mirror_zfs/lib/libspl/include/sys/simd.h

520 lines
12 KiB
C
Raw Normal View History

/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License, Version 1.0 only
* (the "License"). You may not use this file except in compliance
* with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or https://opensource.org/licenses/CDDL-1.0.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2006 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2022 Tino Reichardt <milky-zfs@mcmilk.de>
*/
#ifndef _LIBSPL_SYS_SIMD_H
#define _LIBSPL_SYS_SIMD_H
#include <sys/isa_defs.h>
#include <sys/types.h>
#if defined(__x86)
#include <cpuid.h>
#define kfpu_allowed() 1
#define kfpu_begin() do {} while (0)
#define kfpu_end() do {} while (0)
#define kfpu_init() 0
#define kfpu_fini() ((void) 0)
/*
* CPUID feature tests for user-space.
*
* x86 registers used implicitly by CPUID
*/
typedef enum cpuid_regs {
EAX = 0,
EBX,
ECX,
EDX,
CPUID_REG_CNT = 4
} cpuid_regs_t;
/*
* List of instruction sets identified by CPUID
*/
typedef enum cpuid_inst_sets {
SSE = 0,
SSE2,
SSE3,
SSSE3,
SSE4_1,
SSE4_2,
OSXSAVE,
AVX,
AVX2,
BMI1,
BMI2,
AVX512F,
AVX512CD,
AVX512DQ,
AVX512BW,
AVX512IFMA,
AVX512VBMI,
AVX512PF,
AVX512ER,
AVX512VL,
AES,
ICP: Improve AES-GCM performance Currently SIMD accelerated AES-GCM performance is limited by two factors: a. The need to disable preemption and interrupts and save the FPU state before using it and to do the reverse when done. Due to the way the code is organized (see (b) below) we have to pay this price twice for each 16 byte GCM block processed. b. Most processing is done in C, operating on single GCM blocks. The use of SIMD instructions is limited to the AES encryption of the counter block (AES-NI) and the Galois multiplication (PCLMULQDQ). This leads to the FPU not being fully utilized for crypto operations. To solve (a) we do crypto processing in larger chunks while owning the FPU. An `icp_gcm_avx_chunk_size` module parameter was introduced to make this chunk size tweakable. It defaults to 32 KiB. This step alone roughly doubles performance. (b) is tackled by porting and using the highly optimized openssl AES-GCM assembler routines, which do all the processing (CTR, AES, GMULT) in a single routine. Both steps together result in up to 32x reduction of the time spend in the en/decryption routines, leading up to approximately 12x throughput increase for large (128 KiB) blocks. Lastly, this commit changes the default encryption algorithm from AES-CCM to AES-GCM when setting the `encryption=on` property. Reviewed-By: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-By: Jason King <jason.king@joyent.com> Reviewed-By: Tom Caputi <tcaputi@datto.com> Reviewed-By: Richard Laager <rlaager@wiktel.com> Signed-off-by: Attila Fülöp <attila@fueloep.org> Closes #9749
2020-02-10 23:59:50 +03:00
PCLMULQDQ,
MOVBE
} cpuid_inst_sets_t;
/*
* Instruction set descriptor.
