[PATCH v8 2/2] hash: unify crc32 selection for x86 and Arm

[Wang, Yipeng1]

> -----Original Message-----
> From: Pavan Nikhilesh <pbhagavatula at marvell.com>
> Sent: Friday, April 29, 2022 9:17 AM
> To: ruifeng.wang at arm.com; Wang, Yipeng1 <yipeng1.wang at intel.com>;
> Gobriel, Sameh <sameh.gobriel at intel.com>; Richardson, Bruce
> <bruce.richardson at intel.com>; Medvedkin, Vladimir
> <vladimir.medvedkin at intel.com>
> Cc: jerinj at marvell.com; dev at dpdk.org; Pavan Nikhilesh
> <pbhagavatula at marvell.com>
> Subject: [PATCH v8 2/2] hash: unify crc32 selection for x86 and Arm
> 
> Merge crc32 hash calculation public API implementation for x86 and Arm.
> Select the best available CRC32 algorithm when unsupported algorithm on a
> given CPU architecture is requested by an application.
> 
> Previously, if an application directly includes `rte_crc_arm64.h` without
> including `rte_hash_crc.h` it will fail to compile.
> 
> Signed-off-by: Pavan Nikhilesh <pbhagavatula at marvell.com>
> Reviewed-by: Ruifeng Wang <ruifeng.wang at arm.com>
> ---
>  lib/hash/meson.build       |   1 +
>  lib/hash/rte_crc_arm64.h   |  69 +++---------------
>  lib/hash/rte_crc_generic.h |  72 ++++++++++++++++++
>  lib/hash/rte_crc_x86.h     |  89 +++++++++++++++++++++++
>  lib/hash/rte_hash_crc.h    | 145 ++++++++++---------------------------
>  5 files changed, 210 insertions(+), 166 deletions(-)  create mode 100644
> lib/hash/rte_crc_generic.h
> 
> diff --git a/lib/hash/meson.build b/lib/hash/meson.build index
> ff13e2c7f9..b36c8b0c01 100644
> --- a/lib/hash/meson.build
> +++ b/lib/hash/meson.build
> @@ -13,6 +13,7 @@ indirect_headers += files(
>          'rte_crc_sw.h',
>          'rte_crc_x86.h',
>          'rte_crc_arm64.h',
> +        'rte_crc_generic.h',
>          'rte_thash_x86_gfni.h',
>  )
> 
> diff --git a/lib/hash/rte_crc_arm64.h b/lib/hash/rte_crc_arm64.h index
> b4628cfc09..172894335f 100644
> --- a/lib/hash/rte_crc_arm64.h
> +++ b/lib/hash/rte_crc_arm64.h
> @@ -2,23 +2,8 @@
>   * Copyright(c) 2015 Cavium, Inc
>   */
> 
> -#ifndef _RTE_CRC_ARM64_H_
> -#define _RTE_CRC_ARM64_H_
> -
> -/**
> - * @file
> - *
> - * RTE CRC arm64 Hash
> - */
> -
> -#ifdef __cplusplus
> -extern "C" {
> -#endif
> -
> -#include <stdint.h>
> -#include <rte_cpuflags.h>
> -#include <rte_branch_prediction.h>
> -#include <rte_common.h>
> +#ifndef _HASH_CRC_ARM64_H_
> +#define _HASH_CRC_ARM64_H_
> 
>  static inline uint32_t
>  crc32c_arm64_u8(uint8_t data, uint32_t init_val) @@ -61,40 +46,8 @@
> crc32c_arm64_u64(uint64_t data, uint32_t init_val)  }
> 
>  /**
> - * Allow or disallow use of arm64 SIMD instrinsics for CRC32 hash
> - * calculation.
> - *
> - * @param alg
> - *   An OR of following flags:
> - *   - (CRC32_SW) Don't use arm64 crc intrinsics
> - *   - (CRC32_ARM64) Use ARMv8 CRC intrinsic if available
> - *
> - */
> -static inline void
> -rte_hash_crc_set_alg(uint8_t alg)
> -{
> -	switch (alg) {
> -	case CRC32_ARM64:
> -		if (!rte_cpu_get_flag_enabled(RTE_CPUFLAG_CRC32))
> -			alg = CRC32_SW;
> -		/* fall-through */
> -	case CRC32_SW:
> -		crc32_alg = alg;
> -		/* fall-through */
> -	default:
> -		break;
> -	}
> -}
> -
> -/* Setting the best available algorithm */
> -RTE_INIT(rte_hash_crc_init_alg)
> -{
> -	rte_hash_crc_set_alg(CRC32_ARM64);
> -}
> -
> -/**
> - * Use single crc32 instruction to perform a hash on a 1 byte value.
