iv/linux
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
linux/arch/powerpc/lib/memcpy_power7.S

642 lines
10 KiB
ArmAsm
Raw Normal View History

treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 156 Based on 1 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not write to the free software foundation inc 59 temple place suite 330 boston ma 02111 1307 usa extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 1334 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Richard Fontana <rfontana@redhat.com> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190527070033.113240726@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2019-05-27 08:55:05 +02:00
/* SPDX-License-Identifier: GPL-2.0-or-later */
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
/*
*
* Copyright (C) IBM Corporation, 2012
*
* Author: Anton Blanchard <anton@au.ibm.com>
*/
#include <asm/ppc_asm.h>
selftests/powerpc/64: Test all paths through copy routines The hand-coded assembler 64-bit copy routines include feature sections that select one code path or another depending on which CPU we are executing on. The self-tests for these copy routines end up testing just one path. This adds a mechanism for selecting any desired code path at compile time, and makes 2 or 3 versions of each test, each using a different code path, so as to cover all the possible paths. Signed-off-by: Paul Mackerras <paulus@ozlabs.org> [mpe: Add -mcpu=power4 to CFLAGS for older compilers] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2018-08-03 20:13:04 +10:00
#ifndef SELFTEST_CASE
/* 0 == don't use VMX, 1 == use VMX */
#define SELFTEST_CASE 0
#endif
powerpc: Fix endian issues in VMX copy loops Fix the permute loops for little endian. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2013-09-23 12:04:35 +10:00
#ifdef __BIG_ENDIAN__
#define LVS(VRT,RA,RB) lvsl VRT,RA,RB
#define VPERM(VRT,VRA,VRB,VRC) vperm VRT,VRA,VRB,VRC
#else
#define LVS(VRT,RA,RB) lvsr VRT,RA,RB
#define VPERM(VRT,VRA,VRB,VRC) vperm VRT,VRB,VRA,VRC
#endif
selftests/powerpc/64: Test all paths through copy routines The hand-coded assembler 64-bit copy routines include feature sections that select one code path or another depending on which CPU we are executing on. The self-tests for these copy routines end up testing just one path. This adds a mechanism for selecting any desired code path at compile time, and makes 2 or 3 versions of each test, each using a different code path, so as to cover all the possible paths. Signed-off-by: Paul Mackerras <paulus@ozlabs.org> [mpe: Add -mcpu=power4 to CFLAGS for older compilers] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2018-08-03 20:13:04 +10:00
_GLOBAL(memcpy_power7)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
cmpldi r5,16
cmpldi cr1,r5,4096
powerpc: Fix unsafe accesses to parameter area in ELFv2 Some of the assembler files in lib/ make use of the fact that in the ELFv1 ABI, the caller guarantees to provide stack space to save the parameter registers r3 ... r10. This guarantee is no longer present in ELFv2 for functions that have no variable argument list and no more than 8 arguments. Change the affected routines to temporarily store registers in the red zone and/or the top of their own stack frame (in the space provided to save r31 .. r29, which is actually not used in these routines). In opal_query_takeover, simply always allocate a stack frame; the routine is not performance critical. Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com> Signed-off-by: Anton Blanchard <anton@samba.org>
2014-02-14 19:21:03 +01:00
std r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
blt .Lshort_copy
selftests/powerpc/64: Test all paths through copy routines The hand-coded assembler 64-bit copy routines include feature sections that select one code path or another depending on which CPU we are executing on. The self-tests for these copy routines end up testing just one path. This adds a mechanism for selecting any desired code path at compile time, and makes 2 or 3 versions of each test, each using a different code path, so as to cover all the possible paths. Signed-off-by: Paul Mackerras <paulus@ozlabs.org> [mpe: Add -mcpu=power4 to CFLAGS for older compilers] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2018-08-03 20:13:04 +10:00
#ifdef CONFIG_ALTIVEC
test_feature = SELFTEST_CASE
BEGIN_FTR_SECTION
bgt cr1, .Lvmx_copy
END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
#endif
.Lnonvmx_copy:
/* Get the source 8B aligned */
neg r6,r4
mtocrf 0x01,r6
clrldi r6,r6,(64-3)
bf cr7*4+3,1f
lbz r0,0(r4)
addi r4,r4,1
stb r0,0(r3)
addi r3,r3,1
