From: "Zhao, Weiming" <weimingz@codeaurora.org>
To: musl@lists.openwall.com
Subject: Re: build musl for armv7m
Date: Wed, 22 Jun 2016 13:37:38 -0700 [thread overview]
Message-ID: <34aab271-7305-3376-f5ad-7ae161c93428@codeaurora.org> (raw)
In-Reply-To: <20160622191940.GS10893@brightrain.aerifal.cx>
[-- Attachment #1: Type: text/plain, Size: 2252 bytes --]
How about using the following test?
!defined(__thumb__) || (__ARM_ARCH_7A__ || __ARM_ARCH_7R__ || __ARM_ARCH
> 7)
Because I think those instructions are ISA related, not just arm/thumb.
For example, using -mcpu=cortex-a8 -mthumb, the original code still
works. So if only testing __thumb__, the armv7-a thumb mode will fall
into the slow path too.
On 6/22/2016 12:19 PM, Rich Felker wrote:
> On Wed, Jun 22, 2016 at 12:08:13PM -0700, Zhao, Weiming wrote:
>> Thanks for reviewing.
>>
>> I add tests for ARMv7m for memcpy.
>>
>> For atomics.s, I think the below are equivalent:
>>
>> ldr ip,1f ==> assembler will computes the offset from current inst to the label
>>
>> - ldr ip,[pc,ip] ==> here the address to be loaded is current PC + ip
>> + add ip,pc,ip ==> here, the PC is the same as above
>> + ldr ip,[ip]
> In the original code, pc reads as the address of the instruction
> following the ".word" mnemonic (2 ARM insns ahead of the current
> instruction). In your version, I believe pc reads as the address of
> the .word (2 thumb insns ahead of the current insn) which would be
> wrong, but I may be wrong; one source I saw suggested that arithmetic
> on pc like this was not even defined in thumb mode on some models.
>
>> But I'm not familiar with the CP15 issue you mentioned.
>> So, anyway, I skip the change for atomics.s in this patch.
> OK. But just know that you can't expect it to work unless that's
> implemented.
>> diff --git a/src/string/arm/memcpy_le.S b/src/string/arm/memcpy_le.S
>> index 4db4844..1137f55 100644
>> --- a/src/string/arm/memcpy_le.S
>> +++ b/src/string/arm/memcpy_le.S
>> @@ -241,7 +241,12 @@ non_congruent:
>> beq 2f
>> ldr r5, [r1], #4
>> sub r2, r2, #4
>> +#if (__ARM_ARCH_7A || __ARM_ARCH_7R || __ARM_ARCH > 7)
>> orr r4, r3, r5, lsl lr
>> +#else
>> + lsl r4, r5, lr
>> + orr r4, r3, r4
>> +#endif
> These are not the right conditions. The selection should be made based
> on whether the code is being assembled as thumb, not on the ISA
> revision level. As written your patch uses the thumb code on all
> pre-v7 targets.
> Rich
--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation
[-- Attachment #2: patch.diff --]
[-- Type: text/plain, Size: 1600 bytes --]
diff --git a/src/setjmp/arm/longjmp.s b/src/setjmp/arm/longjmp.s
index e28d8f3..e9b9b32 100644
--- a/src/setjmp/arm/longjmp.s
+++ b/src/setjmp/arm/longjmp.s
@@ -8,7 +8,9 @@ longjmp:
mov ip,r0
movs r0,r1
moveq r0,#1
- ldmia ip!, {v1,v2,v3,v4,v5,v6,sl,fp,sp,lr}
+ ldmia ip!, {v1,v2,v3,v4,v5,v6,sl,fp}
+ ldr sp, [ip]!
+ ldr lr, [ip]!
adr r1,1f
ldr r2,1f
diff --git a/src/setjmp/arm/setjmp.s b/src/setjmp/arm/setjmp.s
index 8779163..fd380b0 100644
--- a/src/setjmp/arm/setjmp.s
+++ b/src/setjmp/arm/setjmp.s
@@ -9,7 +9,9 @@ __setjmp:
_setjmp:
setjmp:
mov ip,r0
- stmia ip!,{v1,v2,v3,v4,v5,v6,sl,fp,sp,lr}
+ stmia ip!,{v1,v2,v3,v4,v5,v6,sl,fp}
+ str sp, [ip]!
+ str lr, [ip]!
mov r0,#0
adr r1,1f
diff --git a/src/string/arm/memcpy_le.S b/src/string/arm/memcpy_le.S
index 4db4844..7c61979 100644
--- a/src/string/arm/memcpy_le.S
+++ b/src/string/arm/memcpy_le.S
@@ -241,7 +241,12 @@ non_congruent:
beq 2f
ldr r5, [r1], #4
sub r2, r2, #4
+#if !defined(__thumb__) || (__ARM_ARCH_7A__ || __ARM_ARCH_7R__ || __ARM_ARCH > 7)
orr r4, r3, r5, lsl lr
+#else
+ lsl r4, r5, lr
+ orr r4, r3, r4
+#endif
mov r3, r5, lsr r12
str r4, [r0], #4
cmp r2, #4
@@ -348,7 +353,12 @@ less_than_thirtytwo:
1: ldr r5, [r1], #4
sub r2, r2, #4
+#if !defined(__thumb__) || (__ARM_ARCH_7A__ || __ARM_ARCH_7R__ || __ARM_ARCH > 7)
orr r4, r3, r5, lsl lr
+#else
+ lsl r4, r5, lr
+ orr r4, r3, r4
+#endif
mov r3, r5, lsr r12
str r4, [r0], #4
cmp r2, #4
next prev parent reply other threads:[~2016-06-22 20:37 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-06-14 8:49 weimingz
2016-06-14 13:00 ` Rich Felker
2016-06-14 16:12 ` Zhao, Weiming
2016-06-14 16:32 ` Szabolcs Nagy
2016-06-14 16:58 ` Zhao, Weiming
2016-06-14 17:40 ` Zhao, Weiming
2016-06-16 18:34 ` Zhao, Weiming
2016-06-20 19:58 ` Rich Felker
2016-06-22 19:08 ` Zhao, Weiming
2016-06-22 19:19 ` Rich Felker
2016-06-22 20:37 ` Zhao, Weiming [this message]
2016-06-22 23:26 ` Rich Felker
2016-06-23 0:21 ` Zhao, Weiming
2016-06-23 4:22 ` Rich Felker
2016-06-23 6:04 ` weimingz
2016-06-23 9:57 ` Szabolcs Nagy
2016-06-23 14:22 ` weimingz
2016-07-05 20:08 ` Rich Felker
2016-06-20 17:17 ` Zhao, Weiming
2016-06-14 14:38 ` Rich Felker
2016-06-14 16:35 ` Zhao, Weiming
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-all 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=34aab271-7305-3376-f5ad-7ae161c93428@codeaurora.org \
--to=weimingz@codeaurora.org \
--cc=musl@lists.openwall.com \
/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.
Code repositories for project(s) associated with this public inbox
https://git.vuxu.org/mirror/musl/
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).