> - * Fall back to software crc32 implementation in case arm64 crc intrinsics is
> + * Use single crc32 instruction to perform a hash on a byte value.
> + * Fall back to software crc32 implementation in case ARM CRC is
>   * not supported
>   *
>   * @param data
> @@ -115,7 +68,7 @@ rte_hash_crc_1byte(uint8_t data, uint32_t init_val)
> 
>  /**
>   * Use single crc32 instruction to perform a hash on a 2 bytes value.
> - * Fall back to software crc32 implementation in case arm64 crc intrinsics is
> + * Fall back to software crc32 implementation in case ARM CRC is
>   * not supported
>   *
>   * @param data
> @@ -136,7 +89,7 @@ rte_hash_crc_2byte(uint16_t data, uint32_t init_val)
> 
>  /**
>   * Use single crc32 instruction to perform a hash on a 4 byte value.
> - * Fall back to software crc32 implementation in case arm64 crc intrinsics is
> + * Fall back to software crc32 implementation in case ARM CRC is
>   * not supported
>   *
>   * @param data
> @@ -157,7 +110,7 @@ rte_hash_crc_4byte(uint32_t data, uint32_t init_val)
> 
>  /**
>   * Use single crc32 instruction to perform a hash on a 8 byte value.
> - * Fall back to software crc32 implementation in case arm64 crc intrinsics is
> + * Fall back to software crc32 implementation in case ARM CRC is
>   * not supported
>   *
>   * @param data
> @@ -170,14 +123,10 @@ rte_hash_crc_4byte(uint32_t data, uint32_t init_val)
> static inline uint32_t  rte_hash_crc_8byte(uint64_t data, uint32_t init_val)  {
> -	if (likely(crc32_alg == CRC32_ARM64))
> +	if (likely(crc32_alg & CRC32_ARM64))
>  		return crc32c_arm64_u64(data, init_val);
> 
>  	return crc32c_2words(data, init_val);
>  }
> 
> -#ifdef __cplusplus
> -}
> -#endif
> -
> -#endif /* _RTE_CRC_ARM64_H_ */
> +#endif /* _HASH_CRC_ARM64_H_ */
> diff --git a/lib/hash/rte_crc_generic.h b/lib/hash/rte_crc_generic.h new file
> mode 100644 index 0000000000..0c55947896
> --- /dev/null
> +++ b/lib/hash/rte_crc_generic.h
> @@ -0,0 +1,72 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2022 Marvell.
> + */
> +
> +#ifndef _HASH_CRC_GENERIC_H_
> +#define _HASH_CRC_GENERIC_H_
> +
> +/**
> + * Software crc32 implementation for 1 byte value.
> + *
> + * @param data
> + *   Data to perform hash on.
> + * @param init_val
> + *   Value to initialise hash generator.
> + * @return
> + *   32bit calculated hash value.
> + */
> +static inline uint32_t
> +rte_hash_crc_1byte(uint8_t data, uint32_t init_val) {
> +	return crc32c_1byte(data, init_val);
> +}
> +
> +/**
> + * Software crc32 implementation for 2 byte value.
> + *
> + * @param data
> + *   Data to perform hash on.
> + * @param init_val
> + *   Value to initialise hash generator.
> + * @return
> + *   32bit calculated hash value.
> + */
> +static inline uint32_t
> +rte_hash_crc_2byte(uint16_t data, uint32_t init_val) {
> +	return crc32c_2bytes(data, init_val);
> +}
> +
> +/**
> + * Software crc32 implementation for 4 byte value.
> + *
> + * @param data
> + *   Data to perform hash on.
> + * @param init_val
> + *   Value to initialise hash generator.
> + * @return
> + *   32bit calculated hash value.
> + */
> +static inline uint32_t
> +rte_hash_crc_4byte(uint32_t data, uint32_t init_val) {
> +	return crc32c_1word(data, init_val);
> +}
> +
> +/**
> + * Software crc32 implementation for 8 byte value.
