This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions

From: Andrew Senkevich <andrew dot n dot senkevich at gmail dot com>
To: "H.J. Lu" <hjl dot tools at gmail dot com>
Cc: Joseph Myers <joseph at codesourcery dot com>, libc-alpha <libc-alpha at sourceware dot org>
Date: Tue, 16 Feb 2016 15:04:12 +0300
Subject: Re: [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions
Authentication-results: sourceware.org; auth=none
References: <CAMXFM3tML81iuKQMKRU-T4Fw0+=sYk0q_BNavMGagt21VcYvzQ at mail dot gmail dot com> <CAMe9rOp7BF2avdWbGTbsxuYQV_rmXSxzDUAwz+nHK-GhWqPKJA at mail dot gmail dot com> <CAMXFM3sG90jn3Em-REfaqtj3OGAjh51OcO=yn1qHqJ4aStkPrg at mail dot gmail dot com> <alpine dot DEB dot 2 dot 10 dot 1602111641250 dot 29940 at digraph dot polyomino dot org dot uk> <CAMXFM3uxH=0DHnwikPjs2AobQ0kxOqKg+j=CuOWX=RLB_i8hxg at mail dot gmail dot com> <CAMe9rOoWy+hqiCOKqFx0nOFVRT_kBETJ5hYE+cY6pFWUke=tkw at mail dot gmail dot com>

2016-02-11 20:25 GMT+03:00 H.J. Lu <hjl.tools@gmail.com>:
> On Thu, Feb 11, 2016 at 9:21 AM, Andrew Senkevich
> <andrew.n.senkevich@gmail.com> wrote:
>> 2016-02-11 19:43 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
>>> On Thu, 11 Feb 2016, Andrew Senkevich wrote:
>>>
>>>> If we need runtime tests with calls to finite aliases it looks better
>>>> to adopt build of existing libmvec tests. We can call finite aliases
>>>> in *-wrappers.c and build shared library from them and link with it
>>>> test binaries. Is this approach looks OK?
>>>
>>> I don't think the finite aliases should be considered part of the API to
>>> test; it's an implementation detail of the header that they may get
>>> referenced in certain circumstances.
>>
>> I agree.
>>
>>>  The relevant thing to test is
>>> whether building a program that directly calls the scalar functions, with
>>> options such that the calls get vectorized, works (including with variant
>>> options for e.g. LTO).
>>
>> But it looks more like compiler test, not library. Or do you mean some
>> ABI test for vector functions?
>> In any case is it necessary for this patch or the following patch is
>> Ok for thrunk?
>>
>> diff --git a/ChangeLog b/ChangeLog
>> index 11c3156..6ebcefb 100644
>> --- a/ChangeLog
>> +++ b/ChangeLog
>> @@ -1,3 +1,9 @@
>> +2016-02-10  Andrew Senkevich  <andrew.senkevich@intel.com>
>> +           Carlos O'Donell  <carlos@redhat.com>
>> +
>> +       [BZ #19590]
>> +       * sysdeps/x86_64/fpu/svml_finite_alias.S (ALIAS_IMPL): Use PLT.
>> +
>>  2016-02-04  Rajalakshmi Srinivasaraghavan  <raji@linux.vnet.ibm.com>
>>
>>         * sysdeps/powerpc/fpu/libm-test-ulps: Regenerated.
>> diff --git a/sysdeps/x86_64/fpu/svml_finite_alias.S
>> b/sysdeps/x86_64/fpu/svml_finite_alias.S
>> index 0062fe4..8314cf4 100644
>> --- a/sysdeps/x86_64/fpu/svml_finite_alias.S
>> +++ b/sysdeps/x86_64/fpu/svml_finite_alias.S
>> @@ -23,8 +23,7 @@
>>
>>  #define ALIAS_IMPL(alias, target) \
>>  ENTRY (alias); \
>> -       call target; \
>> -       ret; \
>> +       jmp target@PLT; \
>>  END (alias)
>>
>>         .text
>>
>
> You need to a test to show your change fixes something.

