* [PATCH 2/N] [x86_64] Vectorized math functions
@ 2014-11-20 17:40 Andrew Senkevich
2014-11-21 17:55 ` Joseph Myers
0 siblings, 1 reply; 13+ messages in thread
From: Andrew Senkevich @ 2014-11-20 17:40 UTC (permalink / raw)
To: libc-alpha
[-- Attachment #1: Type: text/plain, Size: 4956 bytes --]
Hi,
this is addition of several macros: for condition in tests for vector
functions, for forming name of tested function, for runtime
architecture check.
ChangeLog
2014-11-20 Andrew Senkevich <andrew.n.senkevich@gmail.com>
* math/libm-test.inc (TEST_MATHVEC, VEC_PREFIX, INIT_ARCH_EXT,
CHECK_ARCH_EXT, STR_CONCAT, STR_CON3): New macros. START refactored.
* sysdeps/x86_64/fpu/math-tests.h: New file.
diff --git a/math/libm-test.inc b/math/libm-test.inc
index df11343..94931e2 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -678,13 +678,17 @@ test_exceptions (const char *test_name, int exception)
feclearexcept (FE_ALL_EXCEPT);
}
+#ifndef TEST_MATHVEC
+# define TEST_MATHVEC 0
+#endif
+
/* Test whether errno for TEST_NAME, set to ERRNO_VALUE, has value
EXPECTED_VALUE (description EXPECTED_NAME). */
static void
test_single_errno (const char *test_name, int errno_value,
int expected_value, const char *expected_name)
{
-#ifndef TEST_INLINE
+#if !defined TEST_INLINE && !TEST_MATHVEC
if (errno_value == expected_value)
{
if (print_screen (1))
@@ -1690,13 +1694,27 @@ struct test_fFF_11_data
(ARRAY)[i].RM_##ROUNDING_MODE.extra2_expected); \
ROUND_RESTORE_ ## ROUNDING_MODE
+#ifndef INIT_ARCH_EXT
+# define INIT_ARCH_EXT
+# define CHECK_ARCH_EXT
+#endif
+
+#ifndef VEC_PREFIX
+# define VEC_PREFIX
+#endif
+
#ifndef FUNC_TEST
# define FUNC_TEST FUNC
#endif
+#define STR_CONCAT(a, b, c) __STRING (a##b##c)
+#define STR_CON3(a, b, c) STR_CONCAT (a, b, c)
+
/* Start and end the tests for a given function. */
-#define START(FUNC, EXACT) \
- const char *this_func = #FUNC; \
+#define START(FUNC, SUFF, EXACT) \
+ CHECK_ARCH_EXT \
+ if (TEST_MATHVEC) return; \
+ const char *this_func = STR_CON3 (VEC_PREFIX, FUNC, SUFF); \
init_max_error (this_func, EXACT)
#define END \
print_max_error (this_func)
@@ -1709,28 +1727,28 @@ struct test_fFF_11_data
{ \
do \
{ \
- START (FUNC, EXACT); \
+ START (FUNC,, EXACT); \
LOOP_MACRO (FUNC, ARRAY, , ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _downward, EXACT); \
+ START (FUNC, _downward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_DOWNWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _towardzero, EXACT); \
+ START (FUNC, _towardzero, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_TOWARDZERO, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _upward, EXACT); \
+ START (FUNC, _upward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_UPWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
@@ -6038,7 +6056,7 @@ static const struct test_c_c_data cexp_test_data[] =
static void
cexp_test (void)
{
- START (cexp, 0);
+ START (cexp,, 0);
RUN_TEST_LOOP_c_c (cexp, cexp_test_data, );
END_COMPLEX;
}
@@ -7552,7 +7570,7 @@ static const struct test_if_f_data jn_test_data[] =
static void
jn_test (void)
{
- START (jn, 0);
+ START (jn,, 0);
RUN_TEST_LOOP_if_f (jn, jn_test_data, );
END;
}
@@ -9378,7 +9396,7 @@ static const struct test_f_f_data tgamma_test_data[] =
static void
tgamma_test (void)
{
- START (tgamma, 0);
+ START (tgamma,, 0);
RUN_TEST_LOOP_f_f (tgamma, tgamma_test_data, );
END;
}
@@ -9828,6 +9846,8 @@ main (int argc, char **argv)
initialize ();
printf (TEST_MSG);
+ INIT_ARCH_EXT
+
check_ulp ();
/* Keep the tests a wee bit ordered (according to ISO C99). */
diff --git a/sysdeps/x86_64/fpu/math-tests.h b/sysdeps/x86_64/fpu/math-tests.h
new file mode 100644
index 0000000..466b97b
--- /dev/null
+++ b/sysdeps/x86_64/fpu/math-tests.h
@@ -0,0 +1,34 @@
+/* Configuration for math tests. x86_64 version.
+ Copyright (C) 2013-2014 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/>. */
+
+#ifdef REQUIRE_AVX2
+# include <init-arch.h>
+
+ static int avx2_usable; /* Set to 1 if AVX2 supported */
+
+# define INIT_ARCH_EXT \
+ __init_cpu_features (); \
+ avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
+ & bit_AVX2_Usable;
+
+# define CHECK_ARCH_EXT \
+ if (!avx2_usable) return;
+
+#endif
+
+#include_next <math-tests.h>
Attached the same.
--
WBR,
Andrew
[-- Attachment #2: libmvec_2_of_N.patch --]
[-- Type: application/octet-stream, Size: 4714 bytes --]
diff --git a/math/libm-test.inc b/math/libm-test.inc
index df11343..94931e2 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -678,13 +678,17 @@ test_exceptions (const char *test_name, int exception)
feclearexcept (FE_ALL_EXCEPT);
}
+#ifndef TEST_MATHVEC
+# define TEST_MATHVEC 0
+#endif
+
/* Test whether errno for TEST_NAME, set to ERRNO_VALUE, has value
EXPECTED_VALUE (description EXPECTED_NAME). */
static void
test_single_errno (const char *test_name, int errno_value,
int expected_value, const char *expected_name)
{
-#ifndef TEST_INLINE
+#if !defined TEST_INLINE && !TEST_MATHVEC
if (errno_value == expected_value)
{
if (print_screen (1))
@@ -1690,13 +1694,27 @@ struct test_fFF_11_data
(ARRAY)[i].RM_##ROUNDING_MODE.extra2_expected); \
ROUND_RESTORE_ ## ROUNDING_MODE
+#ifndef INIT_ARCH_EXT
+# define INIT_ARCH_EXT
+# define CHECK_ARCH_EXT
+#endif
+
+#ifndef VEC_PREFIX
+# define VEC_PREFIX
+#endif
+
#ifndef FUNC_TEST
# define FUNC_TEST FUNC
#endif
+#define STR_CONCAT(a, b, c) __STRING (a##b##c)
+#define STR_CON3(a, b, c) STR_CONCAT (a, b, c)
+
/* Start and end the tests for a given function. */
-#define START(FUNC, EXACT) \
- const char *this_func = #FUNC; \
+#define START(FUNC, SUFF, EXACT) \
+ CHECK_ARCH_EXT \
+ if (TEST_MATHVEC) return; \
+ const char *this_func = STR_CON3 (VEC_PREFIX, FUNC, SUFF); \
init_max_error (this_func, EXACT)
#define END \
print_max_error (this_func)
@@ -1709,28 +1727,28 @@ struct test_fFF_11_data
{ \
do \
{ \
- START (FUNC, EXACT); \
+ START (FUNC,, EXACT); \
LOOP_MACRO (FUNC, ARRAY, , ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _downward, EXACT); \
+ START (FUNC, _downward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_DOWNWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _towardzero, EXACT); \
+ START (FUNC, _towardzero, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_TOWARDZERO, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _upward, EXACT); \
+ START (FUNC, _upward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_UPWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
@@ -6038,7 +6056,7 @@ static const struct test_c_c_data cexp_test_data[] =
static void
cexp_test (void)
{
- START (cexp, 0);
+ START (cexp,, 0);
RUN_TEST_LOOP_c_c (cexp, cexp_test_data, );
END_COMPLEX;
}
@@ -7552,7 +7570,7 @@ static const struct test_if_f_data jn_test_data[] =
static void
jn_test (void)
{
- START (jn, 0);
+ START (jn,, 0);
RUN_TEST_LOOP_if_f (jn, jn_test_data, );
END;
}
@@ -9378,7 +9396,7 @@ static const struct test_f_f_data tgamma_test_data[] =
static void
tgamma_test (void)
{
- START (tgamma, 0);
+ START (tgamma,, 0);
RUN_TEST_LOOP_f_f (tgamma, tgamma_test_data, );
END;
}
@@ -9828,6 +9846,8 @@ main (int argc, char **argv)
initialize ();
printf (TEST_MSG);
+ INIT_ARCH_EXT
+
check_ulp ();
/* Keep the tests a wee bit ordered (according to ISO C99). */
diff --git a/sysdeps/x86_64/fpu/math-tests.h b/sysdeps/x86_64/fpu/math-tests.h
new file mode 100644
index 0000000..466b97b
--- /dev/null
+++ b/sysdeps/x86_64/fpu/math-tests.h
@@ -0,0 +1,34 @@
+/* Configuration for math tests. x86_64 version.
