Skip to content

Commit 6803d9b

Browse files
tipabutipabu
committed
Optimize crc64_rocksoft for aarch64
Closes #326 Signed-off-by: Tim Burke <[email protected]>
1 parent 77d76b6 commit 6803d9b

9 files changed

+524
-44
lines changed

‎Release_notes.txt

+4
Original file line numberDiff line numberDiff line change
@@ -150,6 +150,10 @@ v2.32
150150
* RAID improvements:
151151
- Added new RVV xor_gen, pq_gen implementations.
152152

153+
* CRC improvements:
154+
- CRC64 Rocksoft implementation on aarch64 optimized similar to other CRC64
155+
implementations.
156+
153157
v2.31
154158

155159
* API changes:

‎crc/aarch64/Makefile.am

+2-1
Original file line numberDiff line numberDiff line change
@@ -41,7 +41,8 @@ lsrc_aarch64 += \
4141
crc/aarch64/crc64_iso_norm_pmull.S \
4242
crc/aarch64/crc64_jones_refl_pmull.S \
4343
crc/aarch64/crc64_jones_norm_pmull.S \
44-
crc/aarch64/crc64_rocksoft.c
44+
crc/aarch64/crc64_rocksoft_refl_pmull.S \
45+
crc/aarch64/crc64_rocksoft_norm_pmull.S
4546

4647

4748
#CRC32/CRC32C for micro-architecture

‎crc/aarch64/crc64_rocksoft.c

-43
This file was deleted.

‎crc/aarch64/crc64_rocksoft_norm_pmull.S

+34
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,34 @@
1+
########################################################################
2+
# Copyright(c) 2025 Tim Burke All rights reserved.
3+
#
4+
# Redistribution and use in source and binary forms, with or without
5+
# modification, are permitted provided that the following conditions
6+
# are met:
7+
# * Redistributions of source code must retain the above copyright
8+
# notice, this list of conditions and the following disclaimer.
9+
# * Redistributions in binary form must reproduce the above copyright
10+
# notice, this list of conditions and the following disclaimer in
11+
# the documentation and/or other materials provided with the
12+
# distribution.
13+
# * Neither the name of Arm Corporation nor the names of its
14+
# contributors may be used to endorse or promote products derived
15+
# from this software without specific prior written permission.
16+
#
17+
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18+
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19+
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20+
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
21+
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
22+
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
23+
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24+
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25+
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26+
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
27+
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28+
#########################################################################
29+
30+
#include "../include/aarch64_label.h"
31+
#include "crc64_rocksoft_norm_pmull.h"
32+
#include "crc64_norm_common_pmull.h"
33+
34+
crc64_norm_func crc64_rocksoft_norm_pmull

