Re: [PATCH 5/6] crypto: x86/sm3 - add AVX assembly implementation

[Jussi Kivilinna]

On 20.12.2021 10.22, Tianjia Zhang wrote:
> This patch adds AVX assembly accelerated implementation of SM3 secure
> hash algorithm. From the benchmark data, compared to pure software
> implementation sm3-generic, the performance increase is up to 38%.
>
> The main algorithm implementation based on SM3 AES/BMI2 accelerated
> work by libgcrypt at:
> https://gnupg.org/software/libgcrypt/index.html
>
> Benchmark on Intel i5-6200U 2.30GHz, performance data of two
> implementations, pure software sm3-generic and sm3-avx acceleration.
> The data comes from the 326 mode and 422 mode of tcrypt. The abscissas
> are different lengths of per update. The data is tabulated and the
> unit is Mb/s:
>
> update-size |     16      64     256    1024    2048    4096    8192
> --------------------------------------------------------------------
> sm3-generic | 105.97  129.60  182.12  189.62  188.06  193.66  194.88
> sm3-avx     | 119.87  163.05  244.44  260.92  257.60  264.87  265.88
>
> Signed-off-by: Tianjia Zhang <tianjia.zhang@xxxxxxxxxxxxxxxxx>
> ---
>   arch/x86/crypto/Makefile         |   3 +
>   arch/x86/crypto/sm3-avx-asm_64.S | 521 +++++++++++++++++++++++++++++++
>   arch/x86/crypto/sm3_avx_glue.c   | 134 ++++++++
>   crypto/Kconfig                   |  13 +
>   4 files changed, 671 insertions(+)
>   create mode 100644 arch/x86/crypto/sm3-avx-asm_64.S
>   create mode 100644 arch/x86/crypto/sm3_avx_glue.c
>
> diff --git a/arch/x86/crypto/Makefile b/arch/x86/crypto/Makefile
> index f307c93fc90a..7cbe860f6201 100644
> --- a/arch/x86/crypto/Makefile
> +++ b/arch/x86/crypto/Makefile
> @@ -88,6 +88,9 @@ nhpoly1305-avx2-y := nh-avx2-x86_64.o nhpoly1305-avx2-glue.o
>   
>   obj-$(CONFIG_CRYPTO_CURVE25519_X86) += curve25519-x86_64.o
>   
> +obj-$(CONFIG_CRYPTO_SM3_AVX_X86_64) += sm3-avx-x86_64.o
> +sm3-avx-x86_64-y := sm3-avx-asm_64.o sm3_avx_glue.o
> +
>   obj-$(CONFIG_CRYPTO_SM4_AESNI_AVX_X86_64) += sm4-aesni-avx-x86_64.o
>   sm4-aesni-avx-x86_64-y := sm4-aesni-avx-asm_64.o sm4_aesni_avx_glue.o
>   
> diff --git a/arch/x86/crypto/sm3-avx-asm_64.S b/arch/x86/crypto/sm3-avx-asm_64.S
> new file mode 100644
> index 000000000000..e7a9a37f3609
> --- /dev/null
> +++ b/arch/x86/crypto/sm3-avx-asm_64.S
> @@ -0,0 +1,521 @@
> +/* SPDX-License-Identifier: GPL-2.0-or-later */
> +/*
> + * SM3 AVX accelerated transform.
> + * specified in: https://datatracker.ietf.org/doc/html/draft-sca-cfrg-sm3-02
> + *
> + * Copyright (C) 2021 Jussi Kivilinna <jussi.kivilinna@xxxxxx>
> + * Copyright (C) 2021 Tianjia Zhang <tianjia.zhang@xxxxxxxxxxxxxxxxx>
> + */
<snip>
> +
> +#define R(i, a, b, c, d, e, f, g, h, round, widx, wtype)                      \
> +	/* rol(a, 12) => t0 */                                                \
> +	roll3mov(12, a, t0); /* rorxl here would reduce perf by 6% on zen3 */ \
> +	/* rol (t0 + e + t), 7) => t1 */                                      \
> +	addl3(t0, e, t1);                                                     \
> +	addl $K##round, t1;                                                   \

It's better to use "leal K##round(t0, e, 1), t1;" here and fix K0-K63 macros
instead as I noted at libgcrypt mailing-list:
 https://lists.gnupg.org/pipermail/gcrypt-devel/2021-December/005209.html

-Jussi