*/
typedef struct cpuid_feature_desc {
uint32_t leaf; /* CPUID leaf */
uint32_t subleaf; /* CPUID sub-leaf */
uint32_t flag; /* bit mask of the feature */
cpuid_regs_t reg; /* which CPUID return register to test */
} cpuid_feature_desc_t;
#define _AVX512F_BIT (1U << 16)
#define _AVX512CD_BIT (_AVX512F_BIT | (1U << 28))
#define _AVX512DQ_BIT (_AVX512F_BIT | (1U << 17))
#define _AVX512BW_BIT (_AVX512F_BIT | (1U << 30))
#define _AVX512IFMA_BIT (_AVX512F_BIT | (1U << 21))
#define _AVX512VBMI_BIT (1U << 1) /* AVX512F_BIT is on another leaf */
#define _AVX512PF_BIT (_AVX512F_BIT | (1U << 26))
#define _AVX512ER_BIT (_AVX512F_BIT | (1U << 27))
#define _AVX512VL_BIT (1U << 31) /* if used also check other levels */
#define _AES_BIT (1U << 25)
#define _PCLMULQDQ_BIT (1U << 1)
ICP: Improve AES-GCM performance Currently SIMD accelerated AES-GCM performance is limited by two factors: a. The need to disable preemption and interrupts and save the FPU state before using it and to do the reverse when done. Due to the way the code is organized (see (b) below) we have to pay this price twice for each 16 byte GCM block processed. b. Most processing is done in C, operating on single GCM blocks. The use of SIMD instructions is limited to the AES encryption of the counter block (AES-NI) and the Galois multiplication (PCLMULQDQ). This leads to the FPU not being fully utilized for crypto operations. To solve (a) we do crypto processing in larger chunks while owning the FPU. An `icp_gcm_avx_chunk_size` module parameter was introduced to make this chunk size tweakable. It defaults to 32 KiB. This step alone roughly doubles performance. (b) is tackled by porting and using the highly optimized openssl AES-GCM assembler routines, which do all the processing (CTR, AES, GMULT) in a single routine. Both steps together result in up to 32x reduction of the time spend in the en/decryption routines, leading up to approximately 12x throughput increase for large (128 KiB) blocks. Lastly, this commit changes the default encryption algorithm from AES-CCM to AES-GCM when setting the `encryption=on` property. Reviewed-By: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-By: Jason King <jason.king@joyent.com> Reviewed-By: Tom Caputi <tcaputi@datto.com> Reviewed-By: Richard Laager <rlaager@wiktel.com> Signed-off-by: Attila Fülöp <attila@fueloep.org> Closes #9749
2020-02-10 23:59:50 +03:00
#define _MOVBE_BIT (1U << 22)
/*
* Descriptions of supported instruction sets
*/
static const cpuid_feature_desc_t cpuid_features[] = {
[SSE] = {1U, 0U, 1U << 25, EDX },
[SSE2] = {1U, 0U, 1U << 26, EDX },
[SSE3] = {1U, 0U, 1U << 0, ECX },
[SSSE3] = {1U, 0U, 1U << 9, ECX },
[SSE4_1] = {1U, 0U, 1U << 19, ECX },
[SSE4_2] = {1U, 0U, 1U << 20, ECX },
[OSXSAVE] = {1U, 0U, 1U << 27, ECX },
[AVX] = {1U, 0U, 1U << 28, ECX },
[AVX2] = {7U, 0U, 1U << 5, EBX },
[BMI1] = {7U, 0U, 1U << 3, EBX },
[BMI2] = {7U, 0U, 1U << 8, EBX },
[AVX512F] = {7U, 0U, _AVX512F_BIT, EBX },
[AVX512CD] = {7U, 0U, _AVX512CD_BIT, EBX },
[AVX512DQ] = {7U, 0U, _AVX512DQ_BIT, EBX },
[AVX512BW] = {7U, 0U, _AVX512BW_BIT, EBX },
[AVX512IFMA] = {7U, 0U, _AVX512IFMA_BIT, EBX },
[AVX512VBMI] = {7U, 0U, _AVX512VBMI_BIT, ECX },
[AVX512PF] = {7U, 0U, _AVX512PF_BIT, EBX },
[AVX512ER] = {7U, 0U, _AVX512ER_BIT, EBX },
[AVX512VL] = {7U, 0U, _AVX512ER_BIT, EBX },
[AES] = {1U, 0U, _AES_BIT, ECX },
[PCLMULQDQ] = {1U, 0U, _PCLMULQDQ_BIT, ECX },
ICP: Improve AES-GCM performance Currently SIMD accelerated AES-GCM performance is limited by two factors: a. The need to disable preemption and interrupts and save the FPU state before using it and to do the reverse when done. Due to the way the code is organized (see (b) below) we have to pay this price twice for each 16 byte GCM block processed. b. Most processing is done in C, operating on single GCM blocks. The use of SIMD instructions is limited to the AES encryption of the counter block (AES-NI) and the Galois multiplication (PCLMULQDQ). This leads to the FPU not being fully utilized for crypto operations. To solve (a) we do crypto processing in larger chunks while owning the FPU. An `icp_gcm_avx_chunk_size` module parameter was introduced to make this chunk size tweakable. It defaults to 32 KiB. This step alone roughly doubles performance. (b) is tackled by porting and using the highly optimized openssl AES-GCM assembler routines, which do all the processing (CTR, AES, GMULT) in a single routine. Both steps together result in up to 32x reduction of the time spend in the en/decryption routines, leading up to approximately 12x throughput increase for large (128 KiB) blocks. Lastly, this commit changes the default encryption algorithm from AES-CCM to AES-GCM when setting the `encryption=on` property. Reviewed-By: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-By: Jason King <jason.king@joyent.com> Reviewed-By: Tom Caputi <tcaputi@datto.com> Reviewed-By: Richard Laager <rlaager@wiktel.com> Signed-off-by: Attila Fülöp <attila@fueloep.org> Closes #9749
2020-02-10 23:59:50 +03:00
[MOVBE] = {1U, 0U, _MOVBE_BIT, ECX },
};
/*
* Check if OS supports AVX and AVX2 by checking XCR0
* Only call this function if CPUID indicates that AVX feature is
* supported by the CPU, otherwise it might be an illegal instruction.