1: bf cr7*4+2,2f
lhz r0,0(r4)
addi r4,r4,2
sth r0,0(r3)
addi r3,r3,2
2: bf cr7*4+1,3f
lwz r0,0(r4)
addi r4,r4,4
stw r0,0(r3)
addi r3,r3,4
3: sub r5,r5,r6
cmpldi r5,128
blt 5f
mflr r0
stdu r1,-STACKFRAMESIZE(r1)
powerpc: Fix usage of register macros getting ready for %r0 change Anything that uses a constructed instruction (ie. from ppc-opcode.h), need to use the new R0 macro, as %r0 is not going to work. Also convert usages of macros where we are just determining an offset (usually for a load/store), like: std r14,STK_REG(r14)(r1) Can't use STK_REG(r14) as %r14 doesn't work in the STK_REG macro since it's just calculating an offset. Signed-off-by: Michael Neuling <mikey@neuling.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-06-25 13:33:10 +00:00
std r14,STK_REG(R14)(r1)
std r15,STK_REG(R15)(r1)
std r16,STK_REG(R16)(r1)
std r17,STK_REG(R17)(r1)
std r18,STK_REG(R18)(r1)
std r19,STK_REG(R19)(r1)
std r20,STK_REG(R20)(r1)
std r21,STK_REG(R21)(r1)
std r22,STK_REG(R22)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
std r0,STACKFRAMESIZE+16(r1)
srdi r6,r5,7
mtctr r6
/* Now do cacheline (128B) sized loads and stores. */
.align 5
4:
ld r0,0(r4)
ld r6,8(r4)
ld r7,16(r4)
ld r8,24(r4)
ld r9,32(r4)
ld r10,40(r4)
ld r11,48(r4)
ld r12,56(r4)
ld r14,64(r4)
ld r15,72(r4)
ld r16,80(r4)
ld r17,88(r4)
ld r18,96(r4)
ld r19,104(r4)
ld r20,112(r4)
ld r21,120(r4)
addi r4,r4,128
std r0,0(r3)
std r6,8(r3)
std r7,16(r3)
std r8,24(r3)
std r9,32(r3)
std r10,40(r3)
std r11,48(r3)
std r12,56(r3)
std r14,64(r3)
std r15,72(r3)
std r16,80(r3)
std r17,88(r3)
std r18,96(r3)
std r19,104(r3)
std r20,112(r3)
std r21,120(r3)
addi r3,r3,128
bdnz 4b
clrldi r5,r5,(64-7)
powerpc: Fix usage of register macros getting ready for %r0 change Anything that uses a constructed instruction (ie. from ppc-opcode.h), need to use the new R0 macro, as %r0 is not going to work. Also convert usages of macros where we are just determining an offset (usually for a load/store), like: std r14,STK_REG(r14)(r1) Can't use STK_REG(r14) as %r14 doesn't work in the STK_REG macro since it's just calculating an offset. Signed-off-by: Michael Neuling <mikey@neuling.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-06-25 13:33:10 +00:00
ld r14,STK_REG(R14)(r1)
ld r15,STK_REG(R15)(r1)
ld r16,STK_REG(R16)(r1)
ld r17,STK_REG(R17)(r1)
ld r18,STK_REG(R18)(r1)
ld r19,STK_REG(R19)(r1)
ld r20,STK_REG(R20)(r1)
ld r21,STK_REG(R21)(r1)
ld r22,STK_REG(R22)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r1,r1,STACKFRAMESIZE
/* Up to 127B to go */
5: srdi r6,r5,4
mtocrf 0x01,r6
6: bf cr7*4+1,7f
ld r0,0(r4)
ld r6,8(r4)
ld r7,16(r4)
ld r8,24(r4)
ld r9,32(r4)
ld r10,40(r4)
ld r11,48(r4)
ld r12,56(r4)
addi r4,r4,64
std r0,0(r3)
std r6,8(r3)
std r7,16(r3)
std r8,24(r3)
std r9,32(r3)
std r10,40(r3)
std r11,48(r3)
std r12,56(r3)
addi r3,r3,64
/* Up to 63B to go */
7: bf cr7*4+2,8f
ld r0,0(r4)
ld r6,8(r4)
ld r7,16(r4)
ld r8,24(r4)
addi r4,r4,32
std r0,0(r3)
std r6,8(r3)
std r7,16(r3)
std r8,24(r3)
addi r3,r3,32
/* Up to 31B to go */
8: bf cr7*4+3,9f
ld r0,0(r4)
ld r6,8(r4)
addi r4,r4,16
std r0,0(r3)
std r6,8(r3)
addi r3,r3,16
9: clrldi r5,r5,(64-4)
/* Up to 15B to go */
.Lshort_copy:
mtocrf 0x01,r5
bf cr7*4+0,12f
lwz r0,0(r4) /* Less chance of a reject with word ops */
lwz r6,4(r4)
addi r4,r4,8
stw r0,0(r3)
stw r6,4(r3)
addi r3,r3,8
12: bf cr7*4+1,13f
lwz r0,0(r4)
addi r4,r4,4
stw r0,0(r3)
addi r3,r3,4
13: bf cr7*4+2,14f
lhz r0,0(r4)
addi r4,r4,2
sth r0,0(r3)
addi r3,r3,2
14: bf cr7*4+3,15f
lbz r0,0(r4)
stb r0,0(r3)
powerpc: Fix unsafe accesses to parameter area in ELFv2 Some of the assembler files in lib/ make use of the fact that in the ELFv1 ABI, the caller guarantees to provide stack space to save the parameter registers r3 ... r10. This guarantee is no longer present in ELFv2 for functions that have no variable argument list and no more than 8 arguments. Change the affected routines to temporarily store registers in the red zone and/or the top of their own stack frame (in the space provided to save r31 .. r29, which is actually not used in these routines). In opal_query_takeover, simply always allocate a stack frame; the routine is not performance critical. Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com> Signed-off-by: Anton Blanchard <anton@samba.org>
2014-02-14 19:21:03 +01:00
15: ld r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
blr
.Lunwind_stack_nonvmx_copy:
addi r1,r1,STACKFRAMESIZE
b .Lnonvmx_copy
.Lvmx_copy:
