uclibc-ng-test and math complex functions
Hi, I'm trying to build uclibc-ng-test, but I can't link with complex trigo functions (such as cacos, casin, catan, etc.). They are required for the following test targets: test-double, test-double-finite, test-float, test-float-finite, test-idouble, and test-ifloat. They are declared in uclibc-ng/sysdeps/linux/common/bits/cmathcalls.h, but I can't find their definitions. Does uclibc-ng define them? Thanks, -- Laurent Thévenoux lthevenoux@kalray.eu Kalray S.A. 180 Avenue de l'Europe 38330 Montbonnot FRANCE
hi, you have to enable complex math in your uclibc-ng config. They are optional and default off IIRC. best regards Waldemar
Am 18.07.2018 um 10:34 schrieb Laurent Thévenoux <lthevenoux@kalray.eu>:
Hi,
I'm trying to build uclibc-ng-test, but I can't link with complex trigo functions (such as cacos, casin, catan, etc.). They are required for the following test targets: test-double, test-double-finite, test-float, test-float-finite, test-idouble, and test-ifloat. They are declared in uclibc-ng/sysdeps/linux/common/bits/cmathcalls.h, but I can't find their definitions.
Does uclibc-ng define them?
Thanks,
--
Laurent Thévenoux
lthevenoux@kalray.eu
Kalray S.A. 180 Avenue de l'Europe 38330 Montbonnot FRANCE _______________________________________________ devel mailing list devel@uclibc-ng.org https://mailman.uclibc-ng.org/cgi-bin/mailman/listinfo/devel
I don't find this option, neither in extra/Configs/Config.in or extra/Configs/Config.in.arch. What its name is? What I'm missing? I have set this three math related options: DO_C99_MATH=y DO_XSI_MATH=y UCLIBC_HAS_LONG_DOUBLE_MATH=y Thanks, Laurent ----- Mail original ----- De: "Waldemar Brodkorb" <wbx@uclibc-ng.org> À: "Laurent Thévenoux" <lthevenoux@kalray.eu> Cc: "devel" <devel@uclibc-ng.org> Envoyé: Mercredi 18 Juillet 2018 10:57:37 Objet: Re: [uclibc-ng-devel] uclibc-ng-test and math complex functions hi, you have to enable complex math in your uclibc-ng config. They are optional and default off IIRC. best regards Waldemar
Am 18.07.2018 um 10:34 schrieb Laurent Thévenoux <lthevenoux@kalray.eu>:
Hi,
I'm trying to build uclibc-ng-test, but I can't link with complex trigo functions (such as cacos, casin, catan, etc.). They are required for the following test targets: test-double, test-double-finite, test-float, test-float-finite, test-idouble, and test-ifloat. They are declared in uclibc-ng/sysdeps/linux/common/bits/cmathcalls.h, but I can't find their definitions.
Does uclibc-ng define them?
Thanks,
--
Laurent Thévenoux
lthevenoux@kalray.eu
Kalray S.A. 180 Avenue de l'Europe 38330 Montbonnot FRANCE _______________________________________________ devel mailing list devel@uclibc-ng.org https://mailman.uclibc-ng.org/cgi-bin/mailman/listinfo/devel
Hi, you are right. My last answer was simply wrong. Sorry for the wrong direction. Can you show the command you are using to compile the testsuite and the failing output? best regards Waldemar
Am 18.07.2018 um 11:19 schrieb Laurent Thévenoux <lthevenoux@kalray.eu>:
I don't find this option, neither in extra/Configs/Config.in or extra/Configs/Config.in.arch.
What its name is? What I'm missing?
