nettle-bugs

nettle-bugs@lists.lysator.liu.se

2 participants
2197 discussions

[S390x] Optimize memxor
by Maamoun TK 22 Jul '21

22 Jul '21

I made a merge request !30 <https://git.lysator.liu.se/nettle/nettle/-/merge_requests/30> that optimizes memxor function for S390x architecture. I've attached a benchmark on a Z15 processor in the MR description. regards, Mamone

1 1

Fwd: [S390x] Optimize GHASH
by Maamoun TK 17 Jul '21

17 Jul '21

---------- Forwarded message --------- From: Maamoun TK <maamoun.tk(a)googlemail.com> Date: Sat, Jul 17, 2021 at 2:00 PM Subject: Re: [S390x] Optimize GHASH To: Niels Möller <nisse(a)lysator.liu.se> On Sat, Jul 17, 2021 at 1:37 PM Niels Möller <nisse(a)lysator.liu.se> wrote: > Maamoun TK <maamoun.tk(a)googlemail.com> writes: > > In another topic, I've optimized the SHA-512 algorithm for arm64 > > architecture but it turned out all CFarm variants don't support SHA-512 > > crypto extension so I can't do any performance or correctness testing for > > now. Do you know any CFarm alternative that supports SHA-512 and SHA3 > > extensions for arm64 architectures? > > Can you do correctness tests on qemu? (I've been using a crosscompiler > and qemu-user to test other ARM code, and that's also what the ci tests > do). > Sure, but I'm also concerned about the performance testing. I was always trying to get the best performance out of hardware-accelerated cores. > I have access to the systems listed on > https://gmplib.org/devel/testsystems, is any of those applicable? The > arm64 machines available includes one Cortex-A73 and one Apple M1. > Cortex-A73 doesn't support SHA-512 and SHA3 List of ARM microarchitectures - Wikipedia <https://en.wikipedia.org/wiki/List_of_ARM_microarchitectures> Cortex-A75 is the minimum architecture that has a full implementation of ARMv8.2-A which is the arch that introduced SHA-512 and SHA3 AArch64 Options (Using the GNU Compiler Collection (GCC)) <https://gcc.gnu.org/onlinedocs/gcc/AArch64-Options.html> Cortex-A55 should support those extensions according to the list.

1 0

Re: [Aarch64] Optimize AES
by Maamoun TK 17 Jul '21

17 Jul '21

On Sat, Jul 17, 2021 at 1:46 PM Niels Möller <nisse(a)lysator.liu.se> wrote: > Maamoun TK <maamoun.tk(a)googlemail.com> writes: > > > It fits better to have one implementation per key size, I'll modify this > > patch to apply this approach. However, we need to merge the s390x branch > at > > first since it has a separate implementation for each key size. > > Right. With fat support in, is there anything else that is essential > before merging? > No, AES and GHASH have the full functionalities for S390x architecture in this way.

1 0

[S390x] Optimize GHASH
by Maamoun TK 17 Jul '21

17 Jul '21

I made a merge request !26 <https://git.lysator.liu.se/nettle/nettle/-/merge_requests/26> that optimizes the GHASH algorithm for S390x architecture. I've attached a benchmark in the description of merge request that describes the improvement of using GHASH accelerator over C implementation. I've also made two patches for fat build support of AES and GHASH for S390x architecture in addition to optimize memxor function using 'xc (xor storage-to-storage) instruction' Files · s390x-fat · Maamoun TK / nettle · GitLab (liu.se) <https://git.lysator.liu.se/mamonet/nettle/-/tree/s390x-fat> Files · s390x-memxor · Maamoun TK / nettle · GitLab (liu.se) <https://git.lysator.liu.se/mamonet/nettle/-/tree/s390x-memxor> I'll make merge requests for both patches after the current one being merged since they need to rebase on top of that patch. regards, Mamone