*/
static inline uint64_t
xgetbv(uint32_t index)
{
uint32_t eax, edx;
/* xgetbv - instruction byte code */
__asm__ __volatile__(".byte 0x0f; .byte 0x01; .byte 0xd0"
: "=a" (eax), "=d" (edx)
: "c" (index));
return ((((uint64_t)edx)<<32) | (uint64_t)eax);
}
/*
* Check if CPU supports a feature
*/
static inline boolean_t
__cpuid_check_feature(const cpuid_feature_desc_t *desc)
{
uint32_t r[CPUID_REG_CNT];
if (__get_cpuid_max(0, NULL) >= desc->leaf) {
/*
* __cpuid_count is needed to properly check
* for AVX2. It is a macro, so return parameters
* are passed by value.
*/
__cpuid_count(desc->leaf, desc->subleaf,
r[EAX], r[EBX], r[ECX], r[EDX]);
return ((r[desc->reg] & desc->flag) == desc->flag);
}
return (B_FALSE);
}
#define CPUID_FEATURE_CHECK(name, id) \
static inline boolean_t \
__cpuid_has_ ## name(void) \
{ \
return (__cpuid_check_feature(&cpuid_features[id])); \
}
/*
* Define functions for user-space CPUID features testing
*/
CPUID_FEATURE_CHECK(sse, SSE);
CPUID_FEATURE_CHECK(sse2, SSE2);
CPUID_FEATURE_CHECK(sse3, SSE3);
CPUID_FEATURE_CHECK(ssse3, SSSE3);
CPUID_FEATURE_CHECK(sse4_1, SSE4_1);
CPUID_FEATURE_CHECK(sse4_2, SSE4_2);
CPUID_FEATURE_CHECK(avx, AVX);
CPUID_FEATURE_CHECK(avx2, AVX2);
CPUID_FEATURE_CHECK(osxsave, OSXSAVE);
CPUID_FEATURE_CHECK(bmi1, BMI1);
CPUID_FEATURE_CHECK(bmi2, BMI2);
CPUID_FEATURE_CHECK(avx512f, AVX512F);
CPUID_FEATURE_CHECK(avx512cd, AVX512CD);
CPUID_FEATURE_CHECK(avx512dq, AVX512DQ);
CPUID_FEATURE_CHECK(avx512bw, AVX512BW);
CPUID_FEATURE_CHECK(avx512ifma, AVX512IFMA);
CPUID_FEATURE_CHECK(avx512vbmi, AVX512VBMI);
CPUID_FEATURE_CHECK(avx512pf, AVX512PF);
CPUID_FEATURE_CHECK(avx512er, AVX512ER);
CPUID_FEATURE_CHECK(avx512vl, AVX512VL);
CPUID_FEATURE_CHECK(aes, AES);
CPUID_FEATURE_CHECK(pclmulqdq, PCLMULQDQ);
ICP: Improve AES-GCM performance Currently SIMD accelerated AES-GCM performance is limited by two factors: a. The need to disable preemption and interrupts and save the FPU state before using it and to do the reverse when done. Due to the way the code is organized (see (b) below) we have to pay this price twice for each 16 byte GCM block processed. b. Most processing is done in C, operating on single GCM blocks. The use of SIMD instructions is limited to the AES encryption of the counter block (AES-NI) and the Galois multiplication (PCLMULQDQ). This leads to the FPU not being fully utilized for crypto operations. To solve (a) we do crypto processing in larger chunks while owning the FPU. An `icp_gcm_avx_chunk_size` module parameter was introduced to make this chunk size tweakable. It defaults to 32 KiB. This step alone roughly doubles performance. (b) is tackled by porting and using the highly optimized openssl AES-GCM assembler routines, which do all the processing (CTR, AES, GMULT) in a single routine. Both steps together result in up to 32x reduction of the time spend in the en/decryption routines, leading up to approximately 12x throughput increase for large (128 KiB) blocks. Lastly, this commit changes the default encryption algorithm from AES-CCM to AES-GCM when setting the `encryption=on` property. Reviewed-By: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-By: Jason King <jason.king@joyent.com> Reviewed-By: Tom Caputi <tcaputi@datto.com> Reviewed-By: Richard Laager <rlaager@wiktel.com> Signed-off-by: Attila Fülöp <attila@fueloep.org> Closes #9749
2020-02-10 23:59:50 +03:00
CPUID_FEATURE_CHECK(movbe, MOVBE);
/*
* Detect register set support
*/
static inline boolean_t
__simd_state_enabled(const uint64_t state)
{
boolean_t has_osxsave;
uint64_t xcr0;
has_osxsave = __cpuid_has_osxsave();
if (!has_osxsave)
return (B_FALSE);
xcr0 = xgetbv(0);
return ((xcr0 & state) == state);
}
#define _XSTATE_SSE_AVX (0x2 | 0x4)