Here is patch with tests.

diff --git a/ChangeLog b/ChangeLog
index a2b394e..f88ca52 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,15 @@
+2016-02-16  Andrew Senkevich  <andrew.senkevich@intel.com>
+           H.J. Lu  <hongjiu.lu@intel.com>
+
+       * sysdeps/x86_64/fpu/Makefile: Added new tests.
+       * sysdeps/x86_64/fpu/svml_finite_alias.S (ALIAS_IMPL):: Use PLT.
+       * sysdeps/x86_64/fpu/test-libmvec-alias-mod.c: New.
+       * sysdeps/x86_64/fpu/test-libmvec-alias.c: Likewise.
+       * sysdeps/x86_64/fpu/test-libmvec-alias-avx512-mod.c: Likewise.
+       * sysdeps/x86_64/fpu/test-libmvec-alias-avx512.c: Likewise.
+       * sysdeps/x86_64/fpu/test-libmvec-alias-wrappers.c: Likewise.
+       * sysdeps/x86_64/fpu/test-libmvec-alias-avx512-wrappers.c: Likewise.
+
 2016-02-14  Carlos O'Donelll  <carlos@redhat.com>

        * manual/install.texi: Latest tested is GCC 5.3, texinfo 6.0, gawk
diff --git a/sysdeps/x86_64/fpu/Makefile b/sysdeps/x86_64/fpu/Makefile
index 88742fa..1db3532 100644
--- a/sysdeps/x86_64/fpu/Makefile
+++ b/sysdeps/x86_64/fpu/Makefile
@@ -29,10 +29,22 @@ endif
 ifeq ($(subdir),math)
 ifeq ($(build-mathvec),yes)
 libmvec-tests += double-vlen2 double-vlen4 double-vlen4-avx2 \
-                float-vlen4 float-vlen8 float-vlen8-avx2
+                float-vlen4 float-vlen8 float-vlen8-avx2 libmvec-alias
+modules-names += test-libmvec-alias-mod
+test-libmvec-alias-mod.so-no-z-defs = yes
+
+$(objpfx)test-libmvec-alias: $(objpfx)test-libmvec-alias-mod.so
+$(objpfx)test-libmvec-alias-mod.so: $(objpfx)../mathvec/libmvec_nonshared.a \
+                                   $(libmvec)

 ifeq (yes,$(config-cflags-avx512))
-libmvec-tests += double-vlen8 float-vlen16
+libmvec-tests += double-vlen8 float-vlen16 libmvec-alias-avx512
+modules-names += test-libmvec-alias-avx512-mod
+test-libmvec-alias-avx512-mod.so-no-z-defs = yes
+
+$(objpfx)test-libmvec-alias-avx512: $(objpfx)test-libmvec-alias-avx512-mod.so
+$(objpfx)test-libmvec-alias-avx512-mod.so: \
+                       $(objpfx)../mathvec/libmvec_nonshared.a $(libmvec)
 endif

 double-vlen4-arch-ext-cflags = -mavx
@@ -43,6 +55,9 @@ float-vlen8-arch-ext-cflags = -mavx
 float-vlen8-arch-ext2-cflags = -mavx2
 float-vlen16-arch-ext-cflags = -mavx512f

+CFLAGS-test-libmvec-alias-mod.c = $(double-vlen4-arch-ext2-cflags)
+CFLAGS-test-libmvec-alias-avx512-mod.c = $(double-vlen8-arch-ext-cflags)
+
 CFLAGS-test-double-vlen4-avx2.c = $(libm-test-vec-cflags)
 CFLAGS-test-double-vlen4-avx2-wrappers.c = $(double-vlen4-arch-ext2-cflags)

diff --git a/sysdeps/x86_64/fpu/svml_finite_alias.S
b/sysdeps/x86_64/fpu/svml_finite_alias.S
index 0062fe4..8314cf4 100644
--- a/sysdeps/x86_64/fpu/svml_finite_alias.S
+++ b/sysdeps/x86_64/fpu/svml_finite_alias.S
@@ -23,8 +23,7 @@

 #define ALIAS_IMPL(alias, target) \
 ENTRY (alias); \
-       call target; \
-       ret; \
+       jmp target@PLT; \
 END (alias)