3 14

[Aarch64] Optimize AES
by Maamoun TK 17 Jul '21

17 Jul '21

This patch optimizes nettle_aes_encrypt() and nettle_aes_decrypt() functions for arm64 architecture, it takes advantage of 'aese' and 'aesmc' instructions to optimize the encryption function and 'aesd' and 'aesimc' to optimize the decryption function. The patch passes the testsuite of nettle. I also run the benchmark on gcc117 instance of CFarm by configuring the library with "--disable-fat --enable-arm64-crypto" options then executing examples/nettle-benchmark: aes128 ECB encrypt 2522.67 aes128 ECB decrypt 2522.53 aes192 ECB encrypt 2165.06 aes192 ECB decrypt 2165.04 aes256 ECB encrypt 1866.80 aes256 ECB decrypt 1866.38 openssl aes128 ECB encrypt 1043.52 openssl aes128 ECB decrypt 1043.05 openssl aes192 ECB encrypt 904.08 openssl aes192 ECB decrypt 903.85 openssl aes256 ECB encrypt 787.43 openssl aes256 ECB decrypt 787.20 gcm_aes128 encrypt 955.10 gcm_aes128 decrypt 955.06 gcm_aes128 update 3269.18 gcm_aes192 encrypt 896.26 gcm_aes192 decrypt 896.46 gcm_aes192 update 3270.24 gcm_aes256 encrypt 840.17 gcm_aes256 decrypt 843.53 gcm_aes256 update 3270.08 openssl gcm_aes128 encrypt 894.51 openssl gcm_aes128 decrypt 899.05 openssl gcm_aes128 update 1636.61 openssl gcm_aes192 encrypt 834.94 openssl gcm_aes192 decrypt 841.99 openssl gcm_aes192 update 1631.40 openssl gcm_aes256 encrypt 788.48 openssl gcm_aes256 decrypt 791.31 openssl gcm_aes256 update 1635.18 I'm a little suspicious about the benchmark numbers because as I remember the performance of gcm update doesn't double the openssl number, I repeat running the process but kept giving the same performance margin. --- arm64/crypto/aes-decrypt-internal.asm | 223 ++++++++++++++++++++++++++++++++++ arm64/crypto/aes-encrypt-internal.asm | 223 ++++++++++++++++++++++++++++++++++ 2 files changed, 446 insertions(+) create mode 100644 arm64/crypto/aes-decrypt-internal.asm create mode 100644 arm64/crypto/aes-encrypt-internal.asm diff --git a/arm64/crypto/aes-decrypt-internal.asm b/arm64/crypto/aes-decrypt-internal.asm new file mode 100644 index 00000000..4bfdb314 --- /dev/null +++ b/arm64/crypto/aes-decrypt-internal.asm @@ -0,0 +1,223 @@ +C arm64/crypto/aes-decrypt-internal.asm + +ifelse(` + Copyright (C) 2021 Mamone Tarsha + This file is part of GNU Nettle. + + GNU Nettle is free software: you can redistribute it and/or + modify it under the terms of either: + + * the GNU Lesser General Public License as published by the Free + Software Foundation; either version 3 of the License, or (at your + option) any later version. + + or + + * the GNU General Public License as published by the Free + Software Foundation; either version 2 of the License, or (at your + option) any later version. + + or both in parallel, as here. + + GNU Nettle is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received