From nobody Thu Apr 2 12:33:36 2026 Received: from mail-wm1-f54.google.com (mail-wm1-f54.google.com [209.85.128.54]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id A4E08946C for ; Sun, 29 Mar 2026 07:43:43 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.128.54 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1774770225; cv=none; b=MDbf1cIIPufF7lfZyDzUBXBvexw17aPAHK06tdp8UKmLpISMQAv7/kY3+bZ9MiSa5ZleAkfd3wFQHwXae3YpXyN17Qyv4jaOyVbemS06lZ0yPFpYXqufkgVnGBRAz6Iji4SctEK9YoQlw17msExZF/X4rtEcqP7qDr972/3W2o8= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1774770225; c=relaxed/simple; bh=ZQxTUnGJ7EyXTmiyyB9+ysKAsAt1gRuRaGjAPxV88fc=; h=From:To:Cc:Subject:Date:Message-ID:In-Reply-To:References: MIME-Version; b=ULZh3gsxE21jpLlyscsWH94Qbha8AQK8Ac/S4STbXrPot48RlXVbNV/iNKsnr2DYuVxnh37vX9AFu7RlEcp+m1iQScJufHiplNiZ1WzvbYmqQdOSr4x4mM9iWnbVtOIpI3jGgKxc945QMPQGl3rVD+MbYl9H4Ymp7wNIuUKMjWY= ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com; spf=pass smtp.mailfrom=gmail.com; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b=d34oj+z4; arc=none smtp.client-ip=209.85.128.54 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=gmail.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b="d34oj+z4" Received: by mail-wm1-f54.google.com with SMTP id 5b1f17b1804b1-48557c8ad47so26388465e9.0 for ; Sun, 29 Mar 2026 00:43:43 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20251104; t=1774770222; x=1775375022; darn=vger.kernel.org; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:from:to:cc:subject:date :message-id:reply-to; bh=mR7zW5huiWH1x5CVyoEzlAdCsPYlEqDzaC0AadT8DWY=; b=d34oj+z40cPqo3BbYPhEwncEeElBDQRfbnXoK9DfzdTxxacQGQXnwKTF7EKxYWrxvb RWuUum/z8F31WwQxi7p/qwKmWiS19A9vJjrsZ+soVLyYw6C3kMuxf95g4zD2BUh/4/Kt w1gNZhjC0mJ9bJvxSwQUg3LcZG6ku2gFn3Te+lSnKmD5cQaODiIbu5cw57XvT60Es4// R/r3O2Iv86CqZqe8bOjPQ81gylQnrdCGfhCq/hkqbSO2dTaf79O6WErTbgdqiEagDQyX UJZIuYBEmmx4CgSj3gsJtq4+nsu2Gm8cc8kCXNrms2VGDDIyHz5NTxRHimFR3o+kV77J 9TtA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1774770222; x=1775375022; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:x-gm-gg:x-gm-message-state:from :to:cc:subject:date:message-id:reply-to; bh=mR7zW5huiWH1x5CVyoEzlAdCsPYlEqDzaC0AadT8DWY=; b=LlyfvkagbSQI1TWS2gsxdPL3BUFHlK3aRQv/c1DOinCPI8+bDyJWal0HNNfslrYEo4 nUyUUH3supL1joKWf1EpifDC1O0NzcuivsC+m2dK0MqUFMxhD6xM4HH4h3OKgMzTmbbR GgpJmwUrpMVCTOOQK4ib3N3KU56ws5bD1+OjHRM66IwVVTKCiybb7DxdyNdv4JOamQHl oubeSh2r98poJQa5Fj5PDoLsRZ7rsAjkPFLkV1syUuCZhen/YAieauddYsNIa0YrWnQD pKno98YPhsjZREh0rwJt/symuXfI+oCq/gt6Cu2uABzgJnw2EImBN6esfAcGKixJvDlL ZOyg== X-Forwarded-Encrypted: i=1; AJvYcCVYDv3YcC+5lIm6SmUCbOtHc459IRYKWtQwcsUelwQj7rtXBNPu02ABntFclab+5eaa5Osdv2irMza7CVA=@vger.kernel.org