math test failures for master, 32bit abi of x86-64

math test failures for master, 32bit abi of x86-64

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 2081 bytes --]

Hi, 

with master on my laptop the following tests fail when building the 32bit abi for x86-64:
FAIL: math/test-double-j0
FAIL: math/test-float-ctan
FAIL: math/test-float-j1
FAIL: math/test-float-y0
FAIL: math/test-float-y1
FAIL: math/test-float32-ctan
FAIL: math/test-float32-j1
FAIL: math/test-float32-y0
FAIL: math/test-float32-y1
FAIL: math/test-float32x-j0
FAIL: math/test-float64-j0

The normal 64bit abi passes fine.

huettel@noumea ~ $ gcc --version
gcc (Gentoo 11.3.1_p20221209 p3) 11.3.1 20221209
[...]
huettel@noumea ~ $ ld --version
GNU ld (Gentoo 2.39 p5) 2.39.0
[...]
huettel@noumea ~ $ uname -a
Linux noumea 5.15.85-gentoo #2 SMP PREEMPT Mon Dec 26 23:22:44 CET 2022 x86_64 Intel(R) Core(TM) i7-8665U CPU @ 1.90GHz GenuineIntel GNU/Linux

Here's the result from "make regenerate-ulps":

diff --git a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
index 284e3fa275..8705a7822c 100644
--- a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
+++ b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
@@ -1027,7 +1027,7 @@ ldouble: 3
 
 Function: Imaginary part of "ctan_upward":
 double: 2
-float: 1
+float: 2
 float128: 5
 ldouble: 3
 
@@ -1260,13 +1260,13 @@ float128: 2
 ldouble: 8
 
 Function: "j0_downward":
-double: 5
+double: 6
 float: 9
 float128: 9
 ldouble: 6
 
 Function: "j0_towardzero":
-double: 6
+double: 7
 float: 9
 float128: 9
 ldouble: 6
@@ -1285,13 +1285,13 @@ ldouble: 9
 
 Function: "j1_downward":
 double: 5
-float: 6
+float: 8
 float128: 4
 ldouble: 4
 
 Function: "j1_towardzero":
 double: 4
-float: 6
+float: 8
 float128: 4
 ldouble: 4
 
@@ -1617,7 +1617,7 @@ ldouble: 5
 
 Function: "y0_towardzero":
 double: 4
-float: 4
+float: 8
 float128: 3
 ldouble: 8
 
@@ -1641,7 +1641,7 @@ ldouble: 7
 
 Function: "y1_towardzero":
 double: 3
-float: 4
+float: 9
 float128: 6
 ldouble: 5
 

-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer 
(council, qa, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: math test failures for master, 32bit abi of x86-64

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 2275 bytes --]

Am Montag, 2. Januar 2023, 00:39:19 CET schrieb Andreas K. Huettel via Libc-alpha:
> Hi, 
> 
> with master on my laptop the following tests fail when building the 32bit abi for x86-64:
> FAIL: math/test-double-j0
> FAIL: math/test-float-ctan
> FAIL: math/test-float-j1
> FAIL: math/test-float-y0
> FAIL: math/test-float-y1
> FAIL: math/test-float32-ctan
> FAIL: math/test-float32-j1
> FAIL: math/test-float32-y0
> FAIL: math/test-float32-y1
> FAIL: math/test-float32x-j0
> FAIL: math/test-float64-j0
> 
> The normal 64bit abi passes fine.
> 

FWIW here's the build settings:

 * ==== glibc build environment ========================================================
 *             ABI:   x86
 *          CBUILD:   x86_64-pc-linux-gnu
 *           CHOST:   x86_64-pc-linux-gnu
 *         CTARGET:   x86_64-pc-linux-gnu
 *      CBUILD_OPT:   i686-pc-linux-gnu
 *     CTARGET_OPT:   i686-pc-linux-gnu
 *              CC:   x86_64-pc-linux-gnu-gcc -m32 -mfpmath=sse -mstackrealign -march=native -pipe -O2 -Wl,-O1 -Wl,--as-needed
 *             CXX:   x86_64-pc-linux-gnu-g++ -m32 -mfpmath=sse -mstackrealign -march=native -pipe -O2
 *             CPP:   
 *              LD:   
 *         ASFLAGS:   
 *          CFLAGS:   -march=native -pipe -O2
 *        CPPFLAGS:   
 *        CXXFLAGS:   -march=native -pipe -O2
 *         LDFLAGS:   -Wl,-O1 -Wl,--as-needed
 *        MAKEINFO:   /dev/null
 *              NM:   x86_64-pc-linux-gnu-nm
 *              AR:   
 *              AS:   
 *           STRIP:   
 *          RANLIB:   
 *         OBJCOPY:   
 *         STRINGS:   
 *         OBJDUMP:   
 *         READELF:   x86_64-pc-linux-gnu-readelf
 * =====================================================================================