copies of the GNU General Public License and + the GNU Lesser General Public License along with this program. If + not, see http://www.gnu.org/licenses/. +') + +.file "aes-decrypt-internal.asm" +.arch armv8-a+crypto + +.text + +C Register usage: + +define(`ROUNDS', `x0') +define(`KEYS', `x1') +define(`LENGTH', `x3') +define(`DST', `x4') +define(`SRC', `x5') + +define(`S0', `v0') +define(`S1', `v1') +define(`S2', `v2') +define(`S3', `v3') +define(`K0', `v16') +define(`K1', `v17') +define(`K2', `v18') +define(`K3', `v19') +define(`K4', `v20') +define(`K5', `v21') +define(`K6', `v22') +define(`K7', `v23') +define(`K8', `v24') +define(`K9', `v25') +define(`K10', `v26') +define(`K11', `v27') +define(`K12', `v28') +define(`K13', `v29') +define(`K14', `v30') + +C AES_ROUND_4B(KEY) +define(`AES_ROUND_4B', m4_assert_numargs(1)` + aesd S0.16b,$1.16b + aesimc S0.16b,S0.16b + aesd S1.16b,$1.16b + aesimc S1.16b,S1.16b + aesd S2.16b,$1.16b + aesimc S2.16b,S2.16b + aesd S3.16b,$1.16b + aesimc S3.16b,S3.16b +') + +C AES_LAST_ROUND_4B(KEY) +define(`AES_LAST_ROUND_4B', m4_assert_numargs(1)` + aesd S0.16b,$1.16b + eor S0.16b,S0.16b,K14.16b + aesd S1.16b,$1.16b + eor S1.16b,S1.16b,K14.16b + aesd S2.16b,$1.16b + eor S2.16b,S2.16b,K14.16b + aesd S3.16b,$1.16b + eor S3.16b,S3.16b,K14.16b +') + +C AES_ROUND_1B(KEY) +define(`AES_ROUND_1B', m4_assert_numargs(1)` + aesd S0.16b,$1.16b + aesimc S0.16b,S0.16b +') + +C AES_LAST_ROUND_1B(KEY) +define(`AES_LAST_ROUND_1B', m4_assert_numargs(1)` + aesd S0.16b,$1.16b + eor S0.16b,S0.16b,K14.16b +') + +C _aes_decrypt(unsigned rounds, const uint32_t *keys, +C const struct aes_table *T, +C size_t length, uint8_t *dst, +C const uint8_t *src) + +PROLOGUE(_nettle_aes_decrypt) + ands x6,LENGTH,#-64 + b.eq L1B + + mov x7,KEYS + ld1 {K0.4s,K1.4s,K2.4s,K3.4s},[x7],#64 + ld1 {K4.4s,K5.4s,K6.4s,K7.4s},[x7],#64 + ld1 {K8.4s,K9.4s},[x7],#32 + cmp ROUNDS,#10 + b.eq L4B_last_key + ld1 {K10.4s,K11.4s},[x7],#32 + cmp ROUNDS,#12 + b.eq L4B_last_key + ld1 {K12.4s,K13.4s},[x7],#32 + +L4B_last_key: + ld1 {K14.4s},[x7] + +L4B_loop: + ld1 {S0.16b,S1.16b,S2.16b,S3.16b},[SRC],#64 + + AES_ROUND_4B(K0) + AES_ROUND_4B(K1) + AES_ROUND_4B(K2) + AES_ROUND_4B(K3) + AES_ROUND_4B(K4) + AES_ROUND_4B(K5) + AES_ROUND_4B(K6) + AES_ROUND_4B(K7) + AES_ROUND_4B(K8) + cmp ROUNDS,#10 + b.eq L4B_10_round + cmp ROUNDS,#12 + b.eq L4B_12_round + b L4B_14_round + +L4B_10_round: + AES_LAST_ROUND_4B(K9) + b L4B_done +L4B_12_round: + AES_ROUND_4B(K9) + AES_ROUND_4B(K10) + AES_LAST_ROUND_4B(K11) + b L4B_done +L4B_14_round: + AES_ROUND_4B(K9) + AES_ROUND_4B(K10) + AES_ROUND_4B(K11) + AES_ROUND_4B(K12) + AES_LAST_ROUND_4B(K13) + +L4B_done: + st1 {S0.16b,S1.16b,S2.16b,S3.16b},[DST],#64 + + subs x6,x6,#64 + b.ne L4B_loop + + and LENGTH,LENGTH,#63 + +L1B: + cbz LENGTH,Ldone + + mov