> + *
> + * @param data
> + *   Data to perform hash on.
> + * @param init_val
> + *   Value to initialise hash generator.
> + * @return
> + *   32bit calculated hash value.
> + */
> +static inline uint32_t
> +rte_hash_crc_8byte(uint64_t data, uint32_t init_val) {
> +	return crc32c_2words(data, init_val);
> +}
> +
> +#endif
> diff --git a/lib/hash/rte_crc_x86.h b/lib/hash/rte_crc_x86.h index
> b80a742afa..19eb3584e7 100644
> --- a/lib/hash/rte_crc_x86.h
> +++ b/lib/hash/rte_crc_x86.h
> @@ -59,4 +59,93 @@ crc32c_sse42_u64(uint64_t data, uint64_t init_val)
>  	return (uint32_t)init_val;
>  }
> 
> +/**
> + * Use single crc32 instruction to perform a hash on a byte value.
> + * Fall back to software crc32 implementation in case SSE4.2 is
> + * not supported
> + *
> + * @param data
> + *   Data to perform hash on.
> + * @param init_val
> + *   Value to initialise hash generator.
> + * @return
> + *   32bit calculated hash value.
> + */
> +static inline uint32_t
> +rte_hash_crc_1byte(uint8_t data, uint32_t init_val) {
> +	if (likely(crc32_alg & CRC32_SSE42))
> +		return crc32c_sse42_u8(data, init_val);
> +
> +	return crc32c_1byte(data, init_val);
> +}
> +
> +/**
> + * Use single crc32 instruction to perform a hash on a 2 bytes value.
> + * Fall back to software crc32 implementation in case SSE4.2 is
> + * not supported
> + *
> + * @param data
> + *   Data to perform hash on.
> + * @param init_val
> + *   Value to initialise hash generator.
> + * @return
> + *   32bit calculated hash value.
> + */
> +static inline uint32_t
> +rte_hash_crc_2byte(uint16_t data, uint32_t init_val) {
> +	if (likely(crc32_alg & CRC32_SSE42))
> +		return crc32c_sse42_u16(data, init_val);
> +
> +	return crc32c_2bytes(data, init_val);
> +}
> +
> +/**
> + * Use single crc32 instruction to perform a hash on a 4 byte value.
> + * Fall back to software crc32 implementation in case SSE4.2 is
> + * not supported
> + *
> + * @param data
> + *   Data to perform hash on.
> + * @param init_val
> + *   Value to initialise hash generator.
> + * @return
> + *   32bit calculated hash value.
> + */
> +static inline uint32_t
> +rte_hash_crc_4byte(uint32_t data, uint32_t init_val) {
> +	if (likely(crc32_alg & CRC32_SSE42))
> +		return crc32c_sse42_u32(data, init_val);
> +
> +	return crc32c_1word(data, init_val);
> +}
> +
> +/**
> + * Use single crc32 instruction to perform a hash on a 8 byte value.
> + * Fall back to software crc32 implementation in case SSE4.2 is
> + * not supported
> + *
> + * @param data
> + *   Data to perform hash on.
> + * @param init_val
> + *   Value to initialise hash generator.
> + * @return
> + *   32bit calculated hash value.
> + */
> +static inline uint32_t
> +rte_hash_crc_8byte(uint64_t data, uint32_t init_val) { #ifdef
> +RTE_ARCH_X86_64
> +	if (likely(crc32_alg == CRC32_SSE42_x64))
> +		return crc32c_sse42_u64(data, init_val); #endif
> +
> +	if (likely(crc32_alg & CRC32_SSE42))
> +		return crc32c_sse42_u64_mimic(data, init_val);
> +
> +	return crc32c_2words(data, init_val);
> +}
> +
>  #endif
> diff --git a/lib/hash/rte_hash_crc.h b/lib/hash/rte_hash_crc.h index
> 308fdde414..cd9ec17082 100644
> --- a/lib/hash/rte_hash_crc.h
> +++ b/lib/hash/rte_hash_crc.h
> @@ -16,10 +16,12 @@ extern "C" {
>  #endif
> 
>  #include <stdint.h>
> -#include <rte_config.h>
> -#include <rte_cpuflags.h>
> +
>  #include <rte_branch_prediction.h>
>  #include <rte_common.h>
> +#include <rte_config.h>
> +#include <rte_cpuflags.h>
> +#include <rte_log.h>
> 
>  #include "rte_crc_sw.h"
> 
> @@ -33,136 +35,67 @@ static uint8_t crc32_alg = CRC32_SW;
> 
>  #if defined(RTE_ARCH_ARM64) && defined(__ARM_FEATURE_CRC32)
> #include "rte_crc_arm64.h"
> -#else
> +#elif defined(RTE_ARCH_X86)
>  #include "rte_crc_x86.h"
> +#else
> +#include "rte_crc_generic.h"
> +#endif
> 
>  /**
> - * Allow or disallow use of SSE4.2 instrinsics for CRC32 hash
> + * Allow or disallow use of SSE4.2/ARMv8 intrinsics for CRC32 hash
>   * calculation.