> huettel@noumea ~ $ gcc --version
> gcc (Gentoo 11.3.1_p20221209 p3) 11.3.1 20221209
> [...]
> huettel@noumea ~ $ ld --version
> GNU ld (Gentoo 2.39 p5) 2.39.0
> [...]
> huettel@noumea ~ $ uname -a
> Linux noumea 5.15.85-gentoo #2 SMP PREEMPT Mon Dec 26 23:22:44 CET 2022 x86_64 Intel(R) Core(TM) i7-8665U CPU @ 1.90GHz GenuineIntel GNU/Linux
> 


-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer 
(council, qa, toolchain, base-system, perl, libreoffice)

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: math test failures for master, 32bit abi of x86-64

[Florian Weimer]

* Andreas K. Huettel via Libc-alpha:

> Hi, 
>
> with master on my laptop the following tests fail when building the 32bit abi for x86-64:
> FAIL: math/test-double-j0
> FAIL: math/test-float-ctan
> FAIL: math/test-float-j1
> FAIL: math/test-float-y0
> FAIL: math/test-float-y1
> FAIL: math/test-float32-ctan
> FAIL: math/test-float32-j1
> FAIL: math/test-float32-y0
> FAIL: math/test-float32-y1
> FAIL: math/test-float32x-j0
> FAIL: math/test-float64-j0
>
> The normal 64bit abi passes fine.
>
> huettel@noumea ~ $ gcc --version
> gcc (Gentoo 11.3.1_p20221209 p3) 11.3.1 20221209
> [...]
> huettel@noumea ~ $ ld --version
> GNU ld (Gentoo 2.39 p5) 2.39.0
> [...]
> huettel@noumea ~ $ uname -a
> Linux noumea 5.15.85-gentoo #2 SMP PREEMPT Mon Dec 26 23:22:44 CET 2022 x86_64 Intel(R) Core(TM) i7-8665U CPU @ 1.90GHz GenuineIntel GNU/Linux
>
> Here's the result from "make regenerate-ulps":
>
> diff --git a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
> index 284e3fa275..8705a7822c 100644
> --- a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
> +++ b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
> @@ -1027,7 +1027,7 @@ ldouble: 3

Could you resend this with a proper commit message and Cc:
<fw@deneb.enyo.de>, please?  Then I can apply it with “git am” and push
it.  The patch itself looks fine.

Reviewed-by: Florian Weimer <fweimer@redhat.com>

Thanks,
Florian

[PATCH] i686: Regenerate ulps

[Andreas K. Hüttel]

Reviewed-by: Florian Weimer <fweimer@redhat.com>
---
 sysdeps/i386/i686/fpu/multiarch/libm-test-ulps | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
index 284e3fa275..8705a7822c 100644
--- a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
+++ b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
@@ -1027,7 +1027,7 @@ ldouble: 3
 
 Function: Imaginary part of "ctan_upward":
 double: 2
-float: 1
+float: 2
 float128: 5
 ldouble: 3
 
@@ -1260,13 +1260,13 @@ float128: 2
 ldouble: 8
 
 Function: "j0_downward":
-double: 5
+double: 6
 float: 9
 float128: 9
 ldouble: 6
 
 Function: "j0_towardzero":
-double: 6
+double: 7
 float: 9
 float128: 9
 ldouble: 6
@@ -1285,13 +1285,13 @@ ldouble: 9
 
 Function: "j1_downward":
 double: 5
-float: 6
+float: 8
 float128: 4
 ldouble: 4
 
 Function: "j1_towardzero":
 double: 4
-float: 6
+float: 8
 float128: 4
 ldouble: 4
 
@@ -1617,7 +1617,7 @@ ldouble: 5
 
 Function: "y0_towardzero":
 double: 4
-float: 4
+float: 8
 float128: 3
 ldouble: 8
 
@@ -1641,7 +1641,7 @@ ldouble: 7
 
 Function: "y1_towardzero":
 double: 3
-float: 4
+float: 9
 float128: 6
 ldouble: 5
 
-- 
2.38.2