x6,KEYS + ld1 {K0.4s,K1.4s,K2.4s,K3.4s},[x6],#64 + ld1 {K4.4s,K5.4s,K6.4s,K7.4s},[x6],#64 + ld1 {K8.4s,K9.4s},[x6],#32 + cmp ROUNDS,#10 + b.eq L1B_last_key + ld1 {K10.4s,K11.4s},[x6],#32 + cmp ROUNDS,#12 + b.eq L1B_last_key + ld1 {K12.4s,K13.4s},[x6],#32 + +L1B_last_key: + ld1 {K14.4s},[x6] + +L1B_loop: + ld1 {S0.16b},[SRC],#16 + + AES_ROUND_1B(K0) + AES_ROUND_1B(K1) + AES_ROUND_1B(K2) + AES_ROUND_1B(K3) + AES_ROUND_1B(K4) + AES_ROUND_1B(K5) + AES_ROUND_1B(K6) + AES_ROUND_1B(K7) + AES_ROUND_1B(K8) + cmp ROUNDS,#10 + b.eq L1B_10_round + cmp ROUNDS,#12 + b.eq L1B_12_round + b L1B_14_round + +L1B_10_round: + AES_LAST_ROUND_1B(K9) + b L1B_done +L1B_12_round: + AES_ROUND_1B(K9) + AES_ROUND_1B(K10) + AES_LAST_ROUND_1B(K11) + b L1B_done +L1B_14_round: + AES_ROUND_1B(K9) + AES_ROUND_1B(K10) + AES_ROUND_1B(K11) + AES_ROUND_1B(K12) + AES_LAST_ROUND_1B(K13) + +L1B_done: + st1 {S0.16b},[DST],#16 + + subs LENGTH,LENGTH,#16 + b.ne L1B_loop + +Ldone: + ret +EPILOGUE(_nettle_aes_decrypt) diff --git a/arm64/crypto/aes-encrypt-internal.asm b/arm64/crypto/aes-encrypt-internal.asm new file mode 100644 index 00000000..314f9333 --- /dev/null +++ b/arm64/crypto/aes-encrypt-internal.asm @@ -0,0 +1,223 @@ +C arm64/crypto/aes-encrypt-internal.asm + +ifelse(` + Copyright (C) 2021 Mamone Tarsha + This file is part of GNU Nettle. + + GNU Nettle is free software: you can redistribute it and/or + modify it under the terms of either: + + * the GNU Lesser General Public License as published by the Free + Software Foundation; either version 3 of the License, or (at your + option) any later version. + + or + + * the GNU General Public License as published by the Free + Software Foundation; either version 2 of the License, or (at your + option) any later version. + + or both in parallel, as here. + + GNU Nettle is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received copies of the GNU General Public License and + the GNU Lesser General Public License along with this program. If + not, see http://www.gnu.org/licenses/. +') + +.file "aes-encrypt-internal.asm" +.arch armv8-a+crypto + +.text + +C Register usage: + +define(`ROUNDS', `x0') +define(`KEYS', `x1') +define(`LENGTH', `x3') +define(`DST', `x4') +define(`SRC', `x5') + +define(`S0', `v0') +define(`S1', `v1') +define(`S2', `v2') +define(`S3', `v3') +define(`K0', `v16') +define(`K1', `v17') +define(`K2', `v18') +define(`K3', `v19') +define(`K4', `v20') +define(`K5', `v21') +define(`K6', `v22') +define(`K7', `v23') +define(`K8', `v24') +define(`K9', `v25') +define(`K10', `v26') +define(`K11', `v27') +define(`K12', `v28') +define(`K13', `v29') +define(`K14', `v30') + +C AES_ROUND_4B(KEY) +define(`AES_ROUND_4B', m4_assert_numargs(1)` + aese S0.16b,$1.16b + aesmc S0.16b,S0.16b + aese S1.16b,$1.16b + aesmc S1.16b,S1.16b + aese S2.16b,$1.16b + aesmc S2.16b,S2.16b + aese S3.16b,$1.16b + aesmc S3.16b,S3.16b +') + +C AES_LAST_ROUND_4B(KEY) +define(`AES_LAST_ROUND_4B', m4_assert_numargs(1)` + aese S0.16b,$1.16b + eor S0.16b,S0.16b,K14.16b + aese S1.16b,$1.16b + eor S1.16b,S1.16b,K14.16b + aese S2.16b,$1.16b + eor S2.16b,S2.16b,K14.16b + aese S3.16b,$1.16b + eor S3.16b,S3.16b,K14.16b +') + +C AES_ROUND_1B(KEY) +define(`AES_ROUND_1B', m4_assert_numargs(1)` + aese S0.16b,$1.16b + aesmc S0.16b,S0.16b +') + +C AES_LAST_ROUND_1B(KEY) +define(`AES_LAST_ROUND_1B', m4_assert_numargs(1)` + aese S0.16b,$1.16b + eor S0.16b,S0.16b,K14.16b +') + +C _aes_encrypt(unsigned rounds, const uint32_t *keys, +C const struct aes_table *T, +C size_t length, uint8_t *dst, +C uint8_t *src) + +PROLOGUE(_nettle_aes_encrypt) + ands x6,LENGTH,#-64 + b.eq L1B + + mov x7,KEYS + ld1 {K0.4s,K1.4s,K2.4s,K3.4s},[x7],#64 + ld1 {K4.4s,K5.4s,K6.4s,K7.4s},[x7],#64 + ld1 {K8.4s,K9.4s},[x7],#32 + cmp ROUNDS,#10 + b.eq L4B_last_key + ld1 {K10.4s,K11.4s},[x7],#32 + cmp ROUNDS,#12 + b.eq L4B_last_key + ld1 {K12.4s,K13.4s},[x7],#32 + +L4B_last_key: + ld1 {K14.4s},[x7] + +L4B_loop: + ld1 {S0.16b,S1.16b,S2.16b,S3.16b},[SRC],#64 + + AES_ROUND_4B(K0) + AES_ROUND_4B(K1) + AES_ROUND_4B(K2) + AES_ROUND_4B(K3) + AES_ROUND_4B(K4) + AES_ROUND_4B(K5) + AES_ROUND_4B(K6) + AES_ROUND_4B(K7) + AES_ROUND_4B(K8) + cmp ROUNDS,#10 + b.eq L4B_10_round + cmp ROUNDS,#12 + b.eq L4B_12_round + b L4B_14_round + +L4B_10_round: + AES_LAST_ROUND_4B(K9) + b L4B_done +L4B_12_round: + AES_ROUND_4B(K9) + AES_ROUND_4B(K10) + AES_LAST_ROUND_4B(K11) + b L4B_done +L4B_14_round: + AES_ROUND_4B(K9) + AES_ROUND_4B(K10) + AES_ROUND_4B(K11) + AES_ROUND_4B(K12) + AES_LAST_ROUND_4B(K13) + +L4B_done: + st1 {S0.16b,S1.16b,S2.16b,S3.16b},[DST],#64 + + subs x6,x6,#64 + b.ne L4B_loop + + and LENGTH,LENGTH,#63 + +L1B: + cbz LENGTH,Ldone + + mov x6,KEYS + ld1 {K0.4s,K1.4s,K2.4s,K3.4s},[x6],#64 + ld1 {K4.4s,K5.4s,K6.4s,K7.4s},[x6],#64 + ld1 {K8.4s,K9.4s},[x6],#32 + cmp ROUNDS,#10 + b.eq L1B_last_key + ld1 {K10.4s,K11.4s},[x6],#32 + cmp ROUNDS,#12 + b.eq L1B_last_key + ld1 {K12.4s,K13.4s},[x6],#32 + +L1B_last_key: + ld1 {K14.4s},[x6] + +L1B_loop: + ld1 {S0.16b},[SRC],#16 + + AES_ROUND_1B(K0) + AES_ROUND_1B(K1) + AES_ROUND_1B(K2) + AES_ROUND_1B(K3) + AES_ROUND_1B(K4) + AES_ROUND_1B(K5) + AES_ROUND_1B(K6) + AES_ROUND_1B(K7) + AES_ROUND_1B(K8) + cmp ROUNDS,#10 + b.eq L1B_10_round + cmp ROUNDS,#12 + b.eq L1B_12_round + b L1B_14_round + +L1B_10_round: + AES_LAST_ROUND_1B(K9) + b L1B_done +L1B_12_round: + AES_ROUND_1B(K9) + AES_ROUND_1B(K10) + AES_LAST_ROUND_1B(K11) + b L1B_done +L1B_14_round: + AES_ROUND_1B(K9) + AES_ROUND_1B(K10) + AES_ROUND_1B(K11) + AES_ROUND_1B(K12) + AES_LAST_ROUND_1B(K13) + +L1B_done: + st1 {S0.16b},[DST],#16 + + subs LENGTH,LENGTH,#16 + b.ne L1B_loop + +Ldone: + ret +EPILOGUE(_nettle_aes_encrypt) -- 2.25.1