+ Copyright (C) 2013-2014 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/>. */
+
+#ifdef REQUIRE_AVX2
+# include <init-arch.h>
+
+ static int avx2_usable; /* Set to 1 if AVX2 supported */
+
+# define INIT_ARCH_EXT \
+ __init_cpu_features (); \
+ avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
+ & bit_AVX2_Usable;
+
+# define CHECK_ARCH_EXT \
+ if (!avx2_usable) return;
+
+#endif
+
+#include_next <math-tests.h>
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-11-20 17:40 [PATCH 2/N] [x86_64] Vectorized math functions Andrew Senkevich
@ 2014-11-21 17:55 ` Joseph Myers
2014-11-24 18:05 ` Andrew Senkevich
0 siblings, 1 reply; 13+ messages in thread
From: Joseph Myers @ 2014-11-21 17:55 UTC (permalink / raw)
To: Andrew Senkevich; +Cc: libc-alpha
On Thu, 20 Nov 2014, Andrew Senkevich wrote:
> +#ifndef TEST_MATHVEC
> +# define TEST_MATHVEC 0
> +#endif
Making this value explicit in each relevant test would better accord with
current glibc practice.
> +#ifndef INIT_ARCH_EXT
> +# define INIT_ARCH_EXT
> +# define CHECK_ARCH_EXT
> +#endif
Again it's worth considering how to avoid this pattern. Maybe a separate
math-tests-arch.h header, with the sysdeps/generic version having a
comment explaining the semantics of the macros and defining them to empty
(and the x86_64 version including the generic one when not doing vector
tests)?
> +#ifndef VEC_PREFIX
> +# define VEC_PREFIX
> +#endif
So far we have two ways of distinguishing (in libm-test-ulps) tests for
variants of functions: suffixes (for rounding modes) and type names such
as "double" and "ildoubl" (for the type and for whether the inline
variants are being tested). I don't think adding a third way is a good
idea. I suggest using suffixes (until we can unify things and get down to
a single way).
> +# define INIT_ARCH_EXT \
> + __init_cpu_features (); \
> + avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
> + & bit_AVX2_Usable;
> +
> +# define CHECK_ARCH_EXT \
> + if (!avx2_usable) return;
Any macro expanding to statements should expand to a
do
{
...;
}
while (0)
block unless what it does (e.g. declaring variables to be used outside its
own expansion) is unsuitable for such a block.
--
Joseph S. Myers
joseph@codesourcery.com
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-11-21 17:55 ` Joseph Myers
@ 2014-11-24 18:05 ` Andrew Senkevich
2014-11-26 13:12 ` Joseph Myers
0 siblings, 1 reply; 13+ messages in thread
From: Andrew Senkevich @ 2014-11-24 18:05 UTC (permalink / raw)
To: Joseph Myers; +Cc: libc-alpha
[-- Attachment #1: Type: text/plain, Size: 8840 bytes --]
2014-11-21 20:55 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> On Thu, 20 Nov 2014, Andrew Senkevich wrote:
>
>> +#ifndef TEST_MATHVEC
>> +# define TEST_MATHVEC 0
>> +#endif
>
> Making this value explicit in each relevant test would better accord with
> current glibc practice.
>
>> +#ifndef INIT_ARCH_EXT
>> +# define INIT_ARCH_EXT
>> +# define CHECK_ARCH_EXT
>> +#endif
>
> Again it's worth considering how to avoid this pattern. Maybe a separate
> math-tests-arch.h header, with the sysdeps/generic version having a
> comment explaining the semantics of the macros and defining them to empty
> (and the x86_64 version including the generic one when not doing vector
> tests)?
>
>> +#ifndef VEC_PREFIX
>> +# define VEC_PREFIX
>> +#endif
>
> So far we have two ways of distinguishing (in libm-test-ulps) tests for
> variants of functions: suffixes (for rounding modes) and type names such
> as "double" and "ildoubl" (for the type and for whether the inline
> variants are being tested). I don't think adding a third way is a good
> idea. I suggest using suffixes (until we can unify things and get down to
> a single way).
>
>> +# define INIT_ARCH_EXT \
>> + __init_cpu_features (); \
>> + avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
>> + & bit_AVX2_Usable;
>> +
>> +# define CHECK_ARCH_EXT \
>> + if (!avx2_usable) return;
>
> Any macro expanding to statements should expand to a
>
> do
> {
> ...;
> }
> while (0)
>
> block unless what it does (e.g. declaring variables to be used outside its
> own expansion) is unsuitable for such a block.
Fixed.