#define _XSTATE_AVX512 (0xE0 | _XSTATE_SSE_AVX)
#define __ymm_enabled() __simd_state_enabled(_XSTATE_SSE_AVX)
#define __zmm_enabled() __simd_state_enabled(_XSTATE_AVX512)
/*
* Check if SSE instruction set is available
*/
static inline boolean_t
zfs_sse_available(void)
{
return (__cpuid_has_sse());
}
/*
* Check if SSE2 instruction set is available
*/
static inline boolean_t
zfs_sse2_available(void)
{
return (__cpuid_has_sse2());
}
/*
* Check if SSE3 instruction set is available
*/
static inline boolean_t
zfs_sse3_available(void)
{
return (__cpuid_has_sse3());
}
/*
* Check if SSSE3 instruction set is available
*/
static inline boolean_t
zfs_ssse3_available(void)
{
return (__cpuid_has_ssse3());
}
/*
* Check if SSE4.1 instruction set is available
*/
static inline boolean_t
zfs_sse4_1_available(void)
{
return (__cpuid_has_sse4_1());
}
/*
* Check if SSE4.2 instruction set is available
*/
static inline boolean_t
zfs_sse4_2_available(void)
{
return (__cpuid_has_sse4_2());
}
/*
* Check if AVX instruction set is available
*/
static inline boolean_t
zfs_avx_available(void)
{
return (__cpuid_has_avx() && __ymm_enabled());
}
/*
* Check if AVX2 instruction set is available
*/
static inline boolean_t
zfs_avx2_available(void)
{
return (__cpuid_has_avx2() && __ymm_enabled());
}
/*
* Check if BMI1 instruction set is available
*/
static inline boolean_t
zfs_bmi1_available(void)
{
return (__cpuid_has_bmi1());
}
/*
* Check if BMI2 instruction set is available
*/
static inline boolean_t
zfs_bmi2_available(void)
{
return (__cpuid_has_bmi2());
}
/*
* Check if AES instruction set is available
*/
static inline boolean_t
zfs_aes_available(void)
{
return (__cpuid_has_aes());
}
/*
* Check if PCLMULQDQ instruction set is available
*/
static inline boolean_t
zfs_pclmulqdq_available(void)
{
return (__cpuid_has_pclmulqdq());
}
ICP: Improve AES-GCM performance Currently SIMD accelerated AES-GCM performance is limited by two factors: a. The need to disable preemption and interrupts and save the FPU state before using it and to do the reverse when done. Due to the way the code is organized (see (b) below) we have to pay this price twice for each 16 byte GCM block processed. b. Most processing is done in C, operating on single GCM blocks. The use of SIMD instructions is limited to the AES encryption of the counter block (AES-NI) and the Galois multiplication (PCLMULQDQ). This leads to the FPU not being fully utilized for crypto operations. To solve (a) we do crypto processing in larger chunks while owning the FPU. An `icp_gcm_avx_chunk_size` module parameter was introduced to make this chunk size tweakable. It defaults to 32 KiB. This step alone roughly doubles performance. (b) is tackled by porting and using the highly optimized openssl AES-GCM assembler routines, which do all the processing (CTR, AES, GMULT) in a single routine. Both steps together result in up to 32x reduction of the time spend in the en/decryption routines, leading up to approximately 12x throughput increase for large (128 KiB) blocks. Lastly, this commit changes the default encryption algorithm from AES-CCM to AES-GCM when setting the `encryption=on` property. Reviewed-By: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-By: Jason King <jason.king@joyent.com> Reviewed-By: Tom Caputi <tcaputi@datto.com> Reviewed-By: Richard Laager <rlaager@wiktel.com> Signed-off-by: Attila Fülöp <attila@fueloep.org> Closes #9749
2020-02-10 23:59:50 +03:00
/*
* Check if MOVBE instruction is available
*/
static inline boolean_t
zfs_movbe_available(void)
{
return (__cpuid_has_movbe());
}
/*
* AVX-512 family of instruction sets:
*
* AVX512F Foundation
* AVX512CD Conflict Detection Instructions