2 2

HPKE implementation
by Norbert Pocs 14 Jul '21

14 Jul '21

Hello mailing list, I am a student at Brno University of Technology at Faculty of Information Technology and intern at Red Hat Crypto team. My current project is the implementation of HPKE draft [0]. The first goal is to implement mode_base. Example usage for the project is the encrypted hello message in TLS [1]. Do you have interest in merging the code after completion? [0] https://tools.ietf.org/html/draft-irtf-cfrg-hpke-07 [1] https://gitlab.com/gnutls/gnutls/-/issues/595 Regards Norbert Pócs

2 3

[AArch64] Fat build support for SHA-256 compress
by Maamoun TK 05 Jul '21

05 Jul '21

I made a merge request that adds fat build support for SHA-256 compress function !29 <https://git.lysator.liu.se/nettle/nettle/-/merge_requests/29> regards, Mamone

2 1

[AArch64] Optimize SHA-256 compress
by Maamoun TK 01 Jul '21

01 Jul '21

I made a merge request !28 <https://git.lysator.liu.se/nettle/nettle/-/merge_requests/28> to the 'arm64-sha1' branch that optimizes SHA-256 compress function, I've added a brief description of the patch in addition to benchmark numbers in the MR description. A patch for fat build support will be followed in another merge request. regards, Mamone

