GNU bug report logs - #69770
[PATCH] build: strengthen 16 bit float support checks

Previous Next

Full log

View this message in rfc822 format

From: help-debbugs <at> gnu.org (GNU bug Tracking System)
To: Pádraig Brady <P <at> draigBrady.com>
Cc: tracker <at> debbugs.gnu.org
Subject: bug#69770: closed ([PATCH] build: strengthen 16 bit float support
 checks)
Date: Wed, 13 Mar 2024 12:39:02 +0000

[Message part 1 (text/plain, inline)]

Your message dated Wed, 13 Mar 2024 12:36:51 +0000
with message-id <40cecf61-b314-8648-ede4-a86ffdeda226 <at> draigBrady.com>
and subject line Re: bug#69770: [PATCH] build: strengthen 16 bit float support checks
has caused the debbugs.gnu.org bug report #69770,
regarding [PATCH] build: strengthen 16 bit float support checks
to be marked as done.

(If you believe you have received this mail in error, please contact
help-debbugs <at> gnu.org.)


-- 
69770: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=69770
GNU Bug Tracking System
Contact help-debbugs <at> gnu.org with problems

[Message part 2 (message/rfc822, inline)]

From: Grisha Levit <grishalevit <at> gmail.com>
To: bug-coreutils <at> gnu.org
Cc: Grisha Levit <grishalevit <at> gmail.com>
Subject: [PATCH] build: strengthen 16 bit float support checks
Date: Tue, 12 Mar 2024 22:24:34 -0400

Recent clang provides __bf16 on aarch64 but it is broken.

If built with -O0, the conversion is wrong:

    $ printf '\x3F\x80' | od --end=big -An -tfB | tr -d ' '
    1.875

If built with -O1 or higher, compilation fails:

    fatal error: error in backend: Cannot select: 0xb400007a58d29780: f32 = fp_extend 0xb400007a58d31720
      0xb400007a58d31720: bf16,ch = CopyFromReg 0xb400007b78c53720, Register:bf16 %13
        0xb400007a58d29470: bf16 = Register %13
    In function: print_bfloat

The latter issue does not cause the existing configure test to fail
because the promotion is optimized out.

* configure.ac: Ensure 16 bit float promotion code does not get
optimized out, and produces an expected result.
---
 configure.ac | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/configure.ac b/configure.ac
index 248e30ca2..21bee28d1 100644
--- a/configure.ac
+++ b/configure.ac
@@ -569,13 +569,14 @@ ac_c_werror_flag=$cu_save_c_werror_flag
 
 # Test compiler support for half precision floating point types (for od)
 AC_MSG_CHECKING([IEEE 16 bit floating point])
- AC_COMPILE_IFELSE(
+ AC_RUN_IFELSE(
    [AC_LANG_SOURCE([[
      int
      main (void)
      {
-        _Float16 hf;
+        volatile _Float16 hf = 1;
         float f = hf;  /* Ensure compiler can promote to float.  */
+        return !(f == 1.0f);
      }
   ]])
   ],[
@@ -589,13 +590,14 @@ if test $ieee_16_bit_supported = yes; then
 fi
 
 AC_MSG_CHECKING([Brain 16 bit floating point])
- AC_COMPILE_IFELSE(
+ AC_RUN_IFELSE(
    [AC_LANG_SOURCE([[
      int
      main (void)
      {
-        __bf16 hf;
+        volatile __bf16 hf = 1;
         float f = hf;  /* Ensure compiler can promote to float.  */
+        return !(f == 1.0f);
      }
   ]])
   ],[
-- 
2.44.0

[Message part 3 (message/rfc822, inline)]

From: Pádraig Brady <P <at> draigBrady.com>
To: Grisha Levit <grishalevit <at> gmail.com>, 69770-done <at> debbugs.gnu.org
Subject: Re: bug#69770: [PATCH] build: strengthen 16 bit float support checks
Date: Wed, 13 Mar 2024 12:36:51 +0000

On 13/03/2024 02:24, Grisha Levit wrote:
> Recent clang provides __bf16 on aarch64 but it is broken.
> 
> If built with -O0, the conversion is wrong:
> 
>      $ printf '\x3F\x80' | od --end=big -An -tfB | tr -d ' '
>      1.875
> 
> If built with -O1 or higher, compilation fails:
> 
>      fatal error: error in backend: Cannot select: 0xb400007a58d29780: f32 = fp_extend 0xb400007a58d31720
>        0xb400007a58d31720: bf16,ch = CopyFromReg 0xb400007b78c53720, Register:bf16 %13
>          0xb400007a58d29470: bf16 = Register %13
>      In function: print_bfloat
> 
> The latter issue does not cause the existing configure test to fail
> because the promotion is optimized out.
> 
> * configure.ac: Ensure 16 bit float promotion code does not get
> optimized out, and produces an expected result.

Looks good!

I'll follow up with another patch to all these involved checks to be cached / bypassed,
by wrapping them in AC_CACHE_VAL().

Marking this as done.

thanks!
Pádraig

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.