Add PPC cpu feature tests for FreeBSD and Linux

Add needed cpu feature tests for powerpc architecture.

Overview:
zfs_altivec_available() - needed by RAID-Z
zfs_vsx_available()     - needed by BLAKE3
zfs_isa207_available()  - needed by SHA2

Part 1 - Userspace
- use getauxval() for Linux and elf_aux_info() for FreeBSD
- direct including <sys/auxv.h> fails with double definitions
- so we self define the needed functions and definitions

Part 2 - Kernel space FreeBSD
- use exported cpu_features of <powerpc/cpu.h>

Part 3 - Kernel space Linux
- use cpu_has_feature() function of <asm/cpufeature.h>

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Tino Reichardt <milky-zfs@mcmilk.de>
Closes #13725
This commit is contained in:
Tino Reichardt
2022-09-07 20:33:59 +02:00
committed by Brian Behlendorf
parent eeca9d27d7
commit 48cf170d5a
6 changed files with 204 additions and 117 deletions
+1
View File
@@ -50,6 +50,7 @@ noinst_HEADERS = \
%D%/spl/sys/sid.h \
%D%/spl/sys/sig.h \
%D%/spl/sys/simd.h \
%D%/spl/sys/simd_powerpc.h \
%D%/spl/sys/simd_x86.h \
%D%/spl/sys/spl_condvar.h \
%D%/spl/sys/string.h \
+6 -2
View File
@@ -26,13 +26,16 @@
* $FreeBSD$
*/
#ifndef _FREEBSD_SIMD_H
#define _FREEBSD_SIMD_H
#if defined(__amd64__) || defined(__i386__)
#include <sys/simd_x86.h>
#else
#elif defined(__powerpc__)
#include <sys/simd_powerpc.h>
#else
#define kfpu_allowed() 0
#define kfpu_initialize(tsk) do {} while (0)
#define kfpu_begin() do {} while (0)
@@ -40,4 +43,5 @@
#define kfpu_init() (0)
#define kfpu_fini() do {} while (0)
#endif
#endif
+90
View File
@@ -0,0 +1,90 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (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) 2022 Tino Reichardt <milky-zfs@mcmilk.de>
*/
/*
* USER API:
*
* Kernel fpu methods:
* kfpu_allowed()
* kfpu_begin()
* kfpu_end()
* kfpu_init()
* kfpu_fini()
*
* SIMD support:
*
* Following functions should be called to determine whether CPU feature
* is supported. All functions are usable in kernel and user space.
* If a SIMD algorithm is using more than one instruction set
* all relevant feature test functions should be called.
*
* Supported features:
* zfs_altivec_available()
* zfs_vsx_available()
* zfs_isa207_available()
*/
#ifndef _FREEBSD_SIMD_POWERPC_H
#define _FREEBSD_SIMD_POWERPC_H
#include <sys/types.h>
#include <sys/cdefs.h>
#include <machine/pcb.h>
#include <powerpc/cpu.h>
#define kfpu_allowed() 1
#define kfpu_initialize(tsk) do {} while (0)
#define kfpu_begin() do {} while (0)
#define kfpu_end() do {} while (0)
#define kfpu_init() (0)
#define kfpu_fini() do {} while (0)
/*
* Check if Altivec is available
*/
static inline boolean_t
zfs_altivec_available(void)
{
return ((cpu_features & PPC_FEATURE_HAS_ALTIVEC) != 0);
}
/*
* Check if VSX is available
*/
static inline boolean_t
zfs_vsx_available(void)
{
return ((cpu_features & PPC_FEATURE_HAS_VSX) != 0);
}
/*
* Check if POWER ISA 2.07 is available (SHA2)
*/
static inline boolean_t
zfs_isa207_available(void)
{
return ((cpu_features2 & PPC_FEATURE2_ARCH_2_07) != 0);
}
+25 -25
View File
@@ -77,7 +77,7 @@ __simd_state_enabled(const uint64_t state)
boolean_t has_osxsave;
uint64_t xcr0;
has_osxsave = !!(cpu_feature2 & CPUID2_OSXSAVE);
has_osxsave = (cpu_feature2 & CPUID2_OSXSAVE) != 0;
if (!has_osxsave)
return (B_FALSE);
@@ -99,7 +99,7 @@ __simd_state_enabled(const uint64_t state)
static inline boolean_t
zfs_sse_available(void)
{
return (!!(cpu_feature & CPUID_SSE));
return ((cpu_feature & CPUID_SSE) != 0);
}
/*
@@ -108,7 +108,7 @@ zfs_sse_available(void)
static inline boolean_t
zfs_sse2_available(void)
{
return (!!(cpu_feature & CPUID_SSE2));
return ((cpu_feature & CPUID_SSE2) != 0);
}
/*
@@ -117,7 +117,7 @@ zfs_sse2_available(void)
static inline boolean_t
zfs_sse3_available(void)
{
return (!!(cpu_feature2 & CPUID2_SSE3));
return ((cpu_feature2 & CPUID2_SSE3) != 0);
}
/*
@@ -126,7 +126,7 @@ zfs_sse3_available(void)
static inline boolean_t
zfs_ssse3_available(void)
{
return (!!(cpu_feature2 & CPUID2_SSSE3));
return ((cpu_feature2 & CPUID2_SSSE3) != 0);
}
/*
@@ -135,7 +135,7 @@ zfs_ssse3_available(void)
static inline boolean_t
zfs_sse4_1_available(void)
{
return (!!(cpu_feature2 & CPUID2_SSE41));
return ((cpu_feature2 & CPUID2_SSE41) != 0);
}
/*
@@ -144,7 +144,7 @@ zfs_sse4_1_available(void)
static inline boolean_t
zfs_sse4_2_available(void)
{
return (!!(cpu_feature2 & CPUID2_SSE42));
return ((cpu_feature2 & CPUID2_SSE42) != 0);
}
/*
@@ -155,7 +155,7 @@ zfs_avx_available(void)
{
boolean_t has_avx;
has_avx = !!(cpu_feature2 & CPUID2_AVX);
has_avx = (cpu_feature2 & CPUID2_AVX) != 0;
return (has_avx && __ymm_enabled());
}
@@ -168,7 +168,7 @@ zfs_avx2_available(void)
{
boolean_t has_avx2;
has_avx2 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX2);
has_avx2 = (cpu_stdext_feature & CPUID_STDEXT_AVX2) != 0;
return (has_avx2 && __ymm_enabled());
}
@@ -196,7 +196,7 @@ zfs_avx512f_available(void)
{
boolean_t has_avx512;
has_avx512 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX512F);
has_avx512 = (cpu_stdext_feature & CPUID_STDEXT_AVX512F) != 0;
return (has_avx512 && __zmm_enabled());
}
@@ -207,8 +207,8 @@ zfs_avx512cd_available(void)
{
boolean_t has_avx512;
has_avx512 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX512F) &&
!!(cpu_stdext_feature & CPUID_STDEXT_AVX512CD);
has_avx512 = (cpu_stdext_feature & CPUID_STDEXT_AVX512F) != 0 &&
(cpu_stdext_feature & CPUID_STDEXT_AVX512CD) != 0;
return (has_avx512 && __zmm_enabled());
}
@@ -219,8 +219,8 @@ zfs_avx512er_available(void)
{
boolean_t has_avx512;
has_avx512 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX512F) &&
!!(cpu_stdext_feature & CPUID_STDEXT_AVX512CD);
has_avx512 = (cpu_stdext_feature & CPUID_STDEXT_AVX512F) != 0 &&
(cpu_stdext_feature & CPUID_STDEXT_AVX512CD) != 0;
return (has_avx512 && __zmm_enabled());
}
@@ -231,8 +231,8 @@ zfs_avx512pf_available(void)
{
boolean_t has_avx512;
has_avx512 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX512F) &&
!!(cpu_stdext_feature & CPUID_STDEXT_AVX512PF);
has_avx512 = (cpu_stdext_feature & CPUID_STDEXT_AVX512F) != 0 &&
(cpu_stdext_feature & CPUID_STDEXT_AVX512PF) != 0;
return (has_avx512 && __zmm_enabled());
}
@@ -243,7 +243,7 @@ zfs_avx512bw_available(void)
{
boolean_t has_avx512 = B_FALSE;
has_avx512 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX512BW);
has_avx512 = (cpu_stdext_feature & CPUID_STDEXT_AVX512BW) != 0;
return (has_avx512 && __zmm_enabled());
}
@@ -254,8 +254,8 @@ zfs_avx512dq_available(void)
{
boolean_t has_avx512;
has_avx512 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX512F) &&
!!(cpu_stdext_feature & CPUID_STDEXT_AVX512DQ);
has_avx512 = (cpu_stdext_feature & CPUID_STDEXT_AVX512F) != 0 &&
(cpu_stdext_feature & CPUID_STDEXT_AVX512DQ) != 0;
return (has_avx512 && __zmm_enabled());
}
@@ -266,8 +266,8 @@ zfs_avx512vl_available(void)
{
boolean_t has_avx512;
has_avx512 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX512F) &&
!!(cpu_stdext_feature & CPUID_STDEXT_AVX512VL);
has_avx512 = (cpu_stdext_feature & CPUID_STDEXT_AVX512F) != 0 &&
(cpu_stdext_feature & CPUID_STDEXT_AVX512VL) != 0;
return (has_avx512 && __zmm_enabled());
}
@@ -278,8 +278,8 @@ zfs_avx512ifma_available(void)
{
boolean_t has_avx512;
has_avx512 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX512F) &&
!!(cpu_stdext_feature & CPUID_STDEXT_AVX512IFMA);
has_avx512 = (cpu_stdext_feature & CPUID_STDEXT_AVX512F) != 0 &&
(cpu_stdext_feature & CPUID_STDEXT_AVX512IFMA) != 0;
return (has_avx512 && __zmm_enabled());
}
@@ -290,8 +290,8 @@ zfs_avx512vbmi_available(void)
{
boolean_t has_avx512;
has_avx512 = !!(cpu_stdext_feature & CPUID_STDEXT_AVX512F) &&
!!(cpu_stdext_feature & CPUID_STDEXT_BMI1);
has_avx512 = (cpu_stdext_feature & CPUID_STDEXT_AVX512F) != 0 &&
(cpu_stdext_feature & CPUID_STDEXT_BMI1) != 0;
return (has_avx512 && __zmm_enabled());
}