selftests/powerpc/64: Test all paths through copy routines The hand-coded assembler 64-bit copy routines include feature sections that select one code path or another depending on which CPU we are executing on. The self-tests for these copy routines end up testing just one path. This adds a mechanism for selecting any desired code path at compile time, and makes 2 or 3 versions of each test, each using a different code path, so as to cover all the possible paths. Signed-off-by: Paul Mackerras <paulus@ozlabs.org> [mpe: Add -mcpu=power4 to CFLAGS for older compilers] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2018-08-03 20:13:04 +10:00
#ifdef CONFIG_ALTIVEC
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
mflr r0
powerpc: Fix unsafe accesses to parameter area in ELFv2 Some of the assembler files in lib/ make use of the fact that in the ELFv1 ABI, the caller guarantees to provide stack space to save the parameter registers r3 ... r10. This guarantee is no longer present in ELFv2 for functions that have no variable argument list and no more than 8 arguments. Change the affected routines to temporarily store registers in the red zone and/or the top of their own stack frame (in the space provided to save r31 .. r29, which is actually not used in these routines). In opal_query_takeover, simply always allocate a stack frame; the routine is not performance critical. Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com> Signed-off-by: Anton Blanchard <anton@samba.org>
2014-02-14 19:21:03 +01:00
std r4,-STACKFRAMESIZE+STK_REG(R30)(r1)
std r5,-STACKFRAMESIZE+STK_REG(R29)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
std r0,16(r1)
stdu r1,-STACKFRAMESIZE(r1)
powerpc/64: enhance memcmp() with VMX instruction for long bytes comparision This patch add VMX primitives to do memcmp() in case the compare size is equal or greater than 4K bytes. KSM feature can benefit from this. Test result with following test program(replace the "^>" with ""): ------ ># cat tools/testing/selftests/powerpc/stringloops/memcmp.c >#include <malloc.h> >#include <stdlib.h> >#include <string.h> >#include <time.h> >#include "utils.h" >#define SIZE (1024 * 1024 * 900) >#define ITERATIONS 40 int test_memcmp(const void *s1, const void *s2, size_t n); static int testcase(void) { char *s1; char *s2; unsigned long i; s1 = memalign(128, SIZE); if (!s1) { perror("memalign"); exit(1); } s2 = memalign(128, SIZE); if (!s2) { perror("memalign"); exit(1); } for (i = 0; i < SIZE; i++) { s1[i] = i & 0xff; s2[i] = i & 0xff; } for (i = 0; i < ITERATIONS; i++) { int ret = test_memcmp(s1, s2, SIZE); if (ret) { printf("return %d at[%ld]! should have returned zero\n", ret, i); abort(); } } return 0; } int main(void) { return test_harness(testcase, "memcmp"); } ------ Without this patch (but with the first patch "powerpc/64: Align bytes before fall back to .Lshort in powerpc64 memcmp()." in the series): 4.726728762 seconds time elapsed ( +- 3.54%) With VMX patch: 4.234335473 seconds time elapsed ( +- 2.63%) There is ~+10% improvement. Testing with unaligned and different offset version (make s1 and s2 shift random offset within 16 bytes) can archieve higher improvement than 10%.. Signed-off-by: Simon Guo <wei.guo.simon@gmail.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2018-06-07 09:57:53 +08:00
bl enter_vmx_ops
powerpc: Fix VMX in interrupt check in POWER7 copy loops The enhanced prefetch hint patches corrupt the condition register that was used to check if we are in interrupt. Fix this by using cr1. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-08-07 17:51:41 +00:00
cmpwi cr1,r3,0
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
ld r0,STACKFRAMESIZE+16(r1)
powerpc: Fix unsafe accesses to parameter area in ELFv2 Some of the assembler files in lib/ make use of the fact that in the ELFv1 ABI, the caller guarantees to provide stack space to save the parameter registers r3 ... r10. This guarantee is no longer present in ELFv2 for functions that have no variable argument list and no more than 8 arguments. Change the affected routines to temporarily store registers in the red zone and/or the top of their own stack frame (in the space provided to save r31 .. r29, which is actually not used in these routines). In opal_query_takeover, simply always allocate a stack frame; the routine is not performance critical. Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com> Signed-off-by: Anton Blanchard <anton@samba.org>
2014-02-14 19:21:03 +01:00
ld r3,STK_REG(R31)(r1)
ld r4,STK_REG(R30)(r1)
ld r5,STK_REG(R29)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
mtlr r0
/*
* We prefetch both the source and destination using enhanced touch
* instructions. We use a stream ID of 0 for the load side and
* 1 for the store side.