X-Gm-Message-State: AOJu0Yw6N9TDzi08ohWNamOBK5M4MVcAmIzlURSzJGzZRN0rNtnvCY1N hyb5AzR9FiOQa7hO4PQyFrcWVqU5vXAY0teo9UngXtSl3dQrTOODYyqPiccIsEs7 X-Gm-Gg: ATEYQzx5Xwy6XpVG1gB8G14nWyB7T3Uk5/eniK43HXeldB7+vtiSmw4V2/urbnfast/ 4+GopkoqUgbFOkX/3OPWSs98Gy+BZDAPh+qmIqVHTn/FeJvkshYY8yyVcMqOdUUChhNV2ChBr5b xPIfp6jBJ4geMC7yHiLQrl+yJWI61x33429RJVEXml1Q+wolcfTocIhy4YgaY0ztPmDABjyDNw7 VX84adqnM5D7mnqFEDNjGkxrfSlzg3fdVK5akuqoRGTwd2AIOkcQ9Yv+yRU9f1XLg2K2J3PIM/F lN6ztfUGD9jQINl9b7M0wo7szhxTlxqTsGkI+rEulrAZSOLuukAAQHqlfuoYxiMoq25rv4cbX01 +5PYkTXyqOb8hYP8dQnsK+SaWpJj1XOxiy8XIBe/lqveylpkufGERINDLlzmApCGzB5xExEqlF7 7cZMT5kVCajDhfsv/ClY94DxDdlUOLbx3Z X-Received: by 2002:a05:600c:8b31:b0:485:30f7:6e88 with SMTP id 5b1f17b1804b1-48727efabdemr165719785e9.31.1774770221785; Sun, 29 Mar 2026 00:43:41 -0700 (PDT) Received: from lima-dev.. ([45.89.90.224]) by smtp.gmail.com with ESMTPSA id 5b1f17b1804b1-48722d38a5fsm193112255e9.12.2026.03.29.00.43.40 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Sun, 29 Mar 2026 00:43:41 -0700 (PDT) From: Demian Shulhan To: linux-crypto@vger.kernel.org, linux-kernel@vger.kernel.org, linux-arm-kernel@lists.infradead.org Cc: ebiggers@kernel.org, ardb@kernel.org, Demian Shulhan Subject: [PATCH v3] lib/crc: arm64: add NEON accelerated CRC64-NVMe implementation Date: Sun, 29 Mar 2026 07:43:38 +0000 Message-ID: <20260329074338.1053550-1-demyansh@gmail.com> X-Mailer: git-send-email 2.43.0 In-Reply-To: <20260317065425.2684093-1-demyansh@gmail.com> References: <20260317065425.2684093-1-demyansh@gmail.com> Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" Implement an optimized CRC64 (NVMe) algorithm for ARM64 using NEON Polynomial Multiply Long (PMULL) instructions. The generic shift-and-XOR software implementation is slow, which creates a bottleneck in NVMe and other storage subsystems. The acceleration is implemented using C intrinsics () rather than raw assembly for better readability and maintainability. Key highlights of this implementation: - Uses 4KB chunking inside scoped_ksimd() to avoid preemption latency spikes on large buffers. - Pre-calculates and loads fold constants via vld1q_u64() to minimize register spilling. - Benchmarks show the break-even point against the generic implementation is around 128 bytes. The PMULL path is enabled only for len >=3D 128. Performance results (kunit crc_benchmark on Cortex-A72): - Generic (len=3D4096): ~268 MB/s - PMULL (len=3D4096): ~1556 MB/s (nearly 6x improvement) Signed-off-by: Demian Shulhan --- v2: - Removed KERNEL_MODE_NEON check from Kconfig as it's redundant on arm6= 4. - Added missing prototype for crc64_nvme_arm64_c