diff --git a/math/libm-test.inc b/math/libm-test.inc
index 5d2ed1a..4958385 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -126,6 +126,7 @@
#include <argp.h>
#include <tininess.h>
#include <math-tests.h>
+#include <math-tests-arch.h>
/* Structure for ulp data for a function, or the real or imaginary
part of a function. */
@@ -684,7 +685,7 @@ static void
test_single_errno (const char *test_name, int errno_value,
int expected_value, const char *expected_name)
{
-#ifndef TEST_INLINE
+#if !defined TEST_INLINE && !TEST_MATHVEC
if (errno_value == expected_value)
{
if (print_screen (1))
@@ -1690,9 +1691,18 @@ struct test_fFF_11_data
(ARRAY)[i].RM_##ROUNDING_MODE.extra2_expected); \
ROUND_RESTORE_ ## ROUNDING_MODE
+#if !TEST_MATHVEC
+# define VEC_SUFF
+#endif
+
+#define STR_CONCAT(a, b, c) __STRING (a##b##c)
+#define STR_CON3(a, b, c) STR_CONCAT (a, b, c)
+
/* Start and end the tests for a given function. */
-#define START(FUNC, EXACT) \
- const char *this_func = #FUNC; \
+#define START(FUNC, SUFF, EXACT) \
+ CHECK_ARCH_EXT \
+ if (TEST_MATHVEC) return; \
+ const char *this_func = STR_CON3 (FUNC, SUFF, VEC_SUFF); \
init_max_error (this_func, EXACT)
#define END \
print_max_error (this_func)
@@ -1705,28 +1715,28 @@ struct test_fFF_11_data
{ \
do \
{ \
- START (FUNC, EXACT); \
+ START (FUNC,, EXACT); \
LOOP_MACRO (FUNC, ARRAY, , ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _downward, EXACT); \
+ START (FUNC, _downward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_DOWNWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _towardzero, EXACT); \
+ START (FUNC, _towardzero, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_TOWARDZERO, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _upward, EXACT); \
+ START (FUNC, _upward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_UPWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
@@ -6034,7 +6044,7 @@ static const struct test_c_c_data cexp_test_data[] =
static void
cexp_test (void)
{
- START (cexp, 0);
+ START (cexp,, 0);
RUN_TEST_LOOP_c_c (cexp, cexp_test_data, );
END_COMPLEX;
}
@@ -7548,7 +7558,7 @@ static const struct test_if_f_data jn_test_data[] =
static void
jn_test (void)
{
- START (jn, 0);
+ START (jn,, 0);
RUN_TEST_LOOP_if_f (jn, jn_test_data, );
END;
}
@@ -9374,7 +9384,7 @@ static const struct test_f_f_data tgamma_test_data[] =
static void
tgamma_test (void)
{
- START (tgamma, 0);
+ START (tgamma,, 0);
RUN_TEST_LOOP_f_f (tgamma, tgamma_test_data, );
END;
}
diff --git a/math/test-double.c b/math/test-double.c
index e666197..4aec418 100644
--- a/math/test-double.c
+++ b/math/test-double.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-float.c b/math/test-float.c
index cb68a5c..f646f98 100644
--- a/math/test-float.c
+++ b/math/test-float.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-idouble.c b/math/test-idouble.c
index 665f78c..9131055 100644
--- a/math/test-idouble.c
+++ b/math/test-idouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#define TEST_INLINE
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ifloat.c b/math/test-ifloat.c
index 41255e9..61a5684 100644
--- a/math/test-ifloat.c
+++ b/math/test-ifloat.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#define TEST_INLINE 1
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ildoubl.c b/math/test-ildoubl.c
index 69ad35e..b8f57a7 100644
--- a/math/test-ildoubl.c
+++ b/math/test-ildoubl.c
@@ -27,6 +27,7 @@
#define PRINTF_NEXPR "Lf"
#define TEST_INLINE
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifdef __NO_MATH_INLINES
# undef __NO_MATH_INLINES
diff --git a/math/test-ldouble.c b/math/test-ldouble.c
index 07e4601..4253043 100644
--- a/math/test-ldouble.c
+++ b/math/test-ldouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "La"
#define PRINTF_NEXPR "Lf"
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/sysdeps/generic/math-tests-arch.h
b/sysdeps/generic/math-tests-arch.h
new file mode 100644
index 0000000..c005fad
--- /dev/null
+++ b/sysdeps/generic/math-tests-arch.h
@@ -0,0 +1,21 @@
+/* Runtime architecture check for math tests.
+ Copyright (C) 2014 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/>. */
+
+/* Macros used for architecture check in math tests runtime. */
+#define INIT_ARCH_EXT
+#define CHECK_ARCH_EXT
diff --git a/sysdeps/x86_64/fpu/math-tests-arch.h
b/sysdeps/x86_64/fpu/math-tests-arch.h
new file mode 100644
index 0000000..dfd8221
--- /dev/null
+++ b/sysdeps/x86_64/fpu/math-tests-arch.h
@@ -0,0 +1,43 @@
+/* Runtime architecture check for math tests. x86_64 version.
+ Copyright (C) 2013-2014 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/>. */
+
+#ifdef REQUIRE_AVX2
+# include <init-arch.h>
+
+ /* Set to 1 if AVX2 supported. */
+ static int avx2_usable;
+
+# define INIT_ARCH_EXT \
+ do \
+ { \
+ __init_cpu_features (); \
+ avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
+ & bit_AVX2_Usable; \
+ } \
+ while(0);
+
+# define CHECK_ARCH_EXT \
+ do \
+ { \
+ if (!avx2_usable) return; \
+ } \
+ while(0);
+
+#else
+# include <sysdeps/generic/math-tests-arch.h>
+#endif
--
WBR,
Andrew
[-- Attachment #2: libmvec_2_of_N_v2.patch --]
[-- Type: application/octet-stream, Size: 7501 bytes --]
diff --git a/math/libm-test.inc b/math/libm-test.inc
index 5d2ed1a..4958385 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -126,6 +126,7 @@
#include <argp.h>
#include <tininess.h>
#include <math-tests.h>
+#include <math-tests-arch.h>
/* Structure for ulp data for a function, or the real or imaginary
part of a function. */
@@ -684,7 +685,7 @@ static void
test_single_errno (const char *test_name, int errno_value,
int expected_value, const char *expected_name)
{
-#ifndef TEST_INLINE
+#if !defined TEST_INLINE && !TEST_MATHVEC
if (errno_value == expected_value)
{
if (print_screen (1))
@@ -1690,9 +1691,18 @@ struct test_fFF_11_data
(ARRAY)[i].RM_##ROUNDING_MODE.extra2_expected); \
ROUND_RESTORE_ ## ROUNDING_MODE
+#if !TEST_MATHVEC
+# define VEC_SUFF
+#endif
+
+#define STR_CONCAT(a, b, c) __STRING (a##b##c)
+#define STR_CON3(a, b, c) STR_CONCAT (a, b, c)
+
/* Start and end the tests for a given function. */
-#define START(FUNC, EXACT) \
- const char *this_func = #FUNC; \
+#define START(FUNC, SUFF, EXACT) \
+ CHECK_ARCH_EXT \
+ if (TEST_MATHVEC) return; \
+ const char *this_func = STR_CON3 (FUNC, SUFF, VEC_SUFF); \
init_max_error (this_func, EXACT)
#define END \
print_max_error (this_func)
@@ -1705,28 +1715,28 @@ struct test_fFF_11_data
{ \
do \
{ \
- START (FUNC, EXACT); \
+ START (FUNC,, EXACT); \
LOOP_MACRO (FUNC, ARRAY, , ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _downward, EXACT); \
+ START (FUNC, _downward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_DOWNWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _towardzero, EXACT); \
+ START (FUNC, _towardzero, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_TOWARDZERO, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _upward, EXACT); \
+ START (FUNC, _upward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_UPWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
@@ -6034,7 +6044,7 @@ static const struct test_c_c_data cexp_test_data[] =
static void
cexp_test (void)
{
- START (cexp, 0);
+ START (cexp,, 0);
RUN_TEST_LOOP_c_c (cexp, cexp_test_data, );
END_COMPLEX;
}
@@ -7548,7 +7558,7 @@ static const struct test_if_f_data jn_test_data[] =
static void
jn_test (void)
{
- START (jn, 0);
+ START (jn,, 0);
RUN_TEST_LOOP_if_f (jn, jn_test_data, );
END;
}
@@ -9374,7 +9384,7 @@ static const struct test_f_f_data tgamma_test_data[] =
static void
tgamma_test (void)
{
- START (tgamma, 0);
+ START (tgamma,, 0);
RUN_TEST_LOOP_f_f (tgamma, tgamma_test_data, );
END;
}
diff --git a/math/test-double.c b/math/test-double.c
index e666197..4aec418 100644
--- a/math/test-double.c
+++ b/math/test-double.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-float.c b/math/test-float.c
index cb68a5c..f646f98 100644
--- a/math/test-float.c
+++ b/math/test-float.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-idouble.c b/math/test-idouble.c
index 665f78c..9131055 100644
--- a/math/test-idouble.c
+++ b/math/test-idouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#define TEST_INLINE
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ifloat.c b/math/test-ifloat.c
index 41255e9..61a5684 100644
--- a/math/test-ifloat.c
+++ b/math/test-ifloat.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#define TEST_INLINE 1
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ildoubl.c b/math/test-ildoubl.c
index 69ad35e..b8f57a7 100644
--- a/math/test-ildoubl.c
+++ b/math/test-ildoubl.c
@@ -27,6 +27,7 @@
#define PRINTF_NEXPR "Lf"
#define TEST_INLINE
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifdef __NO_MATH_INLINES
# undef __NO_MATH_INLINES
diff --git a/math/test-ldouble.c b/math/test-ldouble.c
index 07e4601..4253043 100644
--- a/math/test-ldouble.c
+++ b/math/test-ldouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "La"
#define PRINTF_NEXPR "Lf"
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/sysdeps/generic/math-tests-arch.h b/sysdeps/generic/math-tests-arch.h
new file mode 100644
index 0000000..c005fad
--- /dev/null
+++ b/sysdeps/generic/math-tests-arch.h
@@ -0,0 +1,21 @@
+/* Runtime architecture check for math tests.
