Commit a16036e9 authored by David S. Miller's avatar David S. Miller
Browse files

Merge branch 'sparc64-optimized-fls' 



Vijay Kumar says:

====================
sparc64: Optimize fls and __fls

SPARC provides lzcnt instruction (with VIS3) which can be used to
optimize fls, __fls and fls64 functions. For the systems that supports
lzcnt instruction, we now do boot time patching to use sparc
optimized fls, __fls and fls64 functions.

v3->v4:
 -  Fixed a typo.
v2->v3:
 -  Using ENTRY(), ENDPROC() for assembler functions.
 -  Removed BITS_PER_LONG from __fls.
 -  Using generic fls64().
 -  Replaced lzcnt instruction with .word directive.
v1->v2:
 - Fixed delay slot issue.
====================

Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parents 9a08862a 46ad8d2d

arch/sparc/include/asm/bitops_64.h

+3 −2
Original line number Diff line number Diff line
@@ -22,10 +22,11 @@ void set_bit(unsigned long nr, volatile unsigned long *addr); 
void clear_bit(unsigned long nr, volatile unsigned long *addr);

void change_bit(unsigned long nr, volatile unsigned long *addr);



int fls(unsigned int word);

int __fls(unsigned long word);



#include <asm-generic/bitops/non-atomic.h>



#include <asm-generic/bitops/fls.h>

#include <asm-generic/bitops/__fls.h>

#include <asm-generic/bitops/fls64.h>



#ifdef __KERNEL__

arch/sparc/kernel/head_64.S

+2 −0
Original line number Diff line number Diff line
@@ -640,6 +640,8 @@ niagara4_patch: 
	 nop

	call	niagara4_patch_pageops

	 nop

	call	niagara4_patch_fls

	 nop



	ba,a,pt	%xcc, 80f

	 nop

arch/sparc/lib/Makefile

+3 −0
Original line number Diff line number Diff line
@@ -16,6 +16,9 @@ lib-$(CONFIG_SPARC64) += atomic_64.o 
lib-$(CONFIG_SPARC32) += lshrdi3.o ashldi3.o

lib-$(CONFIG_SPARC32) += muldi3.o bitext.o cmpdi2.o

lib-$(CONFIG_SPARC64) += multi3.o

lib-$(CONFIG_SPARC64) += fls.o

lib-$(CONFIG_SPARC64) += fls64.o

obj-$(CONFIG_SPARC64) += NG4fls.o



lib-$(CONFIG_SPARC64) += copy_page.o clear_page.o bzero.o

lib-$(CONFIG_SPARC64) += csum_copy.o csum_copy_from_user.o csum_copy_to_user.o

arch/sparc/lib/NG4fls.S

0 → 100644
+30 −0
Original line number Diff line number Diff line 
/* NG4fls.S: SPARC optimized fls and __fls for T4 and above.

 *

 * Copyright (c) 2017, Oracle and/or its affiliates. All rights reserved.

 */



#include <linux/linkage.h>



#define LZCNT_O0_G2	\

	.word	0x85b002e8



	.text

	.register	%g2, #scratch

	.register	%g3, #scratch



ENTRY(NG4fls)

	LZCNT_O0_G2	!lzcnt	%o0, %g2

	mov	64, %g3

	retl

	 sub	%g3, %g2, %o0

ENDPROC(NG4fls)



ENTRY(__NG4fls)

	brz,pn	%o0, 1f

	LZCNT_O0_G2	!lzcnt	%o0, %g2

	mov	63, %g3

	sub	%g3, %g2, %o0

1:

	retl

	 nop

ENDPROC(__NG4fls)

arch/sparc/lib/NG4patch.S

+9 −0
Original line number Diff line number Diff line
@@ -3,6 +3,8 @@ 
 * Copyright (C) 2012 David S. Miller <davem@davemloft.net>

 */



#include <linux/linkage.h>



#define BRANCH_ALWAYS	0x10680000

#define NOP		0x01000000

#define NG_DO_PATCH(OLD, NEW)	\
@@ -52,3 +54,10 @@ niagara4_patch_pageops: 
	retl

	 nop

	.size	niagara4_patch_pageops,.-niagara4_patch_pageops



ENTRY(niagara4_patch_fls)

	NG_DO_PATCH(fls, NG4fls)

	NG_DO_PATCH(__fls, __NG4fls)

	retl

	 nop

ENDPROC(niagara4_patch_fls)

CRA Git | Maintained and supported by SUSTech CRA and CCSE