* AVX512ER Exponential and Reciprocal Instructions
* AVX512PF Prefetch Instructions
*
* AVX512BW Byte and Word Instructions
* AVX512DQ Double-word and Quadword Instructions
* AVX512VL Vector Length Extensions
*
* AVX512IFMA Integer Fused Multiply Add (Not supported by kernel 4.4)
* AVX512VBMI Vector Byte Manipulation Instructions
*/
/*
* Check if AVX512F instruction set is available
*/
static inline boolean_t
zfs_avx512f_available(void)
{
return (__cpuid_has_avx512f() && __zmm_enabled());
}
/*
* Check if AVX512CD instruction set is available
*/
static inline boolean_t
zfs_avx512cd_available(void)
{
return (__cpuid_has_avx512cd() && __zmm_enabled());
}
/*
* Check if AVX512ER instruction set is available
*/
static inline boolean_t
zfs_avx512er_available(void)
{
return (__cpuid_has_avx512er() && __zmm_enabled());
}
/*
* Check if AVX512PF instruction set is available
*/
static inline boolean_t
zfs_avx512pf_available(void)
{
return (__cpuid_has_avx512pf() && __zmm_enabled());
}
/*
* Check if AVX512BW instruction set is available
*/
static inline boolean_t
zfs_avx512bw_available(void)
{
return (__cpuid_has_avx512bw() && __zmm_enabled());
}
/*
* Check if AVX512DQ instruction set is available
*/
static inline boolean_t
zfs_avx512dq_available(void)
{
return (__cpuid_has_avx512dq() && __zmm_enabled());
}
/*
* Check if AVX512VL instruction set is available
*/
static inline boolean_t
zfs_avx512vl_available(void)
{
return (__cpuid_has_avx512vl() && __zmm_enabled());
}
/*
* Check if AVX512IFMA instruction set is available
*/
static inline boolean_t
zfs_avx512ifma_available(void)
{
return (__cpuid_has_avx512ifma() && __zmm_enabled());
}
/*
* Check if AVX512VBMI instruction set is available
*/
static inline boolean_t
zfs_avx512vbmi_available(void)
{
return (__cpuid_has_avx512f() && __cpuid_has_avx512vbmi() &&
__zmm_enabled());
}
#elif defined(__aarch64__)
#define kfpu_allowed() 1
#define kfpu_initialize(tsk) do {} while (0)
#define kfpu_begin() do {} while (0)
#define kfpu_end() do {} while (0)
#elif defined(__powerpc__)
/* including <sys/auxv.h> clashes with AT_UID and others */
#if defined(__FreeBSD__)
#define AT_HWCAP 25 /* CPU feature flags. */
#define AT_HWCAP2 26 /* CPU feature flags 2. */
extern int elf_aux_info(int aux, void *buf, int buflen);
static inline unsigned long
getauxval(unsigned long key)
{
unsigned long val = 0UL;
if (elf_aux_info((int)key, &val, sizeof (val)) != 0)
return (0UL);
return (val);
}
#elif defined(__linux__)
#define AT_HWCAP 16 /* CPU feature flags. */
#define AT_HWCAP2 26 /* CPU feature flags 2. */
extern unsigned long getauxval(unsigned long type);
#endif
#define kfpu_allowed() 1
#define kfpu_initialize(tsk) do {} while (0)
#define kfpu_begin() do {} while (0)
#define kfpu_end() do {} while (0)
#define PPC_FEATURE_HAS_ALTIVEC 0x10000000
static inline boolean_t
zfs_altivec_available(void)
{
unsigned long hwcap = getauxval(AT_HWCAP);
return (hwcap & PPC_FEATURE_HAS_ALTIVEC);
}
#define PPC_FEATURE_HAS_VSX 0x00000080
static inline boolean_t
zfs_vsx_available(void)
{
unsigned long hwcap = getauxval(AT_HWCAP);
return (hwcap & PPC_FEATURE_HAS_VSX);
}
#define PPC_FEATURE2_ARCH_2_07 0x80000000
Introduce BLAKE3 checksums as an OpenZFS feature This commit adds BLAKE3 checksums to OpenZFS, it has similar performance to Edon-R, but without the caveats around the latter. Homepage of BLAKE3: https://github.com/BLAKE3-team/BLAKE3 Wikipedia: https://en.wikipedia.org/wiki/BLAKE_(hash_function)#BLAKE3 Short description of Wikipedia: BLAKE3 is a cryptographic hash function based on Bao and BLAKE2, created by Jack O'Connor, Jean-Philippe Aumasson, Samuel Neves, and Zooko Wilcox-O'Hearn. It was announced on January 9, 2020, at Real World Crypto. BLAKE3 is a