From: "Ard Biesheuvel" <ard.biesheuvel@arm.com>
To: devel@edk2.groups.io, philmd@redhat.com
Cc: glin@suse.com, leif@nuviainc.com, lersek@redhat.com,
liming.gao@intel.com
Subject: Re: [edk2-devel] [PATCH v2] ArmPkg/CompilerIntrinsicsLib: provide atomics intrinsics
Date: Thu, 21 May 2020 18:45:39 +0200 [thread overview]
Message-ID: <c5ade5be-9c21-9797-df72-c42317445b4d@arm.com> (raw)
In-Reply-To: <6c9ec3a3-8aa1-c4d4-c7ba-1b9e28fd0866@redhat.com>
On 5/21/20 6:40 PM, Philippe Mathieu-Daudé via groups.io wrote:
> On 5/20/20 2:37 PM, Philippe Mathieu-Daudé wrote:
>> Hi Ard,
>>
>> On 5/20/20 1:44 PM, Ard Biesheuvel wrote:
>>> Gary reports the GCC 10 will emit calls to atomics intrinsics routines
>>> unless -mno-outline-atomics is specified. This means GCC-10 introduces
>>> new intrinsics, and even though it would be possible to work around this
>>> by specifying the command line option, this would require a new GCC10
>>> toolchain profile to be created, which we prefer to avoid.
>>>
>>> So instead, add the new intrinsics to our library so they are provided
>>> when necessary.
>>>
>>> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@arm.com>
>>> ---
>>> v2:
>>> - add missing .globl to export the functions from the object file
>>> - add function end markers so the size of each is visible in the ELF
>>> metadata
>>> - add some comments to describe what is going on
>>
>> Thanks, head hurts a bit less...
>>
>>>
>>> ArmPkg/Library/CompilerIntrinsicsLib/CompilerIntrinsicsLib.inf | 3 +
>>> ArmPkg/Library/CompilerIntrinsicsLib/AArch64/Atomics.S |
>>> 142 ++++++++++++++++++++
>>> 2 files changed, 145 insertions(+)
>>>
>>> diff --git
>>> a/ArmPkg/Library/CompilerIntrinsicsLib/CompilerIntrinsicsLib.inf
>>> b/ArmPkg/Library/CompilerIntrinsicsLib/CompilerIntrinsicsLib.inf
>>> index d5bad9467758..fcf48c678119 100644
>>> --- a/ArmPkg/Library/CompilerIntrinsicsLib/CompilerIntrinsicsLib.inf
>>> +++ b/ArmPkg/Library/CompilerIntrinsicsLib/CompilerIntrinsicsLib.inf
>>> @@ -79,6 +79,9 @@ [Sources.ARM]
>>> Arm/ldivmod.asm | MSFT
>>> Arm/llsr.asm | MSFT
>>> +[Sources.AARCH64]
>>> + AArch64/Atomics.S | GCC
>>> +
>>> [Packages]
>>> MdePkg/MdePkg.dec
>>> ArmPkg/ArmPkg.dec
>>> diff --git a/ArmPkg/Library/CompilerIntrinsicsLib/AArch64/Atomics.S
>>> b/ArmPkg/Library/CompilerIntrinsicsLib/AArch64/Atomics.S
>>> new file mode 100644
>>> index 000000000000..dc61d6bb8e52
>>> --- /dev/null
>>> +++ b/ArmPkg/Library/CompilerIntrinsicsLib/AArch64/Atomics.S
>>> @@ -0,0 +1,142 @@
>>> +#------------------------------------------------------------------------------
>>>
>>> +#
>>> +# Copyright (c) 2020, Arm, Limited. All rights reserved.<BR>
>>> +#
>>> +# SPDX-License-Identifier: BSD-2-Clause-Patent
>>> +#
>>> +#------------------------------------------------------------------------------
>>>
>>> +
>>> + /*
>>> + * Provide the GCC intrinsics that are required when using GCC 9 or