+ Copyright (C) 2014 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/>. */
+
+/* Macros used for architecture check in math tests runtime. */
+#define INIT_ARCH_EXT
+#define CHECK_ARCH_EXT
diff --git a/sysdeps/x86_64/fpu/math-tests-arch.h b/sysdeps/x86_64/fpu/math-tests-arch.h
new file mode 100644
index 0000000..dfd8221
--- /dev/null
+++ b/sysdeps/x86_64/fpu/math-tests-arch.h
@@ -0,0 +1,43 @@
+/* Runtime architecture check for math tests. x86_64 version.
+ Copyright (C) 2013-2014 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/>. */
+
+#ifdef REQUIRE_AVX2
+# include <init-arch.h>
+
+ /* Set to 1 if AVX2 supported. */
+ static int avx2_usable;
+
+# define INIT_ARCH_EXT \
+ do \
+ { \
+ __init_cpu_features (); \
+ avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
+ & bit_AVX2_Usable; \
+ } \
+ while(0);
+
+# define CHECK_ARCH_EXT \
+ do \
+ { \
+ if (!avx2_usable) return; \
+ } \
+ while(0);
+
+#else
+# include <sysdeps/generic/math-tests-arch.h>
+#endif
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-11-24 18:05 ` Andrew Senkevich
@ 2014-11-26 13:12 ` Joseph Myers
2014-11-26 16:03 ` Andrew Senkevich
0 siblings, 1 reply; 13+ messages in thread
From: Joseph Myers @ 2014-11-26 13:12 UTC (permalink / raw)
To: Andrew Senkevich; +Cc: libc-alpha
On Mon, 24 Nov 2014, Andrew Senkevich wrote:
> +#define START(FUNC, SUFF, EXACT) \
> + CHECK_ARCH_EXT \
> + if (TEST_MATHVEC) return; \
With a proper definition of CHECK_ARCH_EXT (one that always acts like a
statement without the trailing ';'), the call should be followed by ';'.
> +# define INIT_ARCH_EXT \
> + do \
> + { \
> + __init_cpu_features (); \
> + avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
> + & bit_AVX2_Usable; \
> + } \
> + while(0);
> +
> +# define CHECK_ARCH_EXT \
> + do \
> + { \
> + if (!avx2_usable) return; \
> + } \
> + while(0);
And the definitions here should not have the trailing semicolon, because
that will be provided by the caller. Also note there should be a space
between "while" and "(0)".
--
Joseph S. Myers
joseph@codesourcery.com
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-11-26 13:12 ` Joseph Myers
@ 2014-11-26 16:03 ` Andrew Senkevich
2014-11-26 16:26 ` Joseph Myers
0 siblings, 1 reply; 13+ messages in thread
From: Andrew Senkevich @ 2014-11-26 16:03 UTC (permalink / raw)
To: Joseph Myers; +Cc: libc-alpha
[-- Attachment #1: Type: text/plain, Size: 9129 bytes --]
2014-11-26 16:12 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> On Mon, 24 Nov 2014, Andrew Senkevich wrote:
>
>> +#define START(FUNC, SUFF, EXACT) \
>> + CHECK_ARCH_EXT \
>> + if (TEST_MATHVEC) return; \
>
> With a proper definition of CHECK_ARCH_EXT (one that always acts like a
> statement without the trailing ';'), the call should be followed by ';'.
>
>> +# define INIT_ARCH_EXT \
>> + do \
>> + { \
>> + __init_cpu_features (); \
>> + avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
>> + & bit_AVX2_Usable; \
>> + } \
>> + while(0);
>> +
>> +# define CHECK_ARCH_EXT \
>> + do \
>> + { \
>> + if (!avx2_usable) return; \
>> + } \
>> + while(0);
>
> And the definitions here should not have the trailing semicolon, because
> that will be provided by the caller. Also note there should be a space
> between "while" and "(0)".
Fixed, and also added call of INIT_ARCH_EXT in libm-test.c (will be
removed from patch 6).
ChangeLog
2014-11-26 Andrew Senkevich <andrew.n.senkevich@gmail.com>
* sysdeps/generic/math-tests-arch.h (INIT_ARCH_EXT, CHECK_ARCH_EXT):
New empty defined macros.
* sysdeps/x86_64/fpu/math-tests-arch.h (INIT_ARCH_EXT,CHECK_ARCH_EXT):
Platform-specific definitions.
* math/libm-test.inc (TEST_MATHVEC, VEC_SUFF, INIT_ARCH_EXT,
STR_CONCAT, STR_CON3): New macros. START refactored.
* math/test-double.c: Added TEST_MATHVEC definition.
* math/test-float.c: Likewise.
* math/test-idouble.c: Likewise.
* math/test-ifloat.c: Likewise.
* math/test-ildoubl.c: Likewise.
* math/test-ldouble.c: Likewise.