single algorithm with many desirable features (parallelism, XOF, KDF, PRF and MAC), in contrast to BLAKE and BLAKE2, which are algorithm families with multiple variants. BLAKE3 has a binary tree structure, so it supports a practically unlimited degree of parallelism (both SIMD and multithreading) given enough input. The official Rust and C implementations are dual-licensed as public domain (CC0) and the Apache License. Along with adding the BLAKE3 hash into the OpenZFS infrastructure a new benchmarking file called chksum_bench was introduced. When read it reports the speed of the available checksum functions. On Linux: cat /proc/spl/kstat/zfs/chksum_bench On FreeBSD: sysctl kstat.zfs.misc.chksum_bench This is an example output of an i3-1005G1 test system with Debian 11: implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 1196 1602 1761 1749 1762 1759 1751 skein-generic 546 591 608 615 619 612 616 sha256-generic 240 300 316 314 304 285 276 sha512-generic 353 441 467 476 472 467 426 blake3-generic 308 313 313 313 312 313 312 blake3-sse2 402 1289 1423 1446 1432 1458 1413 blake3-sse41 427 1470 1625 1704 1679 1607 1629 blake3-avx2 428 1920 3095 3343 3356 3318 3204 blake3-avx512 473 2687 4905 5836 5844 5643 5374 Output on Debian 5.10.0-10-amd64 system: (Ryzen 7 5800X) implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 1840 2458 2665 2719 2711 2723 2693 skein-generic 870 966 996 992 1003 1005 1009 sha256-generic 415 442 453 455 457 457 457 sha512-generic 608 690 711 718 719 720 721 blake3-generic 301 313 311 309 309 310 310 blake3-sse2 343 1865 2124 2188 2180 2181 2186 blake3-sse41 364 2091 2396 2509 2463 2482 2488 blake3-avx2 365 2590 4399 4971 4915 4802 4764 Output on Debian 5.10.0-9-powerpc64le system: (POWER 9) implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 1213 1703 1889 1918 1957 1902 1907 skein-generic 434 492 520 522 511 525 525 sha256-generic 167 183 187 188 188 187 188 sha512-generic 186 216 222 221 225 224 224 blake3-generic 153 152 154 153 151 153 153 blake3-sse2 391 1170 1366 1406 1428 1426 1414 blake3-sse41 352 1049 1212 1174 1262 1258 1259 Output on Debian 5.10.0-11-arm64 system: (Pi400) implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 487 603 629 639 643 641 641 skein-generic 271 299 303 308 309 309 307 sha256-generic 117 127 128 130 130 129 130 sha512-generic 145 165 170 172 173 174 175 blake3-generic 81 29 71 89 89 89 89 blake3-sse2 112 323 368 379 380 371 374 blake3-sse41 101 315 357 368 369 364 360 Structurally, the new code is mainly split into these parts: - 1x cross platform generic c variant: blake3_generic.c - 4x assembly for X86-64 (SSE2, SSE4.1, AVX2, AVX512) - 2x assembly for ARMv8 (NEON converted from SSE2) - 2x assembly for PPC64-LE (POWER8 converted from SSE2) - one file for switching between the implementations Note the PPC64 assembly requires the VSX instruction set and the kfpu_begin() / kfpu_end() calls on PowerPC were updated accordingly. Reviewed-by: Felix Dörre <felix@dogcraft.de> Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tino Reichardt <milky-zfs@mcmilk.de> Co-authored-by: Rich Ercolani <rincebrain@gmail.com> Closes #10058 Closes #12918
2022-06-09 01:55:57 +03:00
static inline boolean_t
zfs_isa207_available(void)