        .text
diff --git a/sysdeps/x86_64/fpu/test-libmvec-alias-avx512-mod.c
b/sysdeps/x86_64/fpu/test-libmvec-alias-avx512-mod.c
new file mode 100644
index 0000000..64218f8
--- /dev/null
+++ b/sysdeps/x86_64/fpu/test-libmvec-alias-avx512-mod.c
@@ -0,0 +1,46 @@
+/* Part of test to build shared library to ensure link against
+   *_finite aliases from libmvec.
+   Copyright (C) 2016 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <immintrin.h>
+
+#include "test-double-vlen8.h"
+#define VEC_TYPE __m512d
+
+VECTOR_WRAPPER (WRAPPER_NAME (log), _ZGVeN8v___log_finite)
+VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVeN8v___exp_finite)
+VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVeN8vv___pow_finite)
+
+#undef FUNC
+#undef FLOAT
+#undef BUILD_COMPLEX
+#undef TEST_MSG
+#undef CHOOSE
+#undef FUNC_TEST
+#undef VEC_TYPE
+#undef VECTOR_WRAPPER
+#undef VECTOR_WRAPPER_ff
+#undef VEC_SUFF
+#undef VEC_LEN
+
+#include "test-float-vlen16.h"
+#define VEC_TYPE __m512
+
+VECTOR_WRAPPER (WRAPPER_NAME (logf), _ZGVeN16v___logf_finite)
+VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVeN16v___expf_finite)
+VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVeN16vv___powf_finite)
diff --git a/sysdeps/x86_64/fpu/test-libmvec-alias-avx512-wrappers.c
b/sysdeps/x86_64/fpu/test-libmvec-alias-avx512-wrappers.c
new file mode 100644
index 0000000..9d841c9
--- /dev/null
+++ b/sysdeps/x86_64/fpu/test-libmvec-alias-avx512-wrappers.c
@@ -0,0 +1 @@
+/* Dummy file.  */
diff --git a/sysdeps/x86_64/fpu/test-libmvec-alias-avx512.c
b/sysdeps/x86_64/fpu/test-libmvec-alias-avx512.c
new file mode 100644
index 0000000..8d11234
--- /dev/null
+++ b/sysdeps/x86_64/fpu/test-libmvec-alias-avx512.c
@@ -0,0 +1,43 @@
+/* Part of test to ensure link against *_finite aliases from libmvec.
+   Copyright (C) 2016 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <init-arch.h>
+
+extern double log_vlen8 (double);
+extern double exp_vlen8 (double);
+extern double pow_vlen8 (double, double);
+
+extern float logf_vlen16 (float);
+extern float expf_vlen16 (float);
+extern float powf_vlen16 (float, float);
+
+int main(void)
+{
+  if (!HAS_ARCH_FEATURE (AVX512F_Usable)) return 0;
+
+  if (log_vlen8(1.0) != 0.0
+      || logf_vlen16(1.0) != 0.0) abort();
+
+  if (exp_vlen8(0.0) != 1.0
+      || expf_vlen16(0.0) != 1.0) abort();
+
+  if (pow_vlen8(1.0, 1.0) != 1.0
+      || powf_vlen16(1.0, 1.0) != 1.0) abort();
+
+  return 0;
+}
diff --git a/sysdeps/x86_64/fpu/test-libmvec-alias-mod.c
b/sysdeps/x86_64/fpu/test-libmvec-alias-mod.c
new file mode 100644
index 0000000..37c2d75
--- /dev/null
+++ b/sysdeps/x86_64/fpu/test-libmvec-alias-mod.c
@@ -0,0 +1,86 @@
+/* Part of test to build shared library to ensure link against
+   *_finite aliases from libmvec.
+   Copyright (C) 2016 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <immintrin.h>
+
+#include "test-double-vlen2.h"
+#define VEC_TYPE __m128d
+
+VECTOR_WRAPPER (WRAPPER_NAME (log), _ZGVbN2v___log_finite)
+VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVbN2v___exp_finite)
+VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVbN2vv___pow_finite)
+
+#undef VEC_TYPE
+#undef VECTOR_WRAPPER
+#undef VECTOR_WRAPPER_ff
+#undef VEC_SUFF
+#undef VEC_LEN
+
+#include "test-double-vlen4.h"
+#define VEC_TYPE __m256d
+
+VECTOR_WRAPPER (WRAPPER_NAME (log), _ZGVcN4v___log_finite)
+VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVcN4v___exp_finite)
+VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVcN4vv___pow_finite)
+
+#undef VEC_SUFF
+#define VEC_SUFF _vlen4_avx2
+
+VECTOR_WRAPPER (WRAPPER_NAME (log), _ZGVdN4v___log_finite)
+VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVdN4v___exp_finite)
+VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVdN4vv___pow_finite)
+
+#undef FUNC
+#undef FLOAT
+#undef BUILD_COMPLEX
+#undef TEST_MSG
+#undef CHOOSE
+#undef FUNC_TEST
+#undef VEC_TYPE
+#undef VECTOR_WRAPPER
+#undef VECTOR_WRAPPER_ff
+#undef VEC_SUFF
+#undef VEC_LEN
+
+#include "test-float-vlen4.h"
+#define VEC_TYPE __m128