to fix sparse/W=3D1 warn= ing. - Improved readability in Makefile with extra newlines and comments. - Removed redundant include guards in crc64.h. - Switched to do-while loops for better optimization in hot paths. - Added comments explaining the magic constants (fold/Barrett). --- v3: - Removed big-endian fallback from the commit message. - Rewrote the comment explaining the final Barrett reduction step. - Adjusted the formatting of the scoped_ksimd() call. --- lib/crc/Kconfig | 1 + lib/crc/Makefile | 8 +++- lib/crc/arm64/crc64-neon-inner.c | 78 ++++++++++++++++++++++++++++++++ lib/crc/arm64/crc64.h | 30 ++++++++++++ 4 files changed, 116 insertions(+), 1 deletion(-) create mode 100644 lib/crc/arm64/crc64-neon-inner.c create mode 100644 lib/crc/arm64/crc64.h diff --git a/lib/crc/Kconfig b/lib/crc/Kconfig index 70e7a6016de3..16cb42d5e306 100644 --- a/lib/crc/Kconfig +++ b/lib/crc/Kconfig @@ -82,6 +82,7 @@ config CRC64 config CRC64_ARCH bool depends on CRC64 && CRC_OPTIMIZATIONS + default y if ARM64 default y if RISCV && RISCV_ISA_ZBC && 64BIT default y if X86_64 =20 diff --git a/lib/crc/Makefile b/lib/crc/Makefile index 7543ad295ab6..c9c35419b39c 100644 --- a/lib/crc/Makefile +++ b/lib/crc/Makefile @@ -38,9 +38,15 @@ obj-$(CONFIG_CRC64) +=3D crc64.o crc64-y :=3D crc64-main.o ifeq ($(CONFIG_CRC64_ARCH),y) CFLAGS_crc64-main.o +=3D -I$(src)/$(SRCARCH) + +CFLAGS_REMOVE_arm64/crc64-neon-inner.o +=3D -mgeneral-regs-only +CFLAGS_arm64/crc64-neon-inner.o +=3D -ffreestanding -march=3Darmv8-a+crypto +CFLAGS_arm64/crc64-neon-inner.o +=3D -isystem $(shell $(CC) -print-file-na= me=3Dinclude) +crc64-$(CONFIG_ARM64) +=3D arm64/crc64-neon-inner.o + crc64-$(CONFIG_RISCV) +=3D riscv/crc64_lsb.o riscv/crc64_msb.o crc64-$(CONFIG_X86) +=3D x86/crc64-pclmul.o -endif +endif # CONFIG_CRC64_ARCH =20 obj-y +=3D tests/ =20 diff --git a/lib/crc/arm64/crc64-neon-inner.c b/lib/crc/arm64/crc64-neon-in= ner.c new file mode 100644 index 000000000000..881cdafadb37 --- /dev/null +++ b/lib/crc/arm64/crc64-neon-inner.c @@ -0,0 +1,78 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Accelerated CRC64 (NVMe) using ARM NEON C intrinsics + */ + +#include +#include + +u64 crc64_nvme_arm64_c(u64 crc, const u8 *p, size_t len); + +#define GET_P64_0(v) ((poly64_t)vgetq_lane_u64(vreinterpretq_u64_p64(v), 0= )) +#define GET_P64_1(v) ((poly64_t)vgetq_lane_u64(vreinterpretq_u64_p64(v), 1= )) + +/* x^191 mod G, x^127 mod G */ +static const u64 fold_consts_val[2] =3D { 0xeadc41fd2ba3d420ULL, + 0x21e9761e252621acULL }; +/* floor(x^127 / G), (G - x^64) / x */ +static const u64 bconsts_val[2] =3D { 0x27ecfa329aef9f77ULL, + 0x34d926535897936aULL }; + +u64 