linux/arch/x86/lib/atomic64_386_32.S

197 lines
2.6 KiB
ArmAsm
Raw Normal View History

x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
/*
* atomic64_t for 386/486
*
* Copyright © 2010 Luca Barbieri
*
* 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.
*/
#include <linux/linkage.h>
#include <asm/alternative-asm.h>
#include <asm/dwarf2.h>
/* if you want SMP support, implement these with real spinlocks */
.macro LOCK reg
pushfl
CFI_ADJUST_CFA_OFFSET 4
cli
.endm
.macro UNLOCK reg
popfl
CFI_ADJUST_CFA_OFFSET -4
.endm
#define BEGIN(op) \
.macro endp; \
CFI_ENDPROC; \
ENDPROC(atomic64_##op##_386); \
.purgem endp; \
.endm; \
ENTRY(atomic64_##op##_386); \
CFI_STARTPROC; \
LOCK v;
#define ENDP endp
#define RET \
UNLOCK v; \
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
ret
#define RET_ENDP \
RET; \
ENDP
#define v %ecx
BEGIN(read)
movl (v), %eax
movl 4(v), %edx
RET_ENDP
#undef v
#define v %esi
BEGIN(set)
movl %ebx, (v)
movl %ecx, 4(v)
RET_ENDP
#undef v
#define v %esi
BEGIN(xchg)
movl (v), %eax
movl 4(v), %edx
movl %ebx, (v)
movl %ecx, 4(v)
RET_ENDP
#undef v
#define v %ecx
BEGIN(add)
addl %eax, (v)
adcl %edx, 4(v)
RET_ENDP
#undef v
#define v %ecx
BEGIN(add_return)
addl (v), %eax
adcl 4(v), %edx
movl %eax, (v)
movl %edx, 4(v)
RET_ENDP
#undef v
#define v %ecx
BEGIN(sub)
subl %eax, (v)
sbbl %edx, 4(v)
RET_ENDP
#undef v
#define v %ecx
BEGIN(sub_return)
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
negl %edx
negl %eax
sbbl $0, %edx
addl (v), %eax
adcl 4(v), %edx
movl %eax, (v)
movl %edx, 4(v)
RET_ENDP
#undef v
#define v %esi
BEGIN(inc)
addl $1, (v)
adcl $0, 4(v)
RET_ENDP
#undef v
#define v %esi
BEGIN(inc_return)
movl (v), %eax
movl 4(v), %edx
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
addl $1, %eax
adcl $0, %edx
movl %eax, (v)
movl %edx, 4(v)
RET_ENDP
#undef v
#define v %esi
BEGIN(dec)
subl $1, (v)
sbbl $0, 4(v)
RET_ENDP
#undef v
#define v %esi
BEGIN(dec_return)
movl (v), %eax
movl 4(v), %edx
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
subl $1, %eax
sbbl $0, %edx
movl %eax, (v)
movl %edx, 4(v)
RET_ENDP
#undef v
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
#define v %ecx
BEGIN(add_unless)
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
addl %eax, %esi
adcl %edx, %edi
addl (v), %eax
adcl 4(v), %edx
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
cmpl %eax, %esi
je 3f
1:
movl %eax, (v)
movl %edx, 4(v)
movl $1, %eax
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
2:
RET
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
3:
cmpl %edx, %edi
jne 1b
xorl %eax, %eax
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
jmp 2b
ENDP
#undef v
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
#define v %esi
BEGIN(inc_not_zero)
movl (v), %eax
movl 4(v), %edx
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
testl %eax, %eax
je 3f
1:
addl $1, %eax
adcl $0, %edx
movl %eax, (v)
movl %edx, 4(v)
movl $1, %eax
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
2:
RET
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
3:
testl %edx, %edx
jne 1b
jmp 2b
ENDP
#undef v
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
#define v %esi
BEGIN(dec_if_positive)
movl (v), %eax
movl 4(v), %edx
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
subl $1, %eax
sbbl $0, %edx
js 1f
movl %eax, (v)
movl %edx, 4(v)
x86-32: Rewrite 32-bit atomic64 functions in assembly This patch replaces atomic64_32.c with two assembly implementations, one for 386/486 machines using pushf/cli/popf and one for 586+ machines using cmpxchg8b. The cmpxchg8b implementation provides the following advantages over the current one: 1. Implements atomic64_add_unless, atomic64_dec_if_positive and atomic64_inc_not_zero 2. Uses the ZF flag changed by cmpxchg8b instead of doing a comparison 3. Uses custom register calling conventions that reduce or eliminate register moves to suit cmpxchg8b 4. Reads the initial value instead of using cmpxchg8b to do that. Currently we use lock xaddl and movl, which seems the fastest. 5. Does not use the lock prefix for atomic64_set 64-bit writes are already atomic, so we don't need that. We still need it for atomic64_read to avoid restoring a value changed in the meantime. 6. Allocates registers as well or better than gcc The 386 implementation provides support for 386 and 486 machines. 386/486 SMP is not supported (we dropped it), but such support can be added easily if desired. A pure assembly implementation is required due to the custom calling conventions, and desire to use %ebp in atomic64_add_return (we need 7 registers...), as well as the ability to use pushf/popf in the 386 code without an intermediate pop/push. The parameter names are changed to match the convention in atomic_64.h Changes in v3 (due to rebasing to tip/x86/asm): - Patches atomic64_32.h instead of atomic_32.h - Uses the CALL alternative mechanism from commit 1b1d9258181bae199dc940f4bd0298126b9a73d9 Changes in v2: - Merged 386 and cx8 support in the same patch - 386 support now done in assembly, C code no longer used at all - cmpxchg64 is used for atomic64_cmpxchg - stop using macros, use one-line inline functions instead - miscellanous changes and improvements Signed-off-by: Luca Barbieri <luca@luca-barbieri.com> LKML-Reference: <1267005265-27958-5-git-send-email-luca@luca-barbieri.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
2010-02-24 17:54:25 +08:00
1:
RET_ENDP
#undef v