>   *
>   * @param alg
>   *   An OR of following flags:
> - *   - (CRC32_SW) Don't use SSE4.2 intrinsics
> + *   - (CRC32_SW) Don't use SSE4.2/ARMv8 intrinsics (default non-
> [x86/ARMv8])
>   *   - (CRC32_SSE42) Use SSE4.2 intrinsics if available
> - *   - (CRC32_SSE42_x64) Use 64-bit SSE4.2 intrinsic if available (default)
> + *   - (CRC32_SSE42_x64) Use 64-bit SSE4.2 intrinsic if available (default x86)
> + *   - (CRC32_ARM64) Use ARMv8 CRC intrinsic if available (default ARMv8)
>   *
>   */
>  static inline void
>  rte_hash_crc_set_alg(uint8_t alg)
>  {
> -#if defined(RTE_ARCH_X86)
> -	if (alg == CRC32_SSE42_x64 &&
> -			!rte_cpu_get_flag_enabled(RTE_CPUFLAG_EM64T))
> -		alg = CRC32_SSE42;
> -#endif
> -	crc32_alg = alg;
> -}
> +	crc32_alg = CRC32_SW;
> 
> -/* Setting the best available algorithm */
> -RTE_INIT(rte_hash_crc_init_alg)
> -{
> -	rte_hash_crc_set_alg(CRC32_SSE42_x64);
> -}
> +	if (alg == CRC32_SW)
> +		return;
> 
> -/**
> - * Use single crc32 instruction to perform a hash on a byte value.
> - * Fall back to software crc32 implementation in case SSE4.2 is
> - * not supported
> - *
> - * @param data
> - *   Data to perform hash on.
> - * @param init_val
> - *   Value to initialise hash generator.
> - * @return
> - *   32bit calculated hash value.
> - */
> -static inline uint32_t
> -rte_hash_crc_1byte(uint8_t data, uint32_t init_val) -{ -#if defined
> RTE_ARCH_X86
> -	if (likely(crc32_alg & CRC32_SSE42))
> -		return crc32c_sse42_u8(data, init_val);
> -#endif
> -
> -	return crc32c_1byte(data, init_val);
> -}
> -
> -/**
> - * Use single crc32 instruction to perform a hash on a 2 bytes value.
> - * Fall back to software crc32 implementation in case SSE4.2 is
> - * not supported
> - *
> - * @param data
> - *   Data to perform hash on.
> - * @param init_val
> - *   Value to initialise hash generator.
> - * @return
> - *   32bit calculated hash value.
> - */
> -static inline uint32_t
> -rte_hash_crc_2byte(uint16_t data, uint32_t init_val) -{  #if defined
> RTE_ARCH_X86
> -	if (likely(crc32_alg & CRC32_SSE42))
> -		return crc32c_sse42_u16(data, init_val);
> +	if (!(alg & CRC32_SSE42_x64))
> +		RTE_LOG(WARNING, HASH,
> +			"Unsupported CRC32 algorithm requested using
> CRC32_x64/CRC32_SSE42\n");
[Wang, Yipeng]
I have a question regarding this logic.
For the set_alg API, how about if user specify to use sse42 (not the 64bit version) algorithm on a em64 CPU, does it also warn "unsupported algorithm" and force user to use the x64 version?
It seems behaves differently than the current API definition. 

> +	if (!rte_cpu_get_flag_enabled(RTE_CPUFLAG_EM64T))
> +		crc32_alg = CRC32_SSE42;
> +	else
> +		crc32_alg = CRC32_SSE42_x64;
>  #endif
> 
> -	return crc32c_2bytes(data, init_val);
> -}
> -