‎crc/aarch64/crc64_rocksoft_norm_pmull.h

+210
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,210 @@
1+
########################################################################
2+
# Copyright(c) 2025 Tim Burke All rights reserved.
3+
#
4+
# Redistribution and use in source and binary forms, with or without
5+
# modification, are permitted provided that the following conditions
6+
# are met:
7+
# * Redistributions of source code must retain the above copyright
8+
# notice, this list of conditions and the following disclaimer.
9+
# * Redistributions in binary form must reproduce the above copyright
10+
# notice, this list of conditions and the following disclaimer in
11+
# the documentation and/or other materials provided with the
12+
# distribution.
13+
# * Neither the name of Arm Corporation nor the names of its
14+
# contributors may be used to endorse or promote products derived
15+
# from this software without specific prior written permission.
16+
#
17+
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18+
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19+
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20+
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
21+
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
22+
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
23+
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24+
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25+
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26+
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
27+
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28+
#########################################################################
29+
30+
# rk15
31+
.equ p4_low_b0, 0x488c
32+
.equ p4_low_b1, 0x0488
33+
.equ p4_low_b2, 0x4e6a
34+
.equ p4_low_b3, 0xb441
35+
# rk16
36+
.equ p4_high_b0, 0x9860
37+
.equ p4_high_b1, 0x9b66
38+
.equ p4_high_b2, 0x30f1
39+
.equ p4_high_b3, 0xa42a
40+
41+
# rk1
42+
.equ p1_low_b0, 0x2f08
43+
.equ p1_low_b1, 0xf0dd
44+
.equ p1_low_b2, 0xc948
45+
.equ p1_low_b3, 0x6b08
46+
# rk2
47+
.equ p1_high_b0, 0x76ae
48+
.equ p1_high_b1, 0x7f04
49+
.equ p1_high_b2, 0x8ba9
50+
.equ p1_high_b3, 0x0857
51+
52+
# rk1
53+
.equ p0_low_b0, 0x2f08
54+
.equ p0_low_b1, 0xf0dd
55+
.equ p0_low_b2, 0xc948
56+
.equ p0_low_b3, 0x6b08
57+
.equ p0_high_b0, 0x0000
58+
.equ p0_high_b1, 0x0000
59+
.equ p0_high_b2, 0x0000
60+
.equ p0_high_b3, 0x0000
61+
62+
# rk7
63+
.equ br_low_b0, 0x6fc8
64+
.equ br_low_b1, 0x98be
65+
.equ br_low_b2, 0xeeb2
66+
.equ br_low_b3, 0xddf3
67+
# rk8
68+
.equ br_high_b0, 0x3659
69+
.equ br_high_b1, 0x94c9
70+
.equ br_high_b2, 0xd235
71+
.equ br_high_b3, 0xad93
72+
73+
.text
74+
ASM_DEF_RODATA
75+
.align 4
76+
.set .lanchor_crc_tab,. + 0
77+
#ifndef __APPLE__
78+
.type crc64_tab, %object
79+
.size crc64_tab, 2048
80+
#endif
81+
82+
crc64_tab:
83+
.xword 0x0000000000000000, 0xad93d23594c93659
84+
.xword 0xf6b4765ebd5b5aeb, 0x5b27a46b29926cb2
85+
.xword 0x40fb3e88ee7f838f, 0xed68ecbd7ab6b5d6
86+
.xword 0xb64f48d65324d964, 0x1bdc9ae3c7edef3d
87+
.xword 0x81f67d11dcff071e, 0x2c65af2448363147
88+
.xword 0x77420b4f61a45df5, 0xdad1d97af56d6bac
89+
.xword 0xc10d439932808491, 0x6c9e91aca649b2c8
90+
.xword 0x37b935c78fdbde7a, 0x9a2ae7f21b12e823
91+
.xword 0xae7f28162d373865, 0x03ecfa23b9fe0e3c
92+
.xword 0x58cb5e48906c628e, 0xf5588c7d04a554d7
93+
.xword 0xee84169ec348bbea, 0x4317c4ab57818db3
94+
.xword 0x183060c07e13e101, 0xb5a3b2f5eadad758
95+
.xword 0x2f895507f1c83f7b, 0x821a873265010922
96+
.xword 0xd93d23594c936590, 0x74aef16cd85a53c9
97+
.xword 0x6f726b8f1fb7bcf4, 0xc2e1b9ba8b7e8aad
98+
.xword 0x99c61dd1a2ece61f, 0x3455cfe43625d046
99+
.xword 0xf16d8219cea74693, 0x5cfe502c5a6e70ca
100+
.xword 0x07d9f44773fc1c78, 0xaa4a2672e7352a21
101+