*/
clrrdi r6,r4,7
clrrdi r9,r3,7
ori r9,r9,1 /* stream=1 */
srdi r7,r5,7 /* length in cachelines, capped at 0x3FF */
powerpc: Fix VMX fix for memcpy case In 2fae7cdb60240e2e2d9b378afbf6d9fcce8a3890 ("powerpc: Fix VMX in interrupt check in POWER7 copy loops"), Anton inadvertently introduced a regression for memcpy on POWER7 machines. copyuser and memcpy diverge slightly in their use of cr1 (copyuser doesn't use it, but memcpy does) and you end up clobbering that register with your fix. That results in (taken from an FC18 kernel): [ 18.824604] Unrecoverable VMX/Altivec Unavailable Exception f20 at c000000000052f40 [ 18.824618] Oops: Unrecoverable VMX/Altivec Unavailable Exception, sig: 6 [#1] [ 18.824623] SMP NR_CPUS=1024 NUMA pSeries [ 18.824633] Modules linked in: tg3(+) be2net(+) cxgb4(+) ipr(+) sunrpc xts lrw gf128mul dm_crypt dm_round_robin dm_multipath linear raid10 raid456 async_raid6_recov async_memcpy async_pq raid6_pq async_xor xor async_tx raid1 raid0 scsi_dh_rdac scsi_dh_hp_sw scsi_dh_emc scsi_dh_alua squashfs cramfs [ 18.824705] NIP: c000000000052f40 LR: c00000000020b874 CTR: 0000000000000512 [ 18.824709] REGS: c000001f1fef7790 TRAP: 0f20 Not tainted (3.6.0-0.rc6.git0.2.fc18.ppc64) [ 18.824713] MSR: 8000000000009032 <SF,EE,ME,IR,DR,RI> CR: 4802802e XER: 20000010 [ 18.824726] SOFTE: 0 [ 18.824728] CFAR: 0000000000000f20 [ 18.824731] TASK = c000000fa7128400[0] 'swapper/24' THREAD: c000000fa7480000 CPU: 24 GPR00: 00000000ffffffc0 c000001f1fef7a10 c00000000164edc0 c000000f9b9a8120 GPR04: c000000f9b9a8124 0000000000001438 0000000000000060 03ffffff064657ee GPR08: 0000000080000000 0000000000000010 0000000000000020 0000000000000030 GPR12: 0000000028028022 c00000000ff25400 0000000000000001 0000000000000000 GPR16: 0000000000000000 7fffffffffffffff c0000000016b2180 c00000000156a500 GPR20: c000000f968c7a90 c0000000131c31d8 c000001f1fef4000 c000000001561d00 GPR24: 000000000000000a 0000000000000000 0000000000000001 0000000000000012 GPR28: c000000fa5c04f80 00000000000008bc c0000000015c0a28 000000000000022e [ 18.824792] NIP [c000000000052f40] .memcpy_power7+0x5a0/0x7c4 [ 18.824797] LR [c00000000020b874] .pcpu_free_area+0x174/0x2d0 [ 18.824800] Call Trace: [ 18.824803] [c000001f1fef7a10] [c000000000052c14] .memcpy_power7+0x274/0x7c4 (unreliable) [ 18.824809] [c000001f1fef7b10] [c00000000020b874] .pcpu_free_area+0x174/0x2d0 [ 18.824813] [c000001f1fef7bb0] [c00000000020ba88] .free_percpu+0xb8/0x1b0 [ 18.824819] [c000001f1fef7c50] [c00000000043d144] .throtl_pd_exit+0x94/0xd0 [ 18.824824] [c000001f1fef7cf0] [c00000000043acf8] .blkg_free+0x88/0xe0 [ 18.824829] [c000001f1fef7d90] [c00000000018c048] .rcu_process_callbacks+0x2e8/0x8a0 [ 18.824835] [c000001f1fef7e90] [c0000000000a8ce8] .__do_softirq+0x158/0x4d0 [ 18.824840] [c000001f1fef7f90] [c000000000025ecc] .call_do_softirq+0x14/0x24 [ 18.824845] [c000000fa7483650] [c000000000010e80] .do_softirq+0x160/0x1a0 [ 18.824850] [c000000fa74836f0] [c0000000000a94a4] .irq_exit+0xf4/0x120 [ 18.824854] [c000000fa7483780] [c000000000020c44] .timer_interrupt+0x154/0x4d0 [ 18.824859] [c000000fa7483830] [c000000000003be0] decrementer_common+0x160/0x180 [ 18.824866] --- Exception: 901 at .plpar_hcall_norets+0x84/0xd4 [ 18.824866] LR = .check_and_cede_processor+0x48/0x80 [ 18.824871] [c000000fa7483b20] [c00000000007f018] .check_and_cede_processor+0x18/0x80 (unreliable) [ 18.824877] [c000000fa7483b90] [c00000000007f104] .dedicated_cede_loop+0x84/0x150 [ 18.824883] [c000000fa7483c50] [c0000000006bc030] .cpuidle_enter+0x30/0x50 [ 18.824887] [c000000fa7483cc0] [c0000000006bc9f4] .cpuidle_idle_call+0x104/0x720 [ 18.824892] [c000000fa7483d80] [c000000000070af8] .pSeries_idle+0x18/0x40 [ 18.824897] [c000000fa7483df0] [c000000000019084] .cpu_idle+0x1a4/0x380 [ 18.824902] [c000000fa7483ec0] [c0000000008a4c18] .start_secondary+0x520/0x528 [ 18.824907] [c000000fa7483f90] [c0000000000093f0] .start_secondary_prolog+0x10/0x14 [ 18.824911] Instruction dump: [ 18.824914] 38840008 90030000 90e30004 38630008 7ca62850 7cc300d0 78c7e102 7cf01120 [ 18.824923] 78c60660 39200010 39400020 39600030 <7e00200c> 7c0020ce 38840010 409f001c [ 18.824935] ---[ end trace 0bb95124affaaa45 ]--- [ 18.825046] Unrecoverable VMX/Altivec Unavailable Exception f20 at c000000000052d08 I believe the right fix is to make memcpy match usercopy and not use cr1. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> CC: <stable@kernel.org> [v3.6]