+
+VECTOR_WRAPPER (WRAPPER_NAME (logf), _ZGVbN4v___logf_finite)
+VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVbN4v___expf_finite)
+VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVbN4vv___powf_finite)
+
+#undef VEC_TYPE
+#undef VECTOR_WRAPPER
+#undef VECTOR_WRAPPER_ff
+#undef VEC_SUFF
+#undef VEC_LEN
+
+#include "test-float-vlen8.h"
+#define VEC_TYPE __m256
+
+VECTOR_WRAPPER (WRAPPER_NAME (logf), _ZGVcN8v___logf_finite)
+VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVcN8v___expf_finite)
+VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVcN8vv___powf_finite)
+
+#undef VEC_SUFF
+#define VEC_SUFF _vlen8_avx2
+
+VECTOR_WRAPPER (WRAPPER_NAME (logf), _ZGVdN8v___logf_finite)
+VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVdN8v___expf_finite)
+VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVdN8vv___powf_finite)
diff --git a/sysdeps/x86_64/fpu/test-libmvec-alias-wrappers.c
b/sysdeps/x86_64/fpu/test-libmvec-alias-wrappers.c
new file mode 100644
index 0000000..9d841c9
--- /dev/null
+++ b/sysdeps/x86_64/fpu/test-libmvec-alias-wrappers.c
@@ -0,0 +1 @@
+/* Dummy file.  */
diff --git a/sysdeps/x86_64/fpu/test-libmvec-alias.c
b/sysdeps/x86_64/fpu/test-libmvec-alias.c
new file mode 100644
index 0000000..429b19e
--- /dev/null
+++ b/sysdeps/x86_64/fpu/test-libmvec-alias.c
@@ -0,0 +1,81 @@
+/* Part of test to ensure link against *_finite aliases from libmvec.
+   Copyright (C) 2016 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <init-arch.h>
+
+extern double log_vlen2 (double);
+extern double exp_vlen2 (double);
+extern double pow_vlen2 (double, double);
+
+extern double log_vlen4 (double);
+extern double exp_vlen4 (double);
+extern double pow_vlen4 (double, double);
+
+extern double log_vlen4_avx2 (double);
+extern double exp_vlen4_avx2 (double);
+extern double pow_vlen4_avx2 (double, double);
+
+extern float logf_vlen4 (float);
+extern float expf_vlen4 (float);
+extern float powf_vlen4 (float, float);
+
+extern float logf_vlen8 (float);
+extern float expf_vlen8 (float);
+extern float powf_vlen8 (float, float);
+
+extern float logf_vlen8_avx2 (float);
+extern float expf_vlen8_avx2 (float);
+extern float powf_vlen8_avx2 (float, float);
+
+int main(void)
+{
+  if (log_vlen2(1.0) != 0.0
+      || logf_vlen4(1.0) != 0.0) abort();
+
+  if (exp_vlen2(0.0) != 1.0
+      || expf_vlen4(0.0) != 1.0) abort();
+
+  if (pow_vlen2(1.0, 1.0) != 1.0
+      || powf_vlen4(1.0, 1.0) != 1.0) abort();
+
+  if (HAS_ARCH_FEATURE (AVX_Usable))
+    {
+      if (log_vlen4(1.0) != 0.0
+          || logf_vlen8(1.0) != 0.0) abort();
+
+      if (exp_vlen4(0.0) != 1.0
+          || expf_vlen8(0.0) != 1.0) abort();
+
+      if (pow_vlen4(1.0, 1.0) != 1.0
+          || powf_vlen8(1.0, 1.0) != 1.0) abort();
+    }
+
+  if (HAS_ARCH_FEATURE (AVX2_Usable))
+    {
+      if (log_vlen4_avx2(1.0) != 0.0
+          || logf_vlen8_avx2(1.0) != 0.0) abort();
+
+      if (exp_vlen4_avx2(0.0) != 1.0
+          || expf_vlen8_avx2(0.0) != 1.0) abort();
+
+      if (pow_vlen4_avx2(1.0, 1.0) != 1.0
+          || powf_vlen8_avx2(1.0, 1.0) != 1.0) abort();
+    }
+
+  return 0;
+}

Is it Ok for trunk?


--
WBR,
Andrew

Attachment: bz19590.patch
Description: Binary data

Follow-Ups:
- Re: [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions
  - From: H.J. Lu
- Re: [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions
  - From: Joseph Myers

References:
- [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions
  - From: Andrew Senkevich
- Re: [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions
  - From: H.J. Lu
- Re: [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions
  - From: Andrew Senkevich
- Re: [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions
  - From: Joseph Myers
- Re: [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions
  - From: Andrew Senkevich
- Re: [PATCH] BZ #19590: Fixed build of shared objects that use libmvec.so functions
  - From: H.J. Lu

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]