I have set this three math related options:
DO_C99_MATH=y DO_XSI_MATH=y UCLIBC_HAS_LONG_DOUBLE_MATH=y
Thanks, Laurent
----- Mail original ----- De: "Waldemar Brodkorb" <wbx@uclibc-ng.org> À: "Laurent Thévenoux" <lthevenoux@kalray.eu> Cc: "devel" <devel@uclibc-ng.org> Envoyé: Mercredi 18 Juillet 2018 10:57:37 Objet: Re: [uclibc-ng-devel] uclibc-ng-test and math complex functions
hi,
you have to enable complex math in your uclibc-ng config. They are optional and default off IIRC.
best regards Waldemar
Am 18.07.2018 um 10:34 schrieb Laurent Thévenoux <lthevenoux@kalray.eu>:
Hi,
I'm trying to build uclibc-ng-test, but I can't link with complex trigo functions (such as cacos, casin, catan, etc.). They are required for the following test targets: test-double, test-double-finite, test-float, test-float-finite, test-idouble, and test-ifloat. They are declared in uclibc-ng/sysdeps/linux/common/bits/cmathcalls.h, but I can't find their definitions.
Does uclibc-ng define them?
Thanks,
--
Laurent Thévenoux
lthevenoux@kalray.eu
Kalray S.A. 180 Avenue de l'Europe 38330 Montbonnot FRANCE _______________________________________________ devel mailing list devel@uclibc-ng.org https://mailman.uclibc-ng.org/cgi-bin/mailman/listinfo/devel
On 07/18/2018 02:37 AM, Waldemar Brodkorb wrote:
Hi,
you are right. My last answer was simply wrong. Sorry for the wrong direction.
Can you show the command you are using to compile the testsuite and the failing output?
I have the exact same issue with ARC test suite builds as well so suppose it is a common problem. TEST_LINK math/ test-double test-double.o: In function `cacos_test': test-double.c:(.text+0x3288): undefined reference to `cacos' test-double.c:(.text+0x3288): undefined reference to `cacos' ... And the whole complex stuff is really ocmplex and heavily macro'ized so I agve up on trying to unravel it.
Waldemar, find attached patch which disables complex math tests. Regards, Laurent ----- Mail original ----- De: "Vineet Gupta" <Vineet.Gupta1@synopsys.com> À: "Waldemar Brodkorb" <wbx@uclibc-ng.org>, "Laurent Thévenoux" <lthevenoux@kalray.eu> Cc: "devel" <devel@uclibc-ng.org> Envoyé: Jeudi 19 Juillet 2018 20:44:08 Objet: Re: [uclibc-ng-devel] uclibc-ng-test and math complex functions On 07/18/2018 02:37 AM, Waldemar Brodkorb wrote:
Hi,
you are right. My last answer was simply wrong. Sorry for the wrong direction.
Can you show the command you are using to compile the testsuite and the failing output?
I have the exact same issue with ARC test suite builds as well so suppose it is a common problem. TEST_LINK math/ test-double test-double.o: In function `cacos_test': test-double.c:(.text+0x3288): undefined reference to `cacos' test-double.c:(.text+0x3288): undefined reference to `cacos' ... And the whole complex stuff is really ocmplex and heavily macro'ized so I agve up on trying to unravel it.
Hi, thanks! Laurent Thévenoux wrote,
Waldemar,
find attached patch which disables complex math tests.
Regards, Laurent
----- Mail original ----- De: "Vineet Gupta" <Vineet.Gupta1@synopsys.com> À: "Waldemar Brodkorb" <wbx@uclibc-ng.org>, "Laurent Thévenoux" <lthevenoux@kalray.eu> Cc: "devel" <devel@uclibc-ng.org> Envoyé: Jeudi 19 Juillet 2018 20:44:08 Objet: Re: [uclibc-ng-devel] uclibc-ng-test and math complex functions
On 07/18/2018 02:37 AM, Waldemar Brodkorb wrote:
Hi,
you are right. My last answer was simply wrong. Sorry for the wrong direction.
Can you show the command you are using to compile the testsuite and the failing output?
I have the exact same issue with ARC test suite builds as well so suppose it is a common problem.
TEST_LINK math/ test-double test-double.o: In function `cacos_test': test-double.c:(.text+0x3288): undefined reference to `cacos' test-double.c:(.text+0x3288): undefined reference to `cacos' ...
And the whole complex stuff is really ocmplex and heavily macro'ized so I agve up on trying to unravel it.
diff --git a/test/math/libm-test.inc b/test/math/libm-test.inc index 780b612..04a5e73 100644 --- a/test/math/libm-test.inc +++ b/test/math/libm-test.inc @@ -56,7 +56,8 @@ scalb, scalbn, scalbln, signbit, sin, sincos, sinh, sqrt, tan, tanh, tgamma, trunc, y0, y1, yn, significand
- and for the following complex math functions: + tests for the following complex math functions are disabled because they are + not implemented: cabs, cacos, cacosh, carg, casin, casinh, catan, catanh, ccos, ccosh, cexp, cimag, clog, clog10, conj, cpow, cproj, creal, csin, csinh, csqrt, ctan, ctanh. @@ -2819,11 +2820,14 @@ static const struct test_c_c_data cacos_test_data[] = TEST_c_c (cacos, -2, -3, 2.1414491111159960199416055713254211L, 1.9833870299165354323470769028940395L), };
+#if 0 static void cacos_test (void) { ALL_RM_TEST (cacos, 0, cacos_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif +
static const struct test_c_c_data cacosh_test_data[] = { @@ -3501,11 +3505,13 @@ static const struct test_c_c_data cacosh_test_data[] = TEST_c_c (cacosh, -2, -3, 1.9833870299165354323470769028940395L, -2.1414491111159960199416055713254211L), };
+#if 0 static void cacosh_test (void) { ALL_RM_TEST (cacosh, 0, cacosh_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_f_data carg_test_data[] = @@ -4232,11 +4238,13 @@ static const struct test_c_c_data casin_test_data[] = TEST_c_c (casin, -2, -3, -0.57065278432109940071028387968566963L, -1.9833870299165354323470769028940395L), };
+#if 0 static void casin_test (void) { ALL_RM_TEST (casin, 0, casin_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_c_data casinh_test_data[] = @@ -4922,11 +4930,13 @@ static const struct test_c_c_data casinh_test_data[] = TEST_c_c (casinh, -2, -3, -1.9686379257930962917886650952454982L, -0.96465850440760279204541105949953237L), };
+#if 0 static void casinh_test (void) { ALL_RM_TEST (casinh, 0, casinh_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_c_data catan_test_data[] = @@ -5429,11 +5439,14 @@ static const struct test_c_c_data catan_test_data[] = TEST_c_c (catan, -2, -3, -1.4099210495965755225306193844604208L, -0.22907268296853876629588180294200276L), };
+#if 0 static void catan_test (void) { ALL_RM_TEST (catan, 0, catan_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif +
static const struct test_c_c_data catanh_test_data[] = { @@ -5934,11 +5947,14 @@ static const struct test_c_c_data catanh_test_data[] = TEST_c_c (catanh, -2, -3, -0.14694666622552975204743278515471595L, -1.3389725222944935611241935759091443L), };
+#if 0 static void catanh_test (void) { ALL_RM_TEST (catanh, 0, catanh_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif +
static const struct test_f_f_data cbrt_test_data[] = { @@ -6009,11 +6025,13 @@ static const struct test_c_c_data ccos_test_data[] = AUTO_TESTS_c_c (ccos), };
+#if 0 static void ccos_test (void) { ALL_RM_TEST (ccos, 0, ccos_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_c_data ccosh_test_data[] = @@ -6066,11 +6084,13 @@ static const struct test_c_c_data ccosh_test_data[] = AUTO_TESTS_c_c (ccosh), };
+#if 0 static void ccosh_test (void) { ALL_RM_TEST (ccosh, 0, ccosh_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_f_f_data ceil_test_data[] = @@ -6289,11 +6309,14 @@ static const struct test_c_f_data cimag_test_data[] = TEST_c_f (cimag, 2.0, 3.0, 3.0, NO_INEXACT_EXCEPTION), };
+#if 0 static void cimag_test (void) { ALL_RM_TEST (cimag, 1, cimag_test_data, RUN_TEST_LOOP_c_f, END); } +#endif +
static const struct test_c_c_data clog_test_data[] = { @@ -6349,11 +6372,13 @@ static const struct test_c_c_data clog_test_data[] = AUTO_TESTS_c_c (clog), };
+#if 0 static void clog_test (void) { ALL_RM_TEST (clog, 0, clog_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_c_data clog10_test_data[] = @@ -6409,11 +6434,13 @@ static const struct test_c_c_data clog10_test_data[] = AUTO_TESTS_c_c (clog10), };
+#if 0 static void clog10_test (void) { ALL_RM_TEST (clog10, 0, clog10_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_c_data conj_test_data[] = @@ -6430,11 +6457,13 @@ static const struct test_c_c_data conj_test_data[] = TEST_c_c (conj, 3.0, -4.0, 3.0, 4.0, NO_INEXACT_EXCEPTION), };
+#if 0 static void conj_test (void) { ALL_RM_TEST (conj, 1, conj_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_ff_f_data copysign_test_data[] = @@ -6572,11 +6601,13 @@ static const struct test_cc_c_data cpow_test_data[] = AUTO_TESTS_cc_c (cpow), };
+#if 0 static void cpow_test (void) { ALL_RM_TEST (cpow, 0, cpow_test_data, RUN_TEST_LOOP_cc_c, END_COMPLEX); } +#endif
static const struct test_c_c_data cproj_test_data[] = @@ -6637,11 +6668,13 @@ static const struct test_c_c_data cproj_test_data[] = TEST_c_c (cproj, 2.0, 3.0, 2.0, 3.0, NO_INEXACT_EXCEPTION), };
+#if 0 static void cproj_test (void) { ALL_RM_TEST (cproj, 1, cproj_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_f_data creal_test_data[] = @@ -6659,11 +6692,14 @@ static const struct test_c_f_data creal_test_data[] = TEST_c_f (creal, 2.0, 3.0, 2.0, NO_INEXACT_EXCEPTION), };
+#if 0 static void creal_test (void) { ALL_RM_TEST (creal, 1, creal_test_data, RUN_TEST_LOOP_c_f, END); } +#endif +
static const struct test_c_c_data csin_test_data[] = { @@ -6718,11 +6754,13 @@ static const struct test_c_c_data csin_test_data[] = TEST_c_c (csin, min_value, min_subnorm_value, min_value, min_subnorm_value, UNDERFLOW_EXCEPTION), };
+#if 0 static void csin_test (void) { ALL_RM_TEST (csin, 0, csin_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_c_data csinh_test_data[] = @@ -6778,11 +6816,13 @@ static const struct test_c_c_data csinh_test_data[] = TEST_c_c (csinh, min_value, min_subnorm_value, min_value, min_subnorm_value, UNDERFLOW_EXCEPTION), };
+#if 0 static void csinh_test (void) { ALL_RM_TEST (csinh, 0, csinh_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_c_data csqrt_test_data[] = @@ -6829,11 +6869,14 @@ static const struct test_c_c_data csqrt_test_data[] = AUTO_TESTS_c_c (csqrt), };
+#if 0 static void csqrt_test (void) { ALL_RM_TEST (csqrt, 0, csqrt_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif +
static const struct test_c_c_data ctan_test_data[] = { @@ -6939,11 +6982,13 @@ static const struct test_c_c_data ctan_test_data[] = AUTO_TESTS_c_c (ctan), };
+#if 0 static void ctan_test (void) { ALL_RM_TEST (ctan, 0, ctan_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_c_c_data ctanh_test_data[] = @@ -7049,11 +7094,13 @@ static const struct test_c_c_data ctanh_test_data[] = AUTO_TESTS_c_c (ctanh), };
+#if 0 static void ctanh_test (void) { ALL_RM_TEST (ctanh, 0, ctanh_test_data, RUN_TEST_LOOP_c_c, END_COMPLEX); } +#endif
static const struct test_f_f_data erf_test_data[] =
On 7/27/18 4:43 AM, Laurent Thévenoux wrote:
Waldemar,
find attached patch which disables complex math tests.
Regards, Laurent
So turns out before this, we were NOT effectively testing float at all test-double etc were failing to link. I tried the updated testsuite and now test-double et al go belly up and I can see lack of any float support code for ARC :-(
participants (5)
-
Laurent Thévenoux -
Vineet Gupta -
Vineet Gupta
-
Waldemar Brodkorb -
Waldemar Brodkorb