2012-10-01 14:59:13 +00:00
cmpldi r7,0x3FF
ble 1f
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
li r7,0x3FF
1: lis r0,0x0E00 /* depth=7 */
sldi r7,r7,7
or r7,r7,r0
ori r10,r7,1 /* stream=1 */
lis r8,0x8000 /* GO=1 */
clrldi r8,r8,32
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
dcbt 0,r6,0b01000
dcbt 0,r7,0b01010
dcbtst 0,r9,0b01000
dcbtst 0,r10,0b01010
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
eieio
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
dcbt 0,r8,0b01010 /* GO */
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
powerpc: Fix VMX in interrupt check in POWER7 copy loops The enhanced prefetch hint patches corrupt the condition register that was used to check if we are in interrupt. Fix this by using cr1. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-08-07 17:51:41 +00:00
beq cr1,.Lunwind_stack_nonvmx_copy
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
/*
* If source and destination are not relatively aligned we use a
* slower permute loop.
*/
xor r6,r4,r3
rldicl. r6,r6,0,(64-4)
bne .Lvmx_unaligned_copy
/* Get the destination 16B aligned */
neg r6,r3
mtocrf 0x01,r6
clrldi r6,r6,(64-4)
bf cr7*4+3,1f
lbz r0,0(r4)
addi r4,r4,1
stb r0,0(r3)
addi r3,r3,1
1: bf cr7*4+2,2f
lhz r0,0(r4)
addi r4,r4,2
sth r0,0(r3)
addi r3,r3,2
2: bf cr7*4+1,3f
lwz r0,0(r4)
addi r4,r4,4
stw r0,0(r3)
addi r3,r3,4
3: bf cr7*4+0,4f
ld r0,0(r4)
addi r4,r4,8
std r0,0(r3)
addi r3,r3,8
4: sub r5,r5,r6
/* Get the desination 128B aligned */
neg r6,r3
srdi r7,r6,4
mtocrf 0x01,r7
clrldi r6,r6,(64-7)
li r9,16
li r10,32
li r11,48
bf cr7*4+3,5f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v1,0,r4
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,16
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v1,0,r3
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,16
5: bf cr7*4+2,6f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v1,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
lvx v0,r4,r9
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,32
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v1,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v0,r3,r9
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,32
6: bf cr7*4+1,7f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v3,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
lvx v2,r4,r9
lvx v1,r4,r10
lvx v0,r4,r11
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,64
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v3,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v2,r3,r9
stvx v1,r3,r10
stvx v0,r3,r11
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,64
7: sub r5,r5,r6
srdi r6,r5,7
powerpc: Fix usage of register macros getting ready for %r0 change Anything that uses a constructed instruction (ie. from ppc-opcode.h), need to use the new R0 macro, as %r0 is not going to work. Also convert usages of macros where we are just determining an offset (usually for a load/store), like: std r14,STK_REG(r14)(r1) Can't use STK_REG(r14) as %r14 doesn't work in the STK_REG macro since it's just calculating an offset. Signed-off-by: Michael Neuling <mikey@neuling.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-06-25 13:33:10 +00:00
std r14,STK_REG(R14)(r1)
std r15,STK_REG(R15)(r1)
std r16,STK_REG(R16)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
li r12,64
li r14,80
li r15,96
li r16,112
mtctr r6
/*
* Now do cacheline sized loads and stores. By this stage the
* cacheline stores are also cacheline aligned.