>>> + * later with the -moutline-atomics options (which became the
>>> default
>>> + * in GCC 10)
>>> + */
>>> + .arch armv8-a
>>> +
>>> + .macro reg_alias, pfx, sz
>>> + r0_\sz .req \pfx\()0
>>> + r1_\sz .req \pfx\()1
>>> + tmp0_\sz .req \pfx\()16
>>> + tmp1_\sz .req \pfx\()17
>>> + .endm
>>> +
>>> + /*
>>> + * Define register aliases of the right type for each size
>>> + * (xN for 8 bytes, wN for everything smaller)
>>> + */
>>> + reg_alias w, 1
>>> + reg_alias w, 2
>>> + reg_alias w, 4
>>> + reg_alias x, 8
>>> +
>>> + .macro fn_start, name:req
>>> + .section .text.\name
>>> + .globl \name
>>> + .type \name, %function
>>> +\name\():
>>> + .endm
>>> +
>>> + .macro fn_end, name:req
>>> + .size \name, . - \name
>>> + .endm
>>> +
>>> + /*
>>> + * Emit an atomic helper for \model with operands of size \sz,
>>> using
>>> + * the operation specified by \insn (which is the LSE name), and
>>> which
>>> + * can be implemented using the generic
>>> load-locked/store-conditional
>>> + * (LL/SC) sequence below, using the arithmetic operation given by
>>> + * \opc.
>>> + */
>>> + .macro emit_ld_sz, sz:req, insn:req, opc:req, model:req,
>>> s, a, l
>>> + fn_start __aarch64_\insn\()\sz\()\model
>>> + mov tmp0_\sz, r0_\sz
>>> +0: ld\a\()xr\s r0_\sz, [x1]
>>> + .ifnc \insn, swp
>>> + \opc tmp1_\sz, r0_\sz, tmp0_\sz
>>> + .else
>>> + \opc tmp1_\sz, tmp0_\sz
>>> + .endif
>>> + st\l\()xr\s w15, tmp1_\sz, [x1]
>>> + cbnz w15, 0b
>>
>> I see at the end \s is in {,b,h} range.
>>
>> Don't you need to use x15 on 64-bit?
>
> Ard, I expanded all macros and reviewed this patch, but I am still
> having hard time to figure why w15 temp is OK instead of x15. Any hint?
>
Why do you think it should be x15?
>>
>>> + ret
>>> + fn_end __aarch64_\insn\()\sz\()\model
>>> + .endm
>>> +
>>> + /*
>>> + * Emit atomic helpers for \model for operand sizes in the
>>> + * set {1, 2, 4, 8}, for the instruction pattern given by
>>> + * \insn. (This is the LSE name, but this implementation uses
>>> + * the generic LL/SC sequence using \opc as the arithmetic
>>> + * operation on the target.)
>>> + */
>>> + .macro emit_ld, insn:req, opc:req, model:req, a, l
>>> + emit_ld_sz 1, \insn, \opc, \model, b, \a, \l
>>> + emit_ld_sz 2, \insn, \opc, \model, h, \a, \l
>>> + emit_ld_sz 4, \insn, \opc, \model, , \a, \l
>>> + emit_ld_sz 8, \insn, \opc, \model, , \a, \l
>>> + .endm
>>> +
>>> + /*
>>> + * Emit the compare and swap helper for \model and size \sz
>>> + * using LL/SC instructions.
>>> + */
>>> + .macro emit_cas_sz, sz:req, model:req, uxt:req, s, a, l
>>> + fn_start __aarch64_cas\sz\()\model
>>> + \uxt tmp0_\sz, r0_\sz
>>> +0: ld\a\()xr\s r0_\sz, [x2]
>>> + cmp r0_\sz, tmp0_\sz
>>> + bne 1f
>>> + st\l\()xr\s w15, r1_\sz, [x2]
>>> + cbnz w15, 0b
>>> +1: ret
>>> + fn_end __aarch64_cas\sz\()\model
>>> + .endm
>>> +
>>> + /*
>>> + * Emit compare-and-swap helpers for \model for operand sizes in
>>> the
>>> + * set {1, 2, 4, 8, 16}.
