16b
50	eor	v1.16b,v1.16b,v17.16b
51	eor	v1.16b,v1.16b,v18.16b
52	pmull	v18.1q,v0.1d,v19.1d		//1st phase
53
54	ins	v2.d[0],v1.d[1]
55	ins	v1.d[1],v0.d[0]
56	eor	v0.16b,v1.16b,v18.16b
57
58	ext	v18.16b,v0.16b,v0.16b,#8		//2nd phase
59	pmull	v0.1q,v0.1d,v19.1d
60	eor	v18.16b,v18.16b,v2.16b
61	eor	v22.16b,v0.16b,v18.16b
62
63	ext	v17.16b,v22.16b,v22.16b,#8		//Karatsuba pre-processing
64	eor	v17.16b,v17.16b,v22.16b
65	ext	v21.16b,v16.16b,v17.16b,#8		//pack Karatsuba pre-processed
66	st1	{v21.2d,v22.2d},[x0]		//store Htable[1..2]
67
68	ret
69.size	gcm_init_v8,.-gcm_init_v8
70.globl	gcm_gmult_v8
71.hidden	gcm_gmult_v8
72.type	gcm_gmult_v8,%function
73.align	4
74gcm_gmult_v8:
75	ld1	{v17.2d},[x0]		//load Xi
76	movi	v19.16b,#0xe1
77	ld1	{v20.2d,v21.2d},[x1]	//load twisted H, ...
78	shl	v19.2d,v19.2d,#57
79#ifndef __ARMEB__
80	rev64	v17.16b,v17.16b
81#endif
82	ext	v3.16b,v17.16b,v17.16b,#8
83
84	pmull	v0.1q,v20.1d,v3.1d		//H.lo·Xi.lo
85	eor	v17.16b,v17.16b,v3.16b		//Karatsuba pre-processing
86	pmull2	v2.1q,v20.2d,v3.2d		//H.hXi.hi
87	pmull	v1.1q,v21.1d,v17.1d		//(H.lo+H.hi)·(Xi.lo+Xi.hi)
88
89	ext	v17.16b,v0.16b,v2.16b,#8		//Karatsuba post-processing
90	eor	v18.16b,v0.16b,v2.16b
91	eor	v1.16b,v1.16b,v17.16b
92	eor	v1.16b,v1.16b,v18.16b
93	pmull	v18.1q,v0.1d,v19.1d		//1st phase of reduction
94
95	ins	v2.d[0],v1.d[1]
96	ins	v1.d[1],v0.d[0]
97	eor	v0.16b,v1.16b,v18.16b
98
99	ext	v18.16b,v0.16b,v0.16b,#8		//2nd phase of reduction
100	pmull	v0.1q,v0.1d,v19.1d
101	eor	v18.16b,v18.16b,v2.16b
102	eor	v0.16b,v0.16b,v18.16b
103
104#ifndef __ARMEB__
105	rev64	v0.16b,v0.16b
106#endif
107	ext	v0.16b,v0.16b,v0.16b,#8
108	st1	{v0.2d},[x0]		//write out Xi
109
110	ret
111.size	gcm_gmult_v8,.-gcm_gmult_v8
112.globl	gcm_ghash_v8
113.hidden	gcm_ghash_v8
114.type	gcm_ghash_v8,%function
115.align	4
116gcm_ghash_v8:
117	ld1	{v0.2d},[x0]		//load [rotated] Xi
118						//"[rotated]" means that
119						//loaded value would have
120						//to be rotated in order to
121						//make it appear as in
122						//algorithm specification
123	subs	x3,x3,#32		//see if x3 is 32 or larger
124	mov	x12,#16		//x12 is used as post-
125						//increment for input pointer;
126						//as loop is modulo-scheduled
127						//x12 is zeroed just in time
128						//to preclude overstepping
129						//inp[len], which means that
130						//last block[s] are actually
131						//loaded twice, but last
132						//copy is not processed
133	ld1	{v20.2d,v21.2d},[x1],#32	//load twisted H, ..., H^2
134	movi	v19.16b,#0xe1
135	ld1	{v22.2d},[x1]
136	csel	x12,xzr,x12,eq			//is it time to zero x12?
137	ext	v0.16b,v0.16b,v0.16b,#8		//rotate Xi
138	ld1	{v16.2d},[x2],#16	//load [rotated] I[0]
139	shl	v19.2d,v19.2d,#57		//compose 0xc2.0 constant
140#ifndef __ARMEB__
141	rev64	v16.16b,v16.16b
142	rev64	v0.16b,v0.16b
143#endif
144	ext	v3.16b,v16.16b,v16.16b,#8		//rotate I[0]
145	b.lo	.Lodd_tail_v8		//x3 was less than 32
146	ld1	{v17.2d},[x2],x12	//load [rotated] I[1]
147#ifndef __ARMEB__
148	rev64	v17.16b,v17.16b
149#endif
150	ext	v7.16b,v17.16b,v17.16b,#8
151	eor	v3.16b,v3.16b,v0.16b		//I[i]^=Xi
152	pmull	v4.1q,v20.1d,v7.1d		//H·Ii+1
153	eor	v17.16b,v17.16b,v7.16b		//Karatsuba pre-processing
154	pmull2	v6.1q,v20.2d,v7.2d
155	b	.Loop_mod2x_v8
156
157.align	4
158.Loop_mod2x_v8:
159	ext	v18.16b,v3.16b,v3.16b,#8
160	subs	x3,x3,#32		//is there more data?
161	pmull	v0.1q,v22.1d,v3.1d		//H^2.lo·Xi.lo
162	csel	x12,xzr,x12,lo			//is it time to zero x12?
163
164	pmull	v5.1q,v21.1d,v17.1d
165	eor	v18.16b,v18.16b,v3.16b		//Karatsuba pre-processing
166	pmull2	v2.1q,v22.2d,v3.2d		//H^2.hi·Xi.hi
167	eor	v0.16b,v0.16b,v4.16b		//accumulate
168	pmull2	v1.1q,v21.2d,v18.2d		//(H^2.lo+H^2.hi)·(Xi.lo+Xi.hi)
169	ld1	{v16.2d},[x2],x12	//load [rotated] I[i+2]
170
171	eor	v2.16b,v2.16b,v6.16b
172	csel	x12,xzr,x12,eq			//is it time to zero x12?
173	eor	v1.16b,v1.16b,v5.16b
174
175	ext	v17.16b,v0.16b,v2.16b,#8		//Karatsuba post-processing
176	eor	v18.16b,v0.16b,v2.16b
177	eor	v1.16b,v1.16b,v17.16b
178	ld1	{v17.2d},[x2],x12	//load [rotated] I[i+3]
179#ifndef __ARMEB__
180	rev64	v16.16b,v16.16b
181#endif
182	eor	v1.16b,v1.16b,v18.16b
183	pmull	v18.1q,v0.1d,v19.1d		//1st phase of reduction
184
185#ifndef __ARMEB__
186	rev64	v17.16b,v17.16b
187#endif
188	ins	v2.d[0],v1.d[1]
189	ins	v1.d[1],v0.d[0]
190	ext	v7.16b,v17.16b,v17.16b,#8
191	ext	v3.16b,v16.16b,v16.16b,#8
192	eor	v0.16b,v1.16b,v18.16b
193	pmull	v4.1q,v20.1d,v7.1d		//H·Ii+1
194	eor	v3.16b,v3.16b,v2.16b		//accumulate v3.16b early
195
196	ext	v18.16b,v0.16b,v0.16b,#8		//2nd phase of reduction
197	pmull	v0.1q,v0.1d,v19.1d
198	eor	v3.16b,v3.16b,v18.16b
199	eor	v17.16b,v17.16b,v7.16b		//Karatsuba pre-processing
200	eor	v3.16b,v3.16b,v0.16b
201	pmull2	v6.1q,v20.2d,v7.2d
202	b.hs	.Loop_mod2x_v8		//there was at least 32 more bytes
203
204	eor	v2.16b,v2.16b,v18.16b
205	ext	v3.16b,v16.16b,v16.16b,#8		//re-construct v3.16b
206	adds	x3,x3,#32		//re-construct x3
207	eor	v0.16b,v0.16b,v2.16b		//re-construct v0.16b
208	b.eq	.Ldone_v8		//is x3 zero?
209.Lodd_tail_v8:
210	ext	v18.16b,v0.16b,v0.16b,#8
211	eor	v3.16b,v3.16b,v0.16b		//inp^=Xi
212	eor	v17.16b,v16.16b,v18.16b		//v17.16b is rotated inp^Xi
213
214	pmull	v0.1q,v20.1d,v3.1d		//H.lo·Xi.lo
215	eor	v17.16b,v17.16b,v3.16b		//Karatsuba pre-processing
216	pmull2	v2.1q,v20.2d,v3.2d		//H.hXi.hi
217	pmull	v1.1q,v21.1d,v17.1d		//(H.lo+H.hi)·(Xi.lo+Xi.hi)
218
219	ext	v17.16b,v0.16b,v2.16b,#8		//Karatsuba post-processing
220	eor	v18.16b,v0.16b,v2.16b
221	eor	v1.16b,v1.16b,v17.16b
222	eor	v1.16b,v1.16b,v18.16b
223	pmull	v18.1q,v0.1d,v19.1d		//1st phase of reduction
224
225	ins	v2.d[0],v1.d[1]
226	ins	v1.d[1],v0.d[0]
227	eor	v0.16b,v1.16b,v18.16b
228
229	ext	v18.16b,v0.16b,v0.16b,#8		//2nd phase of reduction
230	pmull	v0.1q,v0.1d,v19.1d
231	eor	v18.16b,v18.16b,v2.16b
232	eor	v0.16b,v0.16b,v18.16b
233
234.Ldone_v8:
235#ifndef __ARMEB__
236	rev64	v0.16b,v0.16b
237#endif
238	ext	v0.16b,v0.16b,v0.16b,#8
239	st1	{v0.2d},[x0]		//write out Xi
240
241	ret
242.size	gcm_ghash_v8,.-gcm_ghash_v8
243.byte	71,72,65,83,72,32,102,111,114,32,65,82,77,118,56,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0
244.align	2
245.align	2
246#endif
247#endif  // !OPENSSL_NO_ASM
248