Re: [PATCH] clang compatibility: fpu options

Hi,

Please see attached re-formatted patch file.

Sincerely,

Volodymyr Turanskyy | Principal Software Engineer | CE SW LLVM Team

From: Volodymyr Turanskyy <[email protected]>
Date: Monday, 20 January 2025 at 15:06
To: Joel Sherrill <[email protected]>, [email protected] <[email protected]>
Cc: nd <[email protected]>
Subject: Re: [PATCH] clang compatibility: fpu options
Hi Joel,

Let me fix the commit messages.

Clang does not have vfpxd - vfpv3xd might be the closest, however we believe that it is not needed for clang at all, so it is skipped.

Sincerely,

Volodymyr Turanskyy | Principal Software Engineer | CE SW LLVM Team

From: Joel Sherrill <[email protected]>
Date: Monday, 20 January 2025 at 15:01
To: Volodymyr Turanskyy <[email protected]>
Cc: [email protected] <[email protected]>, nd <[email protected]>
Subject: Re: [PATCH] clang compatibility: fpu options
Same with missing commit.

Does the clang assembler have a directive for vfpxd? If I am reading the patch correctly, it just drops the directive rather than replacing it.

--joel

On Mon, Jan 20, 2025 at 8:49 AM Volodymyr Turanskyy <[email protected]<mailto:[email protected]>> wrote:
Hi,

Following up on the RFC [1], please find attached a patch that adds a check to avoid extra assembler directives that are not compatible with clang and are not needed by clang.

References

https://sourceware.org/pipermail/newlib/2025/021452.html

Sincerely,

Volodymyr Turanskyy | Principal Software Engineer | CE SW LLVM Team

From 5184bcdc36ed48cd2074757db5afe2e5de5b28d0 Mon Sep 17 00:00:00 2001
From: Volodymyr Turanskyy <[email protected]>
Date: Mon, 20 Jan 2025 15:18:06 +0000
Subject: [PATCH] Add check for clang to avoid unnecessary FPU directives

This improves compatibility with clang that does not support vfpxd and does not need these extra directives.

Change-Id: Id2027e622aef8457ac9c7e1d6715a9240ce8e3f0
---
 newlib/libc/machine/arm/setjmp.S | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/newlib/libc/machine/arm/setjmp.S b/newlib/libc/machine/arm/setjmp.S
index 0070f17cd..a53f7918e 100644
--- a/newlib/libc/machine/arm/setjmp.S
+++ b/newlib/libc/machine/arm/setjmp.S
@@ -78,10 +78,10 @@
     covers all the cases we need in this file for hardware
     floating-point and should be compatible with all required FPUs
     that we need to support.  */
-# if __ARM_FP
+# if __ARM_FP && !__clang__
 	.fpu vfpxd
 # endif
-# if __ARM_FEATURE_MVE
+# if __ARM_FEATURE_MVE && !__clang__
 	.arch_extension mve
 # endif
 #endif
-- 
2.48.0