*/
.align 5
8:
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v7,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
lvx v6,r4,r9
lvx v5,r4,r10
lvx v4,r4,r11
lvx v3,r4,r12
lvx v2,r4,r14
lvx v1,r4,r15
lvx v0,r4,r16
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,128
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v7,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v6,r3,r9
stvx v5,r3,r10
stvx v4,r3,r11
stvx v3,r3,r12
stvx v2,r3,r14
stvx v1,r3,r15
stvx v0,r3,r16
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,128
bdnz 8b
powerpc: Fix usage of register macros getting ready for %r0 change Anything that uses a constructed instruction (ie. from ppc-opcode.h), need to use the new R0 macro, as %r0 is not going to work. Also convert usages of macros where we are just determining an offset (usually for a load/store), like: std r14,STK_REG(r14)(r1) Can't use STK_REG(r14) as %r14 doesn't work in the STK_REG macro since it's just calculating an offset. Signed-off-by: Michael Neuling <mikey@neuling.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-06-25 13:33:10 +00:00
ld r14,STK_REG(R14)(r1)
ld r15,STK_REG(R15)(r1)
ld r16,STK_REG(R16)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
/* Up to 127B to go */
clrldi r5,r5,(64-7)
srdi r6,r5,4
mtocrf 0x01,r6
bf cr7*4+1,9f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v3,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
lvx v2,r4,r9
lvx v1,r4,r10
lvx v0,r4,r11
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,64
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v3,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v2,r3,r9
stvx v1,r3,r10
stvx v0,r3,r11
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,64
9: bf cr7*4+2,10f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v1,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
lvx v0,r4,r9
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,32
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v1,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v0,r3,r9
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,32
10: bf cr7*4+3,11f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v1,0,r4
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,16
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v1,0,r3
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,16
/* Up to 15B to go */
11: clrldi r5,r5,(64-4)
mtocrf 0x01,r5
bf cr7*4+0,12f
ld r0,0(r4)
addi r4,r4,8
std r0,0(r3)
addi r3,r3,8
12: bf cr7*4+1,13f
lwz r0,0(r4)
addi r4,r4,4
stw r0,0(r3)
addi r3,r3,4
13: bf cr7*4+2,14f
lhz r0,0(r4)
addi r4,r4,2
sth r0,0(r3)
addi r3,r3,2
14: bf cr7*4+3,15f
lbz r0,0(r4)
stb r0,0(r3)
15: addi r1,r1,STACKFRAMESIZE
powerpc: Fix unsafe accesses to parameter area in ELFv2 Some of the assembler files in lib/ make use of the fact that in the ELFv1 ABI, the caller guarantees to provide stack space to save the parameter registers r3 ... r10. This guarantee is no longer present in ELFv2 for functions that have no variable argument list and no more than 8 arguments. Change the affected routines to temporarily store registers in the red zone and/or the top of their own stack frame (in the space provided to save r31 .. r29, which is actually not used in these routines). In opal_query_takeover, simply always allocate a stack frame; the routine is not performance critical. Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com> Signed-off-by: Anton Blanchard <anton@samba.org>
2014-02-14 19:21:03 +01:00
ld r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
powerpc/64: enhance memcmp() with VMX instruction for long bytes comparision This patch add VMX primitives to do memcmp() in case the compare size is equal or greater than 4K bytes. KSM feature can benefit from this. Test result with following test program(replace the "^>" with ""): ------ ># cat tools/testing/selftests/powerpc/stringloops/memcmp.c >#include <malloc.h> >#include <stdlib.h> >#include <string.h> >#include <time.h> >#include "utils.h" >#define SIZE (1024 * 1024 * 900) >#define ITERATIONS 40 int test_memcmp(const void *s1, const void *s2, size_t n); static int testcase(void) { char *s1; char *s2; unsigned long i; s1 = memalign(128, SIZE); if (!s1) { perror("memalign"); exit(1); } s2 = memalign(128, SIZE); if (!s2) { perror("memalign"); exit(1); } for (i = 0; i < SIZE; i++) { s1[i] = i & 0xff; s2[i] = i & 0xff; } for (i = 0; i < ITERATIONS; i++) { int ret = test_memcmp(s1, s2, SIZE); if (ret) { printf("return %d at[%ld]! should have returned zero\n", ret, i); abort(); } } return 0; } int main(void) { return test_harness(testcase, "memcmp"); } ------ Without this patch (but with the first patch "powerpc/64: Align bytes before fall back to .Lshort in powerpc64 memcmp()." in the series): 4.726728762 seconds time elapsed ( +- 3.54%) With VMX patch: 4.234335473 seconds time elapsed ( +- 2.63%) There is ~+10% improvement. Testing with unaligned and different offset version (make s1 and s2 shift random offset within 16 bytes) can archieve higher improvement than 10%.. Signed-off-by: Simon Guo <wei.guo.simon@gmail.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2018-06-07 09:57:53 +08:00
b exit_vmx_ops /* tail call optimise */
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
.Lvmx_unaligned_copy:
/* Get the destination 16B aligned */
neg r6,r3
mtocrf 0x01,r6