Introduce BLAKE3 checksums as an OpenZFS feature This commit adds BLAKE3 checksums to OpenZFS, it has similar performance to Edon-R, but without the caveats around the latter. Homepage of BLAKE3: https://github.com/BLAKE3-team/BLAKE3 Wikipedia: https://en.wikipedia.org/wiki/BLAKE_(hash_function)#BLAKE3 Short description of Wikipedia: BLAKE3 is a cryptographic hash function based on Bao and BLAKE2, created by Jack O'Connor, Jean-Philippe Aumasson, Samuel Neves, and Zooko Wilcox-O'Hearn. It was announced on January 9, 2020, at Real World Crypto. BLAKE3 is a single algorithm with many desirable features (parallelism, XOF, KDF, PRF and MAC), in contrast to BLAKE and BLAKE2, which are algorithm families with multiple variants. BLAKE3 has a binary tree structure, so it supports a practically unlimited degree of parallelism (both SIMD and multithreading) given enough input. The official Rust and C implementations are dual-licensed as public domain (CC0) and the Apache License. Along with adding the BLAKE3 hash into the OpenZFS infrastructure a new benchmarking file called chksum_bench was introduced. When read it reports the speed of the available checksum functions. On Linux: cat /proc/spl/kstat/zfs/chksum_bench On FreeBSD: sysctl kstat.zfs.misc.chksum_bench This is an example output of an i3-1005G1 test system with Debian 11: implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 1196 1602 1761 1749 1762 1759 1751 skein-generic 546 591 608 615 619 612 616 sha256-generic 240 300 316 314 304 285 276 sha512-generic 353 441 467 476 472 467 426 blake3-generic 308 313 313 313 312 313 312 blake3-sse2 402 1289 1423 1446 1432 1458 1413 blake3-sse41 427 1470 1625 1704 1679 1607 1629 blake3-avx2 428 1920 3095 3343 3356 3318 3204 blake3-avx512 473 2687 4905 5836 5844 5643 5374 Output on Debian 5.10.0-10-amd64 system: (Ryzen 7 5800X) implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 1840 2458 2665 2719 2711 2723 2693 skein-generic 870 966 996 992 1003 1005 1009 sha256-generic 415 442 453 455 457 457 457 sha512-generic 608 690 711 718 719 720 721 blake3-generic 301 313 311 309 309 310 310 blake3-sse2 343 1865 2124 2188 2180 2181 2186 blake3-sse41 364 2091 2396 2509 2463 2482 2488 blake3-avx2 365 2590 4399 4971 4915 4802 4764 Output on Debian 5.10.0-9-powerpc64le system: (POWER 9) implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 1213 1703 1889 1918 1957 1902 1907 skein-generic 434 492 520 522 511 525 525 sha256-generic 167 183 187 188 188 187 188 sha512-generic 186 216 222 221 225 224 224 blake3-generic 153 152 154 153 151 153 153 blake3-sse2 391 1170 1366 1406 1428 1426 1414 blake3-sse41 352 1049 1212 1174 1262 1258 1259 Output on Debian 5.10.0-11-arm64 system: (Pi400) implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 487 603 629 639 643 641 641 skein-generic 271 299 303 308 309 309 307 sha256-generic 117 127 128 130 130 129 130 sha512-generic 145 165 170 172 173 174 175 blake3-generic 81 29 71 89 89 89 89 blake3-sse2 112 323 368 379 380 371 374 blake3-sse41 101 315 357 368 369 364 360 Structurally, the new code is mainly split into these parts: - 1x cross platform generic c variant: blake3_generic.c - 4x assembly for X86-64 (SSE2, SSE4.1, AVX2, AVX512) - 2x assembly for ARMv8 (NEON converted from SSE2) - 2x assembly for PPC64-LE (POWER8 converted from SSE2) - one file for switching between the implementations Note the PPC64 assembly requires the VSX instruction set and the kfpu_begin() / kfpu_end() calls on PowerPC were updated accordingly. Reviewed-by: Felix Dörre <felix@dogcraft.de> Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tino Reichardt <milky-zfs@mcmilk.de> Co-authored-by: Rich Ercolani <rincebrain@gmail.com> Closes #10058 Closes #12918
2022-06-09 01:55:57 +03:00
{
unsigned long hwcap = getauxval(AT_HWCAP);
unsigned long hwcap2 = getauxval(AT_HWCAP2);
return ((hwcap & PPC_FEATURE_HAS_VSX) &&
(hwcap2 & PPC_FEATURE2_ARCH_2_07));