diff --git a/math/libm-test.inc b/math/libm-test.inc
index 5d2ed1a..bcedc1e 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -126,6 +126,7 @@
#include <argp.h>
#include <tininess.h>
#include <math-tests.h>
+#include <math-tests-arch.h>
/* Structure for ulp data for a function, or the real or imaginary
part of a function. */
@@ -684,7 +685,7 @@ static void
test_single_errno (const char *test_name, int errno_value,
int expected_value, const char *expected_name)
{
-#ifndef TEST_INLINE
+#if !defined TEST_INLINE && !TEST_MATHVEC
if (errno_value == expected_value)
{
if (print_screen (1))
@@ -1690,9 +1691,18 @@ struct test_fFF_11_data
(ARRAY)[i].RM_##ROUNDING_MODE.extra2_expected); \
ROUND_RESTORE_ ## ROUNDING_MODE
+#if !TEST_MATHVEC
+# define VEC_SUFF
+#endif
+
+#define STR_CONCAT(a, b, c) __STRING (a##b##c)
+#define STR_CON3(a, b, c) STR_CONCAT (a, b, c)
+
/* Start and end the tests for a given function. */
-#define START(FUNC, EXACT) \
- const char *this_func = #FUNC; \
+#define START(FUNC, SUFF, EXACT) \
+ CHECK_ARCH_EXT; \
+ if (TEST_MATHVEC) return; \
+ const char *this_func = STR_CON3 (FUNC, SUFF, VEC_SUFF); \
init_max_error (this_func, EXACT)
#define END \
print_max_error (this_func)
@@ -1705,28 +1715,28 @@ struct test_fFF_11_data
{ \
do \
{ \
- START (FUNC, EXACT); \
+ START (FUNC,, EXACT); \
LOOP_MACRO (FUNC, ARRAY, , ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _downward, EXACT); \
+ START (FUNC, _downward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_DOWNWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _towardzero, EXACT); \
+ START (FUNC, _towardzero, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_TOWARDZERO, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _upward, EXACT); \
+ START (FUNC, _upward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_UPWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
@@ -6034,7 +6044,7 @@ static const struct test_c_c_data cexp_test_data[] =
static void
cexp_test (void)
{
- START (cexp, 0);
+ START (cexp,, 0);
RUN_TEST_LOOP_c_c (cexp, cexp_test_data, );
END_COMPLEX;
}
@@ -7548,7 +7558,7 @@ static const struct test_if_f_data jn_test_data[] =
static void
jn_test (void)
{
- START (jn, 0);
+ START (jn,, 0);
RUN_TEST_LOOP_if_f (jn, jn_test_data, );
END;
}
@@ -9374,7 +9384,7 @@ static const struct test_f_f_data tgamma_test_data[] =
static void
tgamma_test (void)
{
- START (tgamma, 0);
+ START (tgamma,, 0);
RUN_TEST_LOOP_f_f (tgamma, tgamma_test_data, );
END;
}
@@ -9824,6 +9834,8 @@ main (int argc, char **argv)
initialize ();
printf (TEST_MSG);
+ INIT_ARCH_EXT;
+
check_ulp ();
/* Keep the tests a wee bit ordered (according to ISO C99). */
diff --git a/math/test-double.c b/math/test-double.c
index e666197..4aec418 100644
--- a/math/test-double.c
+++ b/math/test-double.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-float.c b/math/test-float.c
index cb68a5c..f646f98 100644
--- a/math/test-float.c
+++ b/math/test-float.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-idouble.c b/math/test-idouble.c
index 665f78c..9131055 100644
--- a/math/test-idouble.c
+++ b/math/test-idouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#define TEST_INLINE
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ifloat.c b/math/test-ifloat.c
index 41255e9..61a5684 100644
--- a/math/test-ifloat.c
+++ b/math/test-ifloat.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#define TEST_INLINE 1
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ildoubl.c b/math/test-ildoubl.c
index 69ad35e..b8f57a7 100644
--- a/math/test-ildoubl.c
+++ b/math/test-ildoubl.c
@@ -27,6 +27,7 @@
#define PRINTF_NEXPR "Lf"
#define TEST_INLINE
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifdef __NO_MATH_INLINES
# undef __NO_MATH_INLINES
diff --git a/math/test-ldouble.c b/math/test-ldouble.c
index 07e4601..4253043 100644
--- a/math/test-ldouble.c
+++ b/math/test-ldouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "La"
#define PRINTF_NEXPR "Lf"
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/sysdeps/generic/math-tests-arch.h
b/sysdeps/generic/math-tests-arch.h
new file mode 100644
index 0000000..c005fad
--- /dev/null
+++ b/sysdeps/generic/math-tests-arch.h
@@ -0,0 +1,21 @@
+/* Runtime architecture check for math tests.
+ Copyright (C) 2014 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/>. */
+
+/* Macros used for architecture check in math tests runtime. */
+#define INIT_ARCH_EXT
+#define CHECK_ARCH_EXT
diff --git a/sysdeps/x86_64/fpu/math-tests-arch.h
b/sysdeps/x86_64/fpu/math-tests-arch.h
new file mode 100644
index 0000000..4c2f372
--- /dev/null
+++ b/sysdeps/x86_64/fpu/math-tests-arch.h
@@ -0,0 +1,43 @@
+/* Runtime architecture check for math tests. x86_64 version.
+ Copyright (C) 2013-2014 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/>. */
+
+#ifdef REQUIRE_AVX2
+# include <init-arch.h>
+
+ /* Set to 1 if AVX2 supported. */
+ static int avx2_usable;
+
+# define INIT_ARCH_EXT \
+ do \
+ { \
+ __init_cpu_features (); \
+ avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
+ & bit_AVX2_Usable; \
+ } \
+ while (0)
+
+# define CHECK_ARCH_EXT \
+ do \
+ { \
+ if (!avx2_usable) return; \
+ } \
+ while (0)
+
+#else
+# include <sysdeps/generic/math-tests-arch.h>
+#endif
Is it OK for commit?
--
WBR,
Andrew
[-- Attachment #2: libmvec_2_of_N_v3.patch --]
[-- Type: application/octet-stream, Size: 7701 bytes --]
diff --git a/math/libm-test.inc b/math/libm-test.inc
index 5d2ed1a..bcedc1e 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -126,6 +126,7 @@
#include <argp.h>
#include <tininess.h>
#include <math-tests.h>
+#include <math-tests-arch.h>
/* Structure for ulp data for a function, or the real or imaginary
part of a function. */
@@ -684,7 +685,7 @@ static void
test_single_errno (const char *test_name, int errno_value,
int expected_value, const char *expected_name)
{
-#ifndef TEST_INLINE
+#if !defined TEST_INLINE && !TEST_MATHVEC
if (errno_value == expected_value)
{
if (print_screen (1))
@@ -1690,9 +1691,18 @@ struct test_fFF_11_data
(ARRAY)[i].RM_##ROUNDING_MODE.extra2_expected); \
ROUND_RESTORE_ ## ROUNDING_MODE
+#if !TEST_MATHVEC
+# define VEC_SUFF
+#endif
+
+#define STR_CONCAT(a, b, c) __STRING (a##b##c)
+#define STR_CON3(a, b, c) STR_CONCAT (a, b, c)
+
/* Start and end the tests for a given function. */
-#define START(FUNC, EXACT) \
- const char *this_func = #FUNC; \
+#define START(FUNC, SUFF, EXACT) \
+ CHECK_ARCH_EXT; \
+ if (TEST_MATHVEC) return; \
+ const char *this_func = STR_CON3 (FUNC, SUFF, VEC_SUFF); \
init_max_error (this_func, EXACT)
#define END \
print_max_error (this_func)
@@ -1705,28 +1715,28 @@ struct test_fFF_11_data
{ \
do \
{ \
- START (FUNC, EXACT); \
+ START (FUNC,, EXACT); \
LOOP_MACRO (FUNC, ARRAY, , ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _downward, EXACT); \
+ START (FUNC, _downward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_DOWNWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _towardzero, EXACT); \
+ START (FUNC, _towardzero, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_TOWARDZERO, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _upward, EXACT); \
+ START (FUNC, _upward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_UPWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
@@ -6034,7 +6044,7 @@ static const struct test_c_c_data cexp_test_data[] =
static void
cexp_test (void)
{
- START (cexp, 0);
+ START (cexp,, 0);
RUN_TEST_LOOP_c_c (cexp, cexp_test_data, );
END_COMPLEX;
}
@@ -7548,7 +7558,7 @@ static const struct test_if_f_data jn_test_data[] =
static void
jn_test (void)
{
- START (jn, 0);
+ START (jn,, 0);
RUN_TEST_LOOP_if_f (jn, jn_test_data, );
END;
}
@@ -9374,7 +9384,7 @@ static const struct test_f_f_data tgamma_test_data[] =
static void
tgamma_test (void)
{
- START (tgamma, 0);
+ START (tgamma,, 0);
RUN_TEST_LOOP_f_f (tgamma, tgamma_test_data, );
END;
}
@@ -9824,6 +9834,8 @@ main (int argc, char **argv)
initialize ();
printf (TEST_MSG);
+ INIT_ARCH_EXT;
+
check_ulp ();
/* Keep the tests a wee bit ordered (according to ISO C99). */
diff --git a/math/test-double.c b/math/test-double.c
index e666197..4aec418 100644
--- a/math/test-double.c
+++ b/math/test-double.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-float.c b/math/test-float.c
index cb68a5c..f646f98 100644
--- a/math/test-float.c
+++ b/math/test-float.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-idouble.c b/math/test-idouble.c
index 665f78c..9131055 100644
--- a/math/test-idouble.c
+++ b/math/test-idouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#define TEST_INLINE
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ifloat.c b/math/test-ifloat.c
index 41255e9..61a5684 100644
--- a/math/test-ifloat.c
+++ b/math/test-ifloat.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#define TEST_INLINE 1
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ildoubl.c b/math/test-ildoubl.c
index 69ad35e..b8f57a7 100644
--- a/math/test-ildoubl.c
+++ b/math/test-ildoubl.c
@@ -27,6 +27,7 @@
#define PRINTF_NEXPR "Lf"
#define TEST_INLINE
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifdef __NO_MATH_INLINES
# undef __NO_MATH_INLINES
diff --git a/math/test-ldouble.c b/math/test-ldouble.c
index 07e4601..4253043 100644
--- a/math/test-ldouble.c
+++ b/math/test-ldouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "La"
#define PRINTF_NEXPR "Lf"
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/sysdeps/generic/math-tests-arch.h b/sysdeps/generic/math-tests-arch.h
new file mode 100644
index 0000000..c005fad
--- /dev/null
+++ b/sysdeps/generic/math-tests-arch.h
@@ -0,0 +1,21 @@
+/* Runtime architecture check for math tests.