clrldi r6,r6,(64-4)
bf cr7*4+3,1f
lbz r0,0(r4)
addi r4,r4,1
stb r0,0(r3)
addi r3,r3,1
1: bf cr7*4+2,2f
lhz r0,0(r4)
addi r4,r4,2
sth r0,0(r3)
addi r3,r3,2
2: bf cr7*4+1,3f
lwz r0,0(r4)
addi r4,r4,4
stw r0,0(r3)
addi r3,r3,4
3: bf cr7*4+0,4f
lwz r0,0(r4) /* Less chance of a reject with word ops */
lwz r7,4(r4)
addi r4,r4,8
stw r0,0(r3)
stw r7,4(r3)
addi r3,r3,8
4: sub r5,r5,r6
/* Get the desination 128B aligned */
neg r6,r3
srdi r7,r6,4
mtocrf 0x01,r7
clrldi r6,r6,(64-7)
li r9,16
li r10,32
li r11,48
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
LVS(v16,0,r4) /* Setup permute control vector */
lvx v0,0,r4
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,16
bf cr7*4+3,5f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v1,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
VPERM(v8,v0,v1,v16)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,16
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v8,0,r3
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,16
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
vor v0,v1,v1
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
5: bf cr7*4+2,6f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v1,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
VPERM(v8,v0,v1,v16)
lvx v0,r4,r9
VPERM(v9,v1,v0,v16)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,32
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v8,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v9,r3,r9
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,32
6: bf cr7*4+1,7f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v3,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
VPERM(v8,v0,v3,v16)
lvx v2,r4,r9
VPERM(v9,v3,v2,v16)
lvx v1,r4,r10
VPERM(v10,v2,v1,v16)
lvx v0,r4,r11
VPERM(v11,v1,v0,v16)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,64
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v8,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v9,r3,r9
stvx v10,r3,r10
stvx v11,r3,r11
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,64
7: sub r5,r5,r6
srdi r6,r5,7
powerpc: Fix usage of register macros getting ready for %r0 change Anything that uses a constructed instruction (ie. from ppc-opcode.h), need to use the new R0 macro, as %r0 is not going to work. Also convert usages of macros where we are just determining an offset (usually for a load/store), like: std r14,STK_REG(r14)(r1) Can't use STK_REG(r14) as %r14 doesn't work in the STK_REG macro since it's just calculating an offset. Signed-off-by: Michael Neuling <mikey@neuling.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-06-25 13:33:10 +00:00
std r14,STK_REG(R14)(r1)
std r15,STK_REG(R15)(r1)
std r16,STK_REG(R16)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
li r12,64
li r14,80
li r15,96
li r16,112
mtctr r6
/*
* Now do cacheline sized loads and stores. By this stage the
* cacheline stores are also cacheline aligned.
*/
.align 5
8:
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v7,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
VPERM(v8,v0,v7,v16)
lvx v6,r4,r9
VPERM(v9,v7,v6,v16)
lvx v5,r4,r10
VPERM(v10,v6,v5,v16)
lvx v4,r4,r11
VPERM(v11,v5,v4,v16)
lvx v3,r4,r12
VPERM(v12,v4,v3,v16)
lvx v2,r4,r14
VPERM(v13,v3,v2,v16)
lvx v1,r4,r15
VPERM(v14,v2,v1,v16)
lvx v0,r4,r16
VPERM(v15,v1,v0,v16)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,128
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v8,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v9,r3,r9
stvx v10,r3,r10
stvx v11,r3,r11
stvx v12,r3,r12
stvx v13,r3,r14
stvx v14,r3,r15
stvx v15,r3,r16
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,128
bdnz 8b
powerpc: Fix usage of register macros getting ready for %r0 change Anything that uses a constructed instruction (ie. from ppc-opcode.h), need to use the new R0 macro, as %r0 is not going to work. Also convert usages of macros where we are just determining an offset (usually for a load/store), like: std r14,STK_REG(r14)(r1) Can't use STK_REG(r14) as %r14 doesn't work in the STK_REG macro since it's just calculating an offset. Signed-off-by: Michael Neuling <mikey@neuling.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-06-25 13:33:10 +00:00
ld r14,STK_REG(R14)(r1)
ld r15,STK_REG(R15)(r1)
ld r16,STK_REG(R16)(r1)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
/* Up to 127B to go */
clrldi r5,r5,(64-7)
srdi r6,r5,4
mtocrf 0x01,r6
bf cr7*4+1,9f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v3,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
VPERM(v8,v0,v3,v16)
lvx v2,r4,r9
VPERM(v9,v3,v2,v16)
lvx v1,r4,r10
VPERM(v10,v2,v1,v16)
lvx v0,r4,r11
VPERM(v11,v1,v0,v16)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,64
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v8,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v9,r3,r9
stvx v10,r3,r10
stvx v11,r3,r11
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,64
9: bf cr7*4+2,10f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v1,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
VPERM(v8,v0,v1,v16)
lvx v0,r4,r9
VPERM(v9,v1,v0,v16)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,32
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v8,0,r3
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
stvx v9,r3,r9
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,32
10: bf cr7*4+3,11f
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