2 1

[Aarch64] Fat build support for SHA1 compress
by Maamoun TK 30 Jun '21

30 Jun '21

This patch added fat build support SHA1 compress function using the regular HWCAP features. --- arm64/fat/sha1-compress-2.asm | 37 +++++++++++++++++++++++++++++++++++++ fat-arm64.c | 32 ++++++++++++++++++++++++++++++-- 2 files changed, 67 insertions(+), 2 deletions(-) create mode 100644 arm64/fat/sha1-compress-2.asm diff --git a/arm64/fat/sha1-compress-2.asm b/arm64/fat/sha1-compress-2.asm new file mode 100644 index 00000000..b53cb63e --- /dev/null +++ b/arm64/fat/sha1-compress-2.asm @@ -0,0 +1,37 @@ +C arm64/fat/sha1-compress-2.asm + + +ifelse(` + Copyright (C) 2021 Mamone Tarsha + + This file is part of GNU Nettle. + + GNU Nettle is free software: you can redistribute it and/or + modify it under the terms of either: + + * the GNU Lesser General Public License as published by the Free + Software Foundation; either version 3 of the License, or (at your + option) any later version. + + or + + * the GNU General Public License as published by the Free + Software Foundation; either version 2 of the License, or (at your + option) any later version. + + or both in parallel, as here. + + GNU Nettle is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received copies of the GNU General Public License and + the GNU Lesser General Public License along with this program. If + not, see http://www.gnu.org/licenses/. +') + +dnl PROLOGUE(nettle_sha1_compress) picked up by configure + +define(`fat_transform', `_$1_arm64') +include_src(`arm64/crypto/sha1-compress.asm') diff --git a/fat-arm64.c b/fat-arm64.c index 9f81951f..c696e9bb 100644 --- a/fat-arm64.c +++ b/fat-arm64.c @@ -61,10 +61,14 @@ #ifndef HWCAP_PMULL #define HWCAP_PMULL (1 << 4) #endif +#ifndef HWCAP_SHA1 +#define HWCAP_SHA1 (1 << 5) +#endif struct arm64_features { int have_pmull; + int have_sha1; }; #define MATCH(s, slen, literal, llen) \ @@ -75,6 +79,7 @@ get_arm64_features (struct arm64_features *features) { const char *s; features->have_pmull = 0; + features->have_sha1 = 0; s = secure_getenv (ENV_OVERRIDE); if (s) @@ -85,6 +90,8 @@ get_arm64_features (struct arm64_features *features) if (MATCH (s, length, "pmull", 5)) features->have_pmull = 1; + else if (MATCH (s, length, "sha1", 4)) + features->have_sha1 = 1; if (!sep) break; s = sep + 1; @@ -95,6 +102,8 @@ get_arm64_features (struct arm64_features *features) unsigned long hwcap = getauxval(AT_HWCAP); features->have_pmull = ((hwcap & (HWCAP_ASIMD | HWCAP_PMULL)) == (HWCAP_ASIMD | HWCAP_PMULL)); + features->have_sha1 + = ((hwcap & (HWCAP_ASIMD | HWCAP_SHA1)) == (HWCAP_ASIMD | HWCAP_SHA1)); #endif } } @@ -109,6 +118,10 @@ DECLARE_FAT_FUNC_VAR(gcm_hash, gcm_hash_func, c) DECLARE_FAT_FUNC_VAR(gcm_hash, gcm_hash_func, arm64) #endif /* GCM_TABLE_BITS == 8 */ +DECLARE_FAT_FUNC(nettle_sha1_compress, sha1_compress_func) +DECLARE_FAT_FUNC_VAR(sha1_compress, sha1_compress_func, c) +DECLARE_FAT_FUNC_VAR(sha1_compress, sha1_compress_func, arm64) + static void CONSTRUCTOR fat_init (void) { @@ -119,8 +132,9 @@ fat_init (void) verbose = getenv (ENV_VERBOSE) != NULL; if (verbose) - fprintf (stderr, "libnettle: cpu features: %s\n", - features.have_pmull ? "polynomial multiply long instructions (PMULL/PMULL2)" : ""); + fprintf (stderr, "libnettle: cpu features:%s%s\n", + features.have_pmull ? " polynomial multiply long instructions (PMULL/PMULL2)" : "", + features.have_sha1 ? " sha1 instructions" : ""); if (features.have_pmull) { @@ -142,6 +156,16 @@ fat_init (void) _nettle_gcm_hash_vec = _nettle_gcm_hash_c; #endif /* GCM_TABLE_BITS == 8 */ } + if (features.have_sha1) + { + if (verbose) + fprintf (stderr, "libnettle: enabling hardware-accelerated sha1 compress code.\n"); + nettle_sha1_compress_vec = _nettle_sha1_compress_arm64; + } + else + { + nettle_sha1_compress_vec = _nettle_sha1_compress_c; + } } #if GCM_TABLE_BITS == 8 @@ -154,3 +178,7 @@ DEFINE_FAT_FUNC(_nettle_gcm_hash, void, size_t length, const uint8_t *data), (key, x, length, data)) #endif /* GCM_TABLE_BITS == 8 */ + +DEFINE_FAT_FUNC(nettle_sha1_compress, void, + (uint32_t *state, const uint8_t *input), + (state, input)) -- 2.25.1

2 1

Re: [S390x] Optimize AES modes
by Maamoun TK 30 Jun '21

30 Jun '21

Hi Niels, Did you get a chance to look at the recently posted patches? Thank you, Mamone

1 0

← Newer
1
...
5
6
7
8
9
10
11
...
220
Older →

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

nettle-bugs