.xword 0xb196bc9120d8c51c, 0x1c056ea4b411f345
102+
.xword 0x4722cacf9d839ff7, 0xeab118fa094aa9ae
103+
.xword 0x709bff081258418d, 0xdd082d3d869177d4
104+
.xword 0x862f8956af031b66, 0x2bbc5b633bca2d3f
105+
.xword 0x3060c180fc27c202, 0x9df313b568eef45b
106+
.xword 0xc6d4b7de417c98e9, 0x6b4765ebd5b5aeb0
107+
.xword 0x5f12aa0fe3907ef6, 0xf281783a775948af
108+
.xword 0xa9a6dc515ecb241d, 0x04350e64ca021244
109+
.xword 0x1fe994870deffd79, 0xb27a46b29926cb20
110+
.xword 0xe95de2d9b0b4a792, 0x44ce30ec247d91cb
111+
.xword 0xdee4d71e3f6f79e8, 0x7377052baba64fb1
112+
.xword 0x2850a14082342303, 0x85c3737516fd155a
113+
.xword 0x9e1fe996d110fa67, 0x338c3ba345d9cc3e
114+
.xword 0x68ab9fc86c4ba08c, 0xc5384dfdf88296d5
115+
.xword 0x4f48d6060987bb7f, 0xe2db04339d4e8d26
116+
.xword 0xb9fca058b4dce194, 0x146f726d2015d7cd
117+
.xword 0x0fb3e88ee7f838f0, 0xa2203abb73310ea9
118+
.xword 0xf9079ed05aa3621b, 0x54944ce5ce6a5442
119+
.xword 0xcebeab17d578bc61, 0x632d792241b18a38
120+
.xword 0x380add496823e68a, 0x95990f7cfcead0d3
121+
.xword 0x8e45959f3b073fee, 0x23d647aaafce09b7
122+
.xword 0x78f1e3c1865c6505, 0xd56231f41295535c
123+
.xword 0xe137fe1024b0831a, 0x4ca42c25b079b543
124+
.xword 0x1783884e99ebd9f1, 0xba105a7b0d22efa8
125+
.xword 0xa1ccc098cacf0095, 0x0c5f12ad5e0636cc
126+
.xword 0x5778b6c677945a7e, 0xfaeb64f3e35d6c27
127+
.xword 0x60c18301f84f8404, 0xcd5251346c86b25d
128+
.xword 0x9675f55f4514deef, 0x3be6276ad1dde8b6
129+
.xword 0x203abd891630078b, 0x8da96fbc82f931d2
130+
.xword 0xd68ecbd7ab6b5d60, 0x7b1d19e23fa26b39
131+
.xword 0xbe25541fc720fdec, 0x13b6862a53e9cbb5
132+
.xword 0x489122417a7ba707, 0xe502f074eeb2915e
133+
.xword 0xfede6a97295f7e63, 0x534db8a2bd96483a
134+
.xword 0x086a1cc994042488, 0xa5f9cefc00cd12d1
135+
.xword 0x3fd3290e1bdffaf2, 0x9240fb3b8f16ccab
136+
.xword 0xc9675f50a684a019, 0x64f48d65324d9640
137+
.xword 0x7f281786f5a0797d, 0xd2bbc5b361694f24
138+
.xword 0x899c61d848fb2396, 0x240fb3eddc3215cf
139+
.xword 0x105a7c09ea17c589, 0xbdc9ae3c7edef3d0
140+
.xword 0xe6ee0a57574c9f62, 0x4b7dd862c385a93b
141+
.xword 0x50a1428104684606, 0xfd3290b490a1705f
142+
.xword 0xa61534dfb9331ced, 0x0b86e6ea2dfa2ab4
143+
.xword 0x91ac011836e8c297, 0x3c3fd32da221f4ce
144+
.xword 0x671877468bb3987c, 0xca8ba5731f7aae25
145+
.xword 0xd1573f90d8974118, 0x7cc4eda54c5e7741
146+
.xword 0x27e349ce65cc1bf3, 0x8a709bfbf1052daa
147+
.xword 0x9e91ac0c130f76fe, 0x33027e3987c640a7
148+
.xword 0x6825da52ae542c15, 0xc5b608673a9d1a4c
149+
.xword 0xde6a9284fd70f571, 0x73f940b169b9c328
150+
.xword 0x28dee4da402baf9a, 0x854d36efd4e299c3
151+
.xword 0x1f67d11dcff071e0, 0xb2f403285b3947b9
152+
.xword 0xe9d3a74372ab2b0b, 0x44407576e6621d52
153+
.xword 0x5f9cef95218ff26f, 0xf20f3da0b546c436
154+
.xword 0xa92899cb9cd4a884, 0x04bb4bfe081d9edd
155+
.xword 0x30ee841a3e384e9b, 0x9d7d562faaf178c2
156+
.xword 0xc65af24483631470, 0x6bc9207117aa2229
157+
.xword 0x7015ba92d047cd14, 0xdd8668a7448efb4d
158+
.xword 0x86a1cccc6d1c97ff, 0x2b321ef9f9d5a1a6
159+
.xword 0xb118f90be2c74985, 0x1c8b2b3e760e7fdc
160+
.xword 0x47ac8f555f9c136e, 0xea3f5d60cb552537
161+
.xword 0xf1e3c7830cb8ca0a, 0x5c7015b69871fc53
162+
.xword 0x0757b1ddb1e390e1, 0xaac463e8252aa6b8
163+
.xword 0x6ffc2e15dda8306d, 0xc26ffc2049610634
164+
.xword 0x9948584b60f36a86, 0x34db8a7ef43a5cdf
165+
.xword 0x2f07109d33d7b3e2, 0x8294c2a8a71e85bb
166+
.xword 0xd9b366c38e8ce909, 0x7420b4f61a45df50
167+
.xword 0xee0a530401573773, 0x43998131959e012a
168+
.xword 0x18be255abc0c6d98, 0xb52df76f28c55bc1
169+
.xword 0xaef16d8cef28b4fc, 0x0362bfb97be182a5
170+
.xword 0x58451bd25273ee17, 0xf5d6c9e7c6bad84e
171+
.xword 0xc1830603f09f0808, 0x6c10d43664563e51
172+
.xword 0x3737705d4dc452e3, 0x9aa4a268d90d64ba
173+
.xword 0x8178388b1ee08b87, 0x2cebeabe8a29bdde
174+
.xword 0x77cc4ed5a3bbd16c, 0xda5f9ce03772e735