+ Copyright (C) 2014 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/>. */
+
+/* Macros used for architecture check in math tests runtime. */
+#define INIT_ARCH_EXT
+#define CHECK_ARCH_EXT
diff --git a/sysdeps/x86_64/fpu/math-tests-arch.h b/sysdeps/x86_64/fpu/math-tests-arch.h
new file mode 100644
index 0000000..4c2f372
--- /dev/null
+++ b/sysdeps/x86_64/fpu/math-tests-arch.h
@@ -0,0 +1,43 @@
+/* Runtime architecture check for math tests. x86_64 version.
+ Copyright (C) 2013-2014 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/>. */
+
+#ifdef REQUIRE_AVX2
+# include <init-arch.h>
+
+ /* Set to 1 if AVX2 supported. */
+ static int avx2_usable;
+
+# define INIT_ARCH_EXT \
+ do \
+ { \
+ __init_cpu_features (); \
+ avx2_usable = __cpu_features.feature[index_AVX2_Usable] \
+ & bit_AVX2_Usable; \
+ } \
+ while (0)
+
+# define CHECK_ARCH_EXT \
+ do \
+ { \
+ if (!avx2_usable) return; \
+ } \
+ while (0)
+
+#else
+# include <sysdeps/generic/math-tests-arch.h>
+#endif
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-11-26 16:03 ` Andrew Senkevich
@ 2014-11-26 16:26 ` Joseph Myers
2014-12-02 16:09 ` Andrew Senkevich
0 siblings, 1 reply; 13+ messages in thread
From: Joseph Myers @ 2014-11-26 16:26 UTC (permalink / raw)
To: Andrew Senkevich; +Cc: libc-alpha
On Wed, 26 Nov 2014, Andrew Senkevich wrote:
> +/* Macros used for architecture check in math tests runtime. */
> +#define INIT_ARCH_EXT
> +#define CHECK_ARCH_EXT
I think the comments in the generic file need to say more about the
semantics of these macros (in particular, that CHECK_ARCH_EXT should
return if it determines, possibly using data set up by INIT_ARCH_EXT, that
the test is running on hardware not supported by the function
implementations to be tested).
--
Joseph S. Myers
joseph@codesourcery.com
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-11-26 16:26 ` Joseph Myers
@ 2014-12-02 16:09 ` Andrew Senkevich
2014-12-02 22:41 ` Joseph Myers
0 siblings, 1 reply; 13+ messages in thread
From: Andrew Senkevich @ 2014-12-02 16:09 UTC (permalink / raw)
To: Joseph Myers; +Cc: libc-alpha
[-- Attachment #1: Type: text/plain, Size: 7170 bytes --]
2014-11-26 19:26 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> On Wed, 26 Nov 2014, Andrew Senkevich wrote:
>
>> +/* Macros used for architecture check in math tests runtime. */
>> +#define INIT_ARCH_EXT
>> +#define CHECK_ARCH_EXT
>
> I think the comments in the generic file need to say more about the
> semantics of these macros (in particular, that CHECK_ARCH_EXT should
> return if it determines, possibly using data set up by INIT_ARCH_EXT, that
> the test is running on hardware not supported by the function
> implementations to be tested).
Comments added.
ChangeLog
* math/libm-test.inc: START refactored.
* math/test-double.c (TEST_MATHVEC): Add define.
* math/test-float.c: Likewise.
* math/test-idouble.c: Likewise.
* math/test-ifloat.c: Likewise.
* math/test-ildoubl.c: Likewise.
* math/test-ldouble.c: Likewise.
* sysdeps/generic/math-tests-arch.h (INIT_ARCH_EXT, CHECK_ARCH_EXT):
New helper macros for runtime architecture check.
diff --git a/math/libm-test.inc b/math/libm-test.inc
index 5d2ed1a..bcedc1e 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -126,6 +126,7 @@
#include <argp.h>
#include <tininess.h>
#include <math-tests.h>
+#include <math-tests-arch.h>
/* Structure for ulp data for a function, or the real or imaginary
part of a function. */
@@ -684,7 +685,7 @@ static void
test_single_errno (const char *test_name, int errno_value,
int expected_value, const char *expected_name)
{
-#ifndef TEST_INLINE
+#if !defined TEST_INLINE && !TEST_MATHVEC
if (errno_value == expected_value)
{
if (print_screen (1))
@@ -1690,9 +1691,18 @@ struct test_fFF_11_data
(ARRAY)[i].RM_##ROUNDING_MODE.extra2_expected); \
ROUND_RESTORE_ ## ROUNDING_MODE
+#if !TEST_MATHVEC
+# define VEC_SUFF
+#endif
+
+#define STR_CONCAT(a, b, c) __STRING (a##b##c)
+#define STR_CON3(a, b, c) STR_CONCAT (a, b, c)
+
/* Start and end the tests for a given function. */
-#define START(FUNC, EXACT) \
- const char *this_func = #FUNC; \
+#define START(FUNC, SUFF, EXACT) \
+ CHECK_ARCH_EXT; \
+ if (TEST_MATHVEC) return; \
+ const char *this_func = STR_CON3 (FUNC, SUFF, VEC_SUFF); \
init_max_error (this_func, EXACT)
#define END \
print_max_error (this_func)
@@ -1705,28 +1715,28 @@ struct test_fFF_11_data
{ \
do \
{ \
- START (FUNC, EXACT); \
+ START (FUNC,, EXACT); \
LOOP_MACRO (FUNC, ARRAY, , ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _downward, EXACT); \
+ START (FUNC, _downward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_DOWNWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _towardzero, EXACT); \
+ START (FUNC, _towardzero, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_TOWARDZERO, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _upward, EXACT); \
+ START (FUNC, _upward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_UPWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
@@ -6034,7 +6044,7 @@ static const struct test_c_c_data cexp_test_data[] =
static void
cexp_test (void)
{
- START (cexp, 0);
+ START (cexp,, 0);
RUN_TEST_LOOP_c_c (cexp, cexp_test_data, );
END_COMPLEX;
}
@@ -7548,7 +7558,7 @@ static const struct test_if_f_data jn_test_data[] =
static void
jn_test (void)
{
- START (jn, 0);
+ START (jn,, 0);
RUN_TEST_LOOP_if_f (jn, jn_test_data, );
END;
}
@@ -9374,7 +9384,7 @@ static const struct test_f_f_data tgamma_test_data[] =
static void
tgamma_test (void)
{
- START (tgamma, 0);
+ START (tgamma,, 0);
RUN_TEST_LOOP_f_f (tgamma, tgamma_test_data, );
END;
}
@@ -9824,6 +9834,8 @@ main (int argc, char **argv)
initialize ();
printf (TEST_MSG);
+ INIT_ARCH_EXT;
+
check_ulp ();
/* Keep the tests a wee bit ordered (according to ISO C99). */
diff --git a/math/test-double.c b/math/test-double.c
index e666197..4aec418 100644
--- a/math/test-double.c
+++ b/math/test-double.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-float.c b/math/test-float.c
index cb68a5c..f646f98 100644
--- a/math/test-float.c
+++ b/math/test-float.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-idouble.c b/math/test-idouble.c
index 665f78c..9131055 100644
--- a/math/test-idouble.c
+++ b/math/test-idouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#define TEST_INLINE
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ifloat.c b/math/test-ifloat.c
index 41255e9..61a5684 100644
--- a/math/test-ifloat.c
+++ b/math/test-ifloat.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#define TEST_INLINE 1
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ildoubl.c b/math/test-ildoubl.c
index 69ad35e..b8f57a7 100644
--- a/math/test-ildoubl.c
+++ b/math/test-ildoubl.c
@@ -27,6 +27,7 @@
#define PRINTF_NEXPR "Lf"
#define TEST_INLINE
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifdef __NO_MATH_INLINES
# undef __NO_MATH_INLINES
diff --git a/math/test-ldouble.c b/math/test-ldouble.c
index 07e4601..4253043 100644
--- a/math/test-ldouble.c
+++ b/math/test-ldouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "La"
#define PRINTF_NEXPR "Lf"
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/sysdeps/generic/math-tests-arch.h
b/sysdeps/generic/math-tests-arch.h
new file mode 100644
index 0000000..770ee27
--- /dev/null
+++ b/sysdeps/generic/math-tests-arch.h
@@ -0,0 +1,24 @@
+/* Runtime architecture check for math tests.