Introduce BLAKE3 checksums as an OpenZFS feature This commit adds BLAKE3 checksums to OpenZFS, it has similar performance to Edon-R, but without the caveats around the latter. Homepage of BLAKE3: https://github.com/BLAKE3-team/BLAKE3 Wikipedia: https://en.wikipedia.org/wiki/BLAKE_(hash_function)#BLAKE3 Short description of Wikipedia: BLAKE3 is a cryptographic hash function based on Bao and BLAKE2, created by Jack O'Connor, Jean-Philippe Aumasson, Samuel Neves, and Zooko Wilcox-O'Hearn. It was announced on January 9, 2020, at Real World Crypto. BLAKE3 is a single algorithm with many desirable features (parallelism, XOF, KDF, PRF and MAC), in contrast to BLAKE and BLAKE2, which are algorithm families with multiple variants. BLAKE3 has a binary tree structure, so it supports a practically unlimited degree of parallelism (both SIMD and multithreading) given enough input. The official Rust and C implementations are dual-licensed as public domain (CC0) and the Apache License. Along with adding the BLAKE3 hash into the OpenZFS infrastructure a new benchmarking file called chksum_bench was introduced. When read it reports the speed of the available checksum functions. On Linux: cat /proc/spl/kstat/zfs/chksum_bench On FreeBSD: sysctl kstat.zfs.misc.chksum_bench This is an example output of an i3-1005G1 test system with Debian 11: implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 1196 1602 1761 1749 1762 1759 1751 skein-generic 546 591 608 615 619 612 616 sha256-generic 240 300 316 314 304 285 276 sha512-generic 353 441 467 476 472 467 426 blake3-generic 308 313 313 313 312 313 312 blake3-sse2 402 1289 1423 1446 1432 1458 1413 blake3-sse41 427 1470 1625 1704 1679 1607 1629 blake3-avx2 428 1920 3095 3343 3356 3318 3204 blake3-avx512 473 2687 4905 5836 5844 5643 5374 Output on Debian 5.10.0-10-amd64 system: (Ryzen 7 5800X) implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 1840 2458 2665 2719 2711 2723 2693 skein-generic 870 966 996 992 1003 1005 1009 sha256-generic 415 442 453 455 457 457 457 sha512-generic 608 690 711 718 719 720 721 blake3-generic 301 313 311 309 309 310 310 blake3-sse2 343 1865 2124 2188 2180 2181 2186 blake3-sse41 364 2091 2396 2509 2463 2482 2488 blake3-avx2 365 2590 4399 4971 4915 4802 4764 Output on Debian 5.10.0-9-powerpc64le system: (POWER 9) implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 1213 1703 1889 1918 1957 1902 1907 skein-generic 434 492 520 522 511 525 525 sha256-generic 167 183 187 188 188 187 188 sha512-generic 186 216 222 221 225 224 224 blake3-generic 153 152 154 153 151 153 153 blake3-sse2 391 1170 1366 1406 1428 1426 1414 blake3-sse41 352 1049 1212 1174 1262 1258 1259 Output on Debian 5.10.0-11-arm64 system: (Pi400) implementation 1k 4k 16k 64k 256k 1m 4m edonr-generic 487 603 629 639 643 641 641 skein-generic 271 299 303 308 309 309 307 sha256-generic 117 127 128 130 130 129 130 sha512-generic 145 165 170 172 173 174 175 blake3-generic 81 29 71 89 89 89 89 blake3-sse2 112 323 368 379 380 371 374 blake3-sse41 101 315 357 368 369 364 360 Structurally, the new code is mainly split into these parts: - 1x cross platform generic c variant: blake3_generic.c - 4x assembly for X86-64 (SSE2, SSE4.1, AVX2, AVX512) - 2x assembly for ARMv8 (NEON converted from SSE2) - 2x assembly for PPC64-LE (POWER8 converted from SSE2) - one file for switching between the implementations Note the PPC64 assembly requires the VSX instruction set and the kfpu_begin() / kfpu_end() calls on PowerPC were updated accordingly. Reviewed-by: Felix Dörre <felix@dogcraft.de> Reviewed-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tino Reichardt <milky-zfs@mcmilk.de> Co-authored-by: Rich Ercolani <rincebrain@gmail.com> Closes #10058 Closes #12918
2022-06-09 01:55:57 +03:00
}
#else
#define kfpu_allowed() 0
#define kfpu_initialize(tsk) do {} while (0)
#define kfpu_begin() do {} while (0)
#define kfpu_end() do {} while (0)
#endif
#endif /* _LIBSPL_SYS_SIMD_H */