lvx v1,0,r4
powerpc: Change vrX register defines to vX to match gcc and glibc As our various loops (copy, string, crypto etc) get more complicated, we want to share implementations between userspace (eg glibc) and the kernel. We also want to write userspace test harnesses to put in tools/testing/selftest. One gratuitous difference between userspace and the kernel is the VMX register definitions - the kernel uses vrX whereas both gcc and glibc use vX. Change the kernel to match userspace. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-10 09:51:22 +11:00
VPERM(v8,v0,v1,v16)
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r4,r4,16
powerpc: Fix invalid use of register expressions binutils >= 2.26 now warns about misuse of register expressions in assembler operands that are actually literals, for example: arch/powerpc/kernel/entry_64.S:535: Warning: invalid register expression In practice these are almost all uses of r0 that should just be a literal 0. Signed-off-by: Andreas Schwab <schwab@linux-m68k.org> [mpe: Mention r0 is almost always the culprit, fold in purgatory change] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2017-08-05 19:55:11 +02:00
stvx v8,0,r3
powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch Implement a POWER7 optimised memcpy using VMX and enhanced prefetch instructions. This is a copy of the POWER7 optimised copy_to_user/copy_from_user loop. Detailed implementation and performance details can be found in commit a66086b8197d (powerpc: POWER7 optimised copy_to_user/copy_from_user using VMX). I noticed memcpy issues when profiling a RAID6 workload: .memcpy .async_memcpy .async_copy_data .__raid_run_ops .handle_stripe .raid5d .md_thread I created a simplified testcase by building a RAID6 array with 4 1GB ramdisks (booting with brd.rd_size=1048576): # mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3] I then timed how long it took to write to the entire array: # dd if=/dev/zero of=/dev/md0 bs=1M Before: 892 MB/s After: 999 MB/s A 12% improvement. Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-05-30 20:22:09 +00:00
addi r3,r3,16
/* Up to 15B to go */
11: clrldi r5,r5,(64-4)
addi r4,r4,-16 /* Unwind the +16 load offset */
mtocrf 0x01,r5
bf cr7*4+0,12f
lwz r0,0(r4) /* Less chance of a reject with word ops */
lwz r6,4(r4)
addi r4,r4,8
stw r0,0(r3)
stw r6,4(r3)
addi r3,r3,8
12: bf cr7*4+1,13f
lwz r0,0(r4)
addi r4,r4,4
stw r0,0(r3)
addi r3,r3,4
13: bf cr7*4+2,14f
lhz r0,0(r4)
addi r4,r4,2
sth r0,0(r3)
addi r3,r3,2
14: bf cr7*4+3,15f
lbz r0,0(r4)
stb r0,0(r3)
15: addi r1,r1,STACKFRAMESIZE
powerpc: Fix unsafe accesses to parameter area in ELFv2 Some of the assembler files in lib/ make use of the fact that in the ELFv1 ABI, the caller guarantees to provide stack space to save the parameter registers r3 ... r10. This guarantee is no longer present in ELFv2 for functions that have no variable argument list and no more than 8 arguments. Change the affected routines to temporarily store registers in the red zone and/or the top of their own stack frame (in the space provided to save r31 .. r29, which is actually not used in these routines). In opal_query_takeover, simply always allocate a stack frame; the routine is not performance critical. Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com> Signed-off-by: Anton Blanchard <anton@samba.org>
2014-02-14 19:21:03 +01:00
ld r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
powerpc/64: enhance memcmp() with VMX instruction for long bytes comparision This patch add VMX primitives to do memcmp() in case the compare size is equal or greater than 4K bytes. KSM feature can benefit from this. Test result with following test program(replace the "^>" with ""): ------ ># cat tools/testing/selftests/powerpc/stringloops/memcmp.c >#include <malloc.h> >#include <stdlib.h> >#include <string.h> >#include <time.h> >#include "utils.h" >#define SIZE (1024 * 1024 * 900) >#define ITERATIONS 40 int test_memcmp(const void *s1, const void *s2, size_t n); static int testcase(void) { char *s1; char *s2; unsigned long i; s1 = memalign(128, SIZE); if (!s1) { perror("memalign"); exit(1); } s2 = memalign(128, SIZE); if (!s2) { perror("memalign"); exit(1); } for (i = 0; i < SIZE; i++) { s1[i] = i & 0xff; s2[i] = i & 0xff; } for (i = 0; i < ITERATIONS; i++) { int ret = test_memcmp(s1, s2, SIZE); if (ret) { printf("return %d at[%ld]! should have returned zero\n", ret, i); abort(); } } return 0; } int main(void) { return test_harness(testcase, "memcmp"); } ------ Without this patch (but with the first patch "powerpc/64: Align bytes before fall back to .Lshort in powerpc64 memcmp()." in the series): 4.726728762 seconds time elapsed ( +- 3.54%) With VMX patch: 4.234335473 seconds time elapsed ( +- 2.63%) There is ~+10% improvement. Testing with unaligned and different offset version (make s1 and s2 shift random offset within 16 bytes) can archieve higher improvement than 10%.. Signed-off-by: Simon Guo <wei.guo.simon@gmail.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2018-06-07 09:57:53 +08:00
b exit_vmx_ops /* tail call optimise */
powerpc: Fix comment typos 'CONFiG_ALTIVEC' Signed-off-by: Paul Bolle <pebolle@tiscali.nl> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
2014-09-26 19:45:34 +02:00
#endif /* CONFIG_ALTIVEC */
Reference in New Issue Copy Permalink