+ Copyright (C) 2014 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/>. */
+
+/* These macros used for architecture check in math tests runtime.
+ INIT_ARCH_EXT should set up for example some global variable which is
+ checked by CHECK_ARCH_EXT which produces return from individual test to
+ prevent run on hardware not supported by tested function implementation. */
+#define INIT_ARCH_EXT
+#define CHECK_ARCH_EXT
Is it OK for trunk?
--
WBR,
Andrew
[-- Attachment #2: libmvec_2_of_N_v4.patch --]
[-- Type: application/octet-stream, Size: 6320 bytes --]
diff --git a/math/libm-test.inc b/math/libm-test.inc
index 5d2ed1a..bcedc1e 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -126,6 +126,7 @@
#include <argp.h>
#include <tininess.h>
#include <math-tests.h>
+#include <math-tests-arch.h>
/* Structure for ulp data for a function, or the real or imaginary
part of a function. */
@@ -684,7 +685,7 @@ static void
test_single_errno (const char *test_name, int errno_value,
int expected_value, const char *expected_name)
{
-#ifndef TEST_INLINE
+#if !defined TEST_INLINE && !TEST_MATHVEC
if (errno_value == expected_value)
{
if (print_screen (1))
@@ -1690,9 +1691,18 @@ struct test_fFF_11_data
(ARRAY)[i].RM_##ROUNDING_MODE.extra2_expected); \
ROUND_RESTORE_ ## ROUNDING_MODE
+#if !TEST_MATHVEC
+# define VEC_SUFF
+#endif
+
+#define STR_CONCAT(a, b, c) __STRING (a##b##c)
+#define STR_CON3(a, b, c) STR_CONCAT (a, b, c)
+
/* Start and end the tests for a given function. */
-#define START(FUNC, EXACT) \
- const char *this_func = #FUNC; \
+#define START(FUNC, SUFF, EXACT) \
+ CHECK_ARCH_EXT; \
+ if (TEST_MATHVEC) return; \
+ const char *this_func = STR_CON3 (FUNC, SUFF, VEC_SUFF); \
init_max_error (this_func, EXACT)
#define END \
print_max_error (this_func)
@@ -1705,28 +1715,28 @@ struct test_fFF_11_data
{ \
do \
{ \
- START (FUNC, EXACT); \
+ START (FUNC,, EXACT); \
LOOP_MACRO (FUNC, ARRAY, , ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _downward, EXACT); \
+ START (FUNC, _downward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_DOWNWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _towardzero, EXACT); \
+ START (FUNC, _towardzero, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_TOWARDZERO, ## __VA_ARGS__); \
END_MACRO; \
} \
while (0); \
do \
{ \
- START (FUNC ## _upward, EXACT); \
+ START (FUNC, _upward, EXACT); \
LOOP_MACRO (FUNC, ARRAY, FE_UPWARD, ## __VA_ARGS__); \
END_MACRO; \
} \
@@ -6034,7 +6044,7 @@ static const struct test_c_c_data cexp_test_data[] =
static void
cexp_test (void)
{
- START (cexp, 0);
+ START (cexp,, 0);
RUN_TEST_LOOP_c_c (cexp, cexp_test_data, );
END_COMPLEX;
}
@@ -7548,7 +7558,7 @@ static const struct test_if_f_data jn_test_data[] =
static void
jn_test (void)
{
- START (jn, 0);
+ START (jn,, 0);
RUN_TEST_LOOP_if_f (jn, jn_test_data, );
END;
}
@@ -9374,7 +9384,7 @@ static const struct test_f_f_data tgamma_test_data[] =
static void
tgamma_test (void)
{
- START (tgamma, 0);
+ START (tgamma,, 0);
RUN_TEST_LOOP_f_f (tgamma, tgamma_test_data, );
END;
}
@@ -9824,6 +9834,8 @@ main (int argc, char **argv)
initialize ();
printf (TEST_MSG);
+ INIT_ARCH_EXT;
+
check_ulp ();
/* Keep the tests a wee bit ordered (according to ISO C99). */
diff --git a/math/test-double.c b/math/test-double.c
index e666197..4aec418 100644
--- a/math/test-double.c
+++ b/math/test-double.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-float.c b/math/test-float.c
index cb68a5c..f646f98 100644
--- a/math/test-float.c
+++ b/math/test-float.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/math/test-idouble.c b/math/test-idouble.c
index 665f78c..9131055 100644
--- a/math/test-idouble.c
+++ b/math/test-idouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_DOUBLE 1
+#define TEST_MATHVEC 0
#define TEST_INLINE
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ifloat.c b/math/test-ifloat.c
index 41255e9..61a5684 100644
--- a/math/test-ifloat.c
+++ b/math/test-ifloat.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "a"
#define PRINTF_NEXPR "f"
#define TEST_FLOAT 1
+#define TEST_MATHVEC 0
#define TEST_INLINE 1
#ifdef __NO_MATH_INLINES
diff --git a/math/test-ildoubl.c b/math/test-ildoubl.c
index 69ad35e..b8f57a7 100644
--- a/math/test-ildoubl.c
+++ b/math/test-ildoubl.c
@@ -27,6 +27,7 @@
#define PRINTF_NEXPR "Lf"
#define TEST_INLINE
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifdef __NO_MATH_INLINES
# undef __NO_MATH_INLINES
diff --git a/math/test-ldouble.c b/math/test-ldouble.c
index 07e4601..4253043 100644
--- a/math/test-ldouble.c
+++ b/math/test-ldouble.c
@@ -26,6 +26,7 @@
#define PRINTF_XEXPR "La"
#define PRINTF_NEXPR "Lf"
#define TEST_LDOUBLE 1
+#define TEST_MATHVEC 0
#ifndef __NO_MATH_INLINES
# define __NO_MATH_INLINES
diff --git a/sysdeps/generic/math-tests-arch.h b/sysdeps/generic/math-tests-arch.h
new file mode 100644
index 0000000..770ee27
--- /dev/null
+++ b/sysdeps/generic/math-tests-arch.h
@@ -0,0 +1,24 @@
+/* Runtime architecture check for math tests.
+ Copyright (C) 2014 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/>. */
+
+/* These macros used for architecture check in math tests runtime.
+ INIT_ARCH_EXT should set up for example some global variable which is
+ checked by CHECK_ARCH_EXT which produces return from individual test to
+ prevent run on hardware not supported by tested function implementation. */
+#define INIT_ARCH_EXT
+#define CHECK_ARCH_EXT
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-12-02 16:09 ` Andrew Senkevich
@ 2014-12-02 22:41 ` Joseph Myers
2014-12-29 9:19 ` Andrew Senkevich
0 siblings, 1 reply; 13+ messages in thread
From: Joseph Myers @ 2014-12-02 22:41 UTC (permalink / raw)
To: Andrew Senkevich; +Cc: libc-alpha
On Tue, 2 Dec 2014, Andrew Senkevich wrote:
> Is it OK for trunk?
OK once the copyright assignment is in.
--
Joseph S. Myers
joseph@codesourcery.com
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-12-02 22:41 ` Joseph Myers
@ 2014-12-29 9:19 ` Andrew Senkevich
2014-12-30 17:05 ` Joseph Myers
0 siblings, 1 reply; 13+ messages in thread
From: Andrew Senkevich @ 2014-12-29 9:19 UTC (permalink / raw)
To: Joseph Myers; +Cc: libc-alpha
2014-12-03 1:41 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> On Tue, 2 Dec 2014, Andrew Senkevich wrote:
>
>> Is it OK for trunk?
>
> OK once the copyright assignment is in.
Hi, Joseph, is it ok for commit with the following ChangeLog?
2014-12-28 Andrew Senkevich <andrew.senkevich@intel.com>
* math/libm-test.inc: START refactored.
* math/test-double.c (TEST_MATHVEC): Add define.
* math/test-float.c: Likewise.
* math/test-idouble.c: Likewise.
* math/test-ifloat.c: Likewise.
* math/test-ildoubl.c: Likewise.
* math/test-ldouble.c: Likewise.
* sysdeps/generic/math-tests-arch.h (INIT_ARCH_EXT, CHECK_ARCH_EXT):
New helper macros for runtime architecture check.
--
WBR,
Andrew
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-12-29 9:19 ` Andrew Senkevich
@ 2014-12-30 17:05 ` Joseph Myers
2014-12-30 17:24 ` Andrew Senkevich
0 siblings, 1 reply; 13+ messages in thread
From: Joseph Myers @ 2014-12-30 17:05 UTC (permalink / raw)
To: Andrew Senkevich; +Cc: libc-alpha
On Sun, 28 Dec 2014, Andrew Senkevich wrote:
> 2014-12-03 1:41 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> > On Tue, 2 Dec 2014, Andrew Senkevich wrote:
> >
> >> Is it OK for trunk?
> >
> > OK once the copyright assignment is in.
>
> Hi, Joseph, is it ok for commit with the following ChangeLog?
I still don't see the assignment in copyright.list.
--
Joseph S. Myers
joseph@codesourcery.com
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-12-30 17:05 ` Joseph Myers
@ 2014-12-30 17:24 ` Andrew Senkevich
2014-12-30 17:34 ` Joseph Myers
0 siblings, 1 reply; 13+ messages in thread
From: Andrew Senkevich @ 2014-12-30 17:24 UTC (permalink / raw)
To: Joseph Myers; +Cc: libc-alpha, Melik-Adamyan, Areg
2014-12-30 20:00 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> On Sun, 28 Dec 2014, Andrew Senkevich wrote:
>
>> 2014-12-03 1:41 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
>> > On Tue, 2 Dec 2014, Andrew Senkevich wrote:
>> >
>> >> Is it OK for trunk?
>> >
>> > OK once the copyright assignment is in.
>>
>> Hi, Joseph, is it ok for commit with the following ChangeLog?
>
> I still don't see the assignment in copyright.list.
Yes, there are no assignment in copyright.list but agreement between
Intel and FSF was achieved about contribution is possible from anyone
having @intel.com email. But may be FSF hasn't signed to formal
documents right now.
What do you think about it?
I have 17 patches (in format suitable for git am) and I am ready to send them.
--
WBR,
Andrew
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-12-30 17:24 ` Andrew Senkevich
@ 2014-12-30 17:34 ` Joseph Myers
2014-12-30 18:30 ` Andrew Senkevich
0 siblings, 1 reply; 13+ messages in thread
From: Joseph Myers @ 2014-12-30 17:34 UTC (permalink / raw)
To: Andrew Senkevich; +Cc: libc-alpha, Melik-Adamyan, Areg
On Tue, 30 Dec 2014, Andrew Senkevich wrote:
> 2014-12-30 20:00 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> > On Sun, 28 Dec 2014, Andrew Senkevich wrote:
> >
> >> 2014-12-03 1:41 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> >> > On Tue, 2 Dec 2014, Andrew Senkevich wrote:
> >> >
> >> >> Is it OK for trunk?
> >> >
> >> > OK once the copyright assignment is in.
> >>
> >> Hi, Joseph, is it ok for commit with the following ChangeLog?
> >
> > I still don't see the assignment in copyright.list.
>
> Yes, there are no assignment in copyright.list but agreement between
> Intel and FSF was achieved about contribution is possible from anyone
> having @intel.com email. But may be FSF hasn't signed to formal
> documents right now.
>
> What do you think about it?
As a GNU maintainer of glibc I need to wait for the assignment to appear
in copyright.list, or for the FSF to notify the GNU maintainers of glibc
in some other way.
--
Joseph S. Myers
joseph@codesourcery.com
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/N] [x86_64] Vectorized math functions
2014-12-30 17:34 ` Joseph Myers
@ 2014-12-30 18:30 ` Andrew Senkevich
0 siblings, 0 replies; 13+ messages in thread
From: Andrew Senkevich @ 2014-12-30 18:30 UTC (permalink / raw)
To: Joseph Myers; +Cc: libc-alpha, Melik-Adamyan, Areg
2014-12-30 20:24 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
> On Tue, 30 Dec 2014, Andrew Senkevich wrote:
>
>> 2014-12-30 20:00 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
>> > On Sun, 28 Dec 2014, Andrew Senkevich wrote:
>> >
>> >> 2014-12-03 1:41 GMT+03:00 Joseph Myers <joseph@codesourcery.com>:
>> >> > On Tue, 2 Dec 2014, Andrew Senkevich wrote:
>> >> >
>> >> >> Is it OK for trunk?
>> >> >
>> >> > OK once the copyright assignment is in.
>> >>
>> >> Hi, Joseph, is it ok for commit with the following ChangeLog?
>> >
>> > I still don't see the assignment in copyright.list.
>>
>> Yes, there are no assignment in copyright.list but agreement between
>> Intel and FSF was achieved about contribution is possible from anyone
>> having @intel.com email. But may be FSF hasn't signed to formal
>> documents right now.
>>
>> What do you think about it?
>
> As a GNU maintainer of glibc I need to wait for the assignment to appear
> in copyright.list, or for the FSF to notify the GNU maintainers of glibc
> in some other way.
Clear. I hope it will be confirmed before commits closing... Will
check for it after several days...
--
WBR,
Andrew
^ permalink raw reply [flat|nested] 13+ messages in thread
end of thread, other threads:[~2014-12-30 17:40 UTC | newest]
Thread overview: 13+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-11-20 17:40 [PATCH 2/N] [x86_64] Vectorized math functions Andrew Senkevich
2014-11-21 17:55 ` Joseph Myers
2014-11-24 18:05 ` Andrew Senkevich
2014-11-26 13:12 ` Joseph Myers
2014-11-26 16:03 ` Andrew Senkevich
2014-11-26 16:26 ` Joseph Myers
2014-12-02 16:09 ` Andrew Senkevich
2014-12-02 22:41 ` Joseph Myers
2014-12-29 9:19 ` Andrew Senkevich
2014-12-30 17:05 ` Joseph Myers
2014-12-30 17:24 ` Andrew Senkevich
2014-12-30 17:34 ` Joseph Myers
2014-12-30 18:30 ` Andrew Senkevich
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 read-only IMAP folder(s) and NNTP newsgroup(s).