175+
.xword 0x40757b122c600f16, 0xede6a927b8a9394f
176+
.xword 0xb6c10d4c913b55fd, 0x1b52df7905f263a4
177+
.xword 0x008e459ac21f8c99, 0xad1d97af56d6bac0
178+
.xword 0xf63a33c47f44d672, 0x5ba9e1f1eb8de02b
179+
.xword 0xd1d97a0a1a88cd81, 0x7c4aa83f8e41fbd8
180+
.xword 0x276d0c54a7d3976a, 0x8afede61331aa133
181+
.xword 0x91224482f4f74e0e, 0x3cb196b7603e7857
182+
.xword 0x679632dc49ac14e5, 0xca05e0e9dd6522bc
183+
.xword 0x502f071bc677ca9f, 0xfdbcd52e52befcc6
184+
.xword 0xa69b71457b2c9074, 0x0b08a370efe5a62d
185+
.xword 0x10d4399328084910, 0xbd47eba6bcc17f49
186+
.xword 0xe6604fcd955313fb, 0x4bf39df8019a25a2
187+
.xword 0x7fa6521c37bff5e4, 0xd2358029a376c3bd
188+
.xword 0x891224428ae4af0f, 0x2481f6771e2d9956
189+
.xword 0x3f5d6c94d9c0766b, 0x92cebea14d094032
190+
.xword 0xc9e91aca649b2c80, 0x647ac8fff0521ad9
191+
.xword 0xfe502f0deb40f2fa, 0x53c3fd387f89c4a3
192+
.xword 0x08e45953561ba811, 0xa5778b66c2d29e48
193+
.xword 0xbeab1185053f7175, 0x1338c3b091f6472c
194+
.xword 0x481f67dbb8642b9e, 0xe58cb5ee2cad1dc7
195+
.xword 0x20b4f813d42f8b12, 0x8d272a2640e6bd4b
196+
.xword 0xd6008e4d6974d1f9, 0x7b935c78fdbde7a0
197+
.xword 0x604fc69b3a50089d, 0xcddc14aeae993ec4
198+
.xword 0x96fbb0c5870b5276, 0x3b6862f013c2642f
199+
.xword 0xa142850208d08c0c, 0x0cd157379c19ba55
200+
.xword 0x57f6f35cb58bd6e7, 0xfa6521692142e0be
201+
.xword 0xe1b9bb8ae6af0f83, 0x4c2a69bf726639da
202+
.xword 0x170dcdd45bf45568, 0xba9e1fe1cf3d6331
203+
.xword 0x8ecbd005f918b377, 0x235802306dd1852e
204+
.xword 0x787fa65b4443e99c, 0xd5ec746ed08adfc5
205+
.xword 0xce30ee8d176730f8, 0x63a33cb883ae06a1
206+
.xword 0x388498d3aa3c6a13, 0x95174ae63ef55c4a
207+
.xword 0x0f3dad1425e7b469, 0xa2ae7f21b12e8230
208+
.xword 0xf989db4a98bcee82, 0x541a097f0c75d8db
209+
.xword 0x4fc6939ccb9837e6, 0xe25541a95f5101bf
210+
.xword 0xb972e5c276c36d0d, 0x14e137f7e20a5b54

‎crc/aarch64/crc64_rocksoft_refl_pmull.S

+34
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,34 @@
1+
########################################################################
2+
# Copyright(c) 2025 Tim Burke All rights reserved.
3+
#
4+
# Redistribution and use in source and binary forms, with or without
5+
# modification, are permitted provided that the following conditions
6+
# are met:
7+
# * Redistributions of source code must retain the above copyright
8+
# notice, this list of conditions and the following disclaimer.
9+
# * Redistributions in binary form must reproduce the above copyright
10+
# notice, this list of conditions and the following disclaimer in
11+
# the documentation and/or other materials provided with the
12+
# distribution.
13+
# * Neither the name of Arm Corporation nor the names of its
14+
# contributors may be used to endorse or promote products derived
15+
# from this software without specific prior written permission.
16+
#
17+
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18+
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19+
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20+
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
21+
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
22+
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
23+
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24+
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25+
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26+
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
27+
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28+
#########################################################################
29+
30+
#include "../include/aarch64_label.h"
31+
#include "crc64_rocksoft_refl_pmull.h"
32+
#include "crc64_refl_common_pmull.h"
33+
34+
crc64_refl_func crc64_rocksoft_refl_pmull

0 commit comments

Comments
 (0)