crc64_nvme_arm64_c(u64 crc, const u8 *p, size_t len) +{ + uint64x2_t v0_u64 =3D { crc, 0 }; + poly64x2_t v0 =3D vreinterpretq_p64_u64(v0_u64); + poly64x2_t fold_consts =3D + vreinterpretq_p64_u64(vld1q_u64(fold_consts_val)); + poly64x2_t v1 =3D vreinterpretq_p64_u8(vld1q_u8(p)); + + v0 =3D vreinterpretq_p64_u8(veorq_u8(vreinterpretq_u8_p64(v0), + vreinterpretq_u8_p64(v1))); + p +=3D 16; + len -=3D 16; + + do { + v1 =3D vreinterpretq_p64_u8(vld1q_u8(p)); + + poly128_t v2 =3D vmull_high_p64(fold_consts, v0); + poly128_t v0_128 =3D + vmull_p64(GET_P64_0(fold_consts), GET_P64_0(v0)); + + uint8x16_t x0 =3D veorq_u8(vreinterpretq_u8_p128(v0_128), + vreinterpretq_u8_p128(v2)); + + x0 =3D veorq_u8(x0, vreinterpretq_u8_p64(v1)); + v0 =3D vreinterpretq_p64_u8(x0); + + p +=3D 16; + len -=3D 16; + } while (len >=3D 16); + + /* Multiply the 128-bit value by x^64 and reduce it back to 128 bits. */ + poly64x2_t v7 =3D vreinterpretq_p64_u64((uint64x2_t){ 0, 0 }); + poly128_t v1_128 =3D vmull_p64(GET_P64_1(fold_consts), GET_P64_0(v0)); + + uint8x16_t ext_v0 =3D + vextq_u8(vreinterpretq_u8_p64(v0), vreinterpretq_u8_p64(v7), 8); + uint8x16_t x0 =3D veorq_u8(ext_v0, vreinterpretq_u8_p128(v1_128)); + + v0 =3D vreinterpretq_p64_u8(x0); + + /* Final Barrett reduction */ + poly64x2_t bconsts =3D vreinterpretq_p64_u64(vld1q_u64(bconsts_val)); + + v1_128 =3D vmull_p64(GET_P64_0(bconsts), GET_P64_0(v0)); + + poly64x2_t v1_64 =3D vreinterpretq_p64_u8(vreinterpretq_u8_p128(v1_128)); + poly128_t v3_128 =3D vmull_p64(GET_P64_1(bconsts), GET_P64_0(v1_64)); + + x0 =3D veorq_u8(vreinterpretq_u8_p64(v0), vreinterpretq_u8_p128(v3_128)); + + uint8x16_t ext_v2 =3D vextq_u8(vreinterpretq_u8_p64(v7), + vreinterpretq_u8_p128(v1_128), 8); + + x0 =3D veorq_u8(x0, ext_v2); + + v0 =3D vreinterpretq_p64_u8(x0); + return vgetq_lane_u64(vreinterpretq_u64_p64(v0), 1); +} diff --git a/lib/crc/arm64/crc64.h b/lib/crc/arm64/crc64.h new file mode 100644 index 000000000000..cc65abeee24c --- /dev/null +++ b/lib/crc/arm64/crc64.h @@ -0,0 +1,30 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * CRC64 using ARM64 PMULL instructions + */ + +#include +#include +#include +#include + +u64 crc64_nvme_arm64_c(u64 crc, const u8 *p, size_t len); + +#define crc64_be_arch crc64_be_generic + +static inline u64 crc64_nvme_arch(u64 crc, const u8 *p, size_t len) +{ + if (len >=3D 128 && cpu_have_named_feature(PMULL) && + likely(may_use_simd())) { + do { + size_t chunk =3D min_t(size_t, len & ~15, SZ_4K); + + scoped_ksimd() + crc =3D crc64_nvme_arm64_c(crc, p, chunk); + + p +=3D chunk; + len -=3D chunk; + } while (len >=3D 128); + } + return crc64_nvme_generic(crc, p, len); +} --=20 2.43.0