>>> + */
>>> + .macro emit_cas, model:req, a, l
>>> + emit_cas_sz 1, \model, uxtb, b, \a, \l
>>> + emit_cas_sz 2, \model, uxth, h, \a, \l
>>> + emit_cas_sz 4, \model, mov , , \a, \l
>>> + emit_cas_sz 8, \model, mov , , \a, \l
>>> +
>>> + /*
>>> + * We cannot use the parameterized sequence for 16 byte CAS, so we
>>> + * need to define it explicitly.
>>> + */
>>> + fn_start __aarch64_cas16\model
>>> + mov x16, x0
>>> + mov x17, x1
>>> +0: ld\a\()xp x0, x1, [x4]
>>> + cmp x0, x16
>>> + ccmp x1, x17, #0, eq
>>> + bne 1f
>>> + st\l\()xp w15, x16, x17, [x4]
>>> + cbnz w15, 0b
>>> +1: ret
>>> + fn_end __aarch64_cas16\model
>>> + .endm
>>> +
>>> + /*
>>> + * Emit the set of GCC outline atomic helper functions for
>>> + * the memory ordering model given by \model:
>>> + * - relax unordered loads and stores
>>> + * - acq load-acquire, unordered store
>>> + * - rel unordered load, store-release
>>> + * - acq_rel load-acquire, store-release
>>> + */
>>> + .macro emit_model, model:req, a, l
>>> + emit_ld ldadd, add, \model, \a, \l
>>> + emit_ld ldclr, bic, \model, \a, \l
>>> + emit_ld ldeor, eor, \model, \a, \l
>>> + emit_ld ldset, orr, \model, \a, \l
>>> + emit_ld swp, mov, \model, \a, \l
>>> + emit_cas \model, \a, \l
>>> + .endm
>>> +
>>> + emit_model _relax
>>> + emit_model _acq, a
>>> + emit_model _rel,, l
>>> + emit_model _acq_rel, a, l
>>>
>
>
>
>
next prev parent reply other threads:[~2020-05-21 16:45 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-20 11:44 [PATCH v2] ArmPkg/CompilerIntrinsicsLib: provide atomics intrinsics Ard Biesheuvel
2020-05-20 12:37 ` [edk2-devel] " Philippe Mathieu-Daudé
2020-05-21 16:40 ` Philippe Mathieu-Daudé
2020-05-21 16:45 ` Ard Biesheuvel [this message]
2020-05-21 16:59 ` Philippe Mathieu-Daudé
2020-05-21 17:02 ` Ard Biesheuvel
2020-05-21 17:07 ` Philippe Mathieu-Daudé
2020-05-20 15:59 ` Laszlo Ersek
2020-05-21 11:23 ` Leif Lindholm
2020-05-21 12:58 ` [edk2-devel] " Ard Biesheuvel
2020-05-21 13:16 ` Leif Lindholm
2020-05-21 13:31 ` Ard Biesheuvel
2020-05-21 14:16 ` Leif Lindholm
2020-05-21 20:22 ` Laszlo Ersek
2020-05-22 8:16 ` Ard Biesheuvel
2020-05-22 10:54 ` Leif Lindholm
2020-05-22 13:27 ` Ard Biesheuvel
2020-05-22 19:04 ` Laszlo Ersek
2020-06-02 0:50 ` Liming Gao
2020-06-02 7:09 ` Ard Biesheuvel
2020-05-28 1:36 ` Gary Lin
2020-05-28 9:49 ` Leif Lindholm
2020-05-28 15:48 ` Liming Gao
2020-05-28 20:09 ` Laszlo Ersek
2020-05-29 3:04 ` Liming Gao
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-list from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=c5ade5be-9c21-9797-df72-c42317445b4d@arm.com \
--to=devel@edk2.groups.io \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox