Test cpow in all rounding modes
diff mbox

Message ID Pine.LNX.4.64.1406102300510.15277@digraph.polyomino.org.uk
State New
Headers show

Commit Message

Joseph Myers June 10, 2014, 11:01 p.m. UTC
This patch enables testing of cpow in all rounding modes using
ALL_RM_TEST.  There were two reasons this was previously deferred:

* MPC has complicated rounding-mode-dependent rules for the signs of
  exact zero real or imaginary parts in the result of mpc_pow.  Annex
  G does not impose any such requirements and I don't think glibc
  should try to implement any particular logic here.  This patch adds
  support for gen-auto-libm-tests passing the IGNORE_ZERO_INF_SIGN
  flag to libm-test.inc.

* Error accumulations in some tests in non-default rounding modes
  exceed the maximum error permitted in libm-test.inc.  This patch
  marks the problem tests with xfail-rounding.  (It might be possible
  to reduce the accumulations a bit by using round-to-nearest when
  cpow calls clog, but I don't think there's much point; the
  implementation approach for cpow is fundamentally deficient, as
  discussed in the existing bug for cpow inaccuracy which can
  reasonably be considered to cover these less-inaccurate cases as
  well.  It's possible that the test "cpow 2 0 10 0" will also need
  xfail-rounding on some platforms.)

Tested x86_64 and x86 and ulps updated accordingly.

(auto-libm-test-out diffs omitted below.)

2014-06-10  Joseph Myers  <joseph@codesourcery.com>

	* math/gen-auto-libm-tests.c: Document use of
	ignore-zero-inf-sign.
	(input_flag_type): Add value flag_ignore_zero_inf_sign.
	(input_flags): Add ignore-zero-inf-sign.
	(output_for_one_input_case): Handle flag_ignore_zero_inf_sign.
	* math/gen-libm-test.pl (generate_testfile): Handle
	ignore-zero-inf-sign.
	* math/auto-libm-test-in: Mark some cpow tests with
	ignore-zero-inf-sign and some with xfail-rounding.
	* math/auto-libm-test-out: Regenerated.
	* math/libm-test.inc (cpow_test): Use ALL_RM_TEST.
	* sysdeps/i386/fpu/libm-test-ulps: Update.
	* sysdeps/x86_64/fpu/libm-test-ulps: Likewise.

Comments

Joseph Myers June 18, 2014, 10:40 a.m. UTC | #1
Ping.  This patch 
<https://sourceware.org/ml/libc-alpha/2014-06/msg00250.html> is pending 
review.
Joseph Myers June 23, 2014, 11:54 a.m. UTC | #2
Ping^2.  This patch 
<https://sourceware.org/ml/libc-alpha/2014-06/msg00250.html> is still 
pending review.
Andreas Jaeger June 23, 2014, 6:23 p.m. UTC | #3
On 06/11/2014 01:01 AM, Joseph S. Myers wrote:
> This patch enables testing of cpow in all rounding modes using
> ALL_RM_TEST.  There were two reasons this was previously deferred:
> 
> * MPC has complicated rounding-mode-dependent rules for the signs of
>   exact zero real or imaginary parts in the result of mpc_pow.  Annex
>   G does not impose any such requirements and I don't think glibc
>   should try to implement any particular logic here.  This patch adds
>   support for gen-auto-libm-tests passing the IGNORE_ZERO_INF_SIGN
>   flag to libm-test.inc.
> 
> * Error accumulations in some tests in non-default rounding modes
>   exceed the maximum error permitted in libm-test.inc.  This patch
>   marks the problem tests with xfail-rounding.  (It might be possible
>   to reduce the accumulations a bit by using round-to-nearest when
>   cpow calls clog, but I don't think there's much point; the
>   implementation approach for cpow is fundamentally deficient, as
>   discussed in the existing bug for cpow inaccuracy which can
>   reasonably be considered to cover these less-inaccurate cases as
>   well.  It's possible that the test "cpow 2 0 10 0" will also need
>   xfail-rounding on some platforms.)
> 
> Tested x86_64 and x86 and ulps updated accordingly.

Looks fine, thanks,

Andreas

Patch
diff mbox

diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in
index 6edad5a..bfbc085 100644
--- a/math/auto-libm-test-in
+++ b/math/auto-libm-test-in
@@ -652,14 +652,14 @@  cosh 0x2.c5d37700c6bb03a6c24b6c9b494ep+12 no-test-inline xfail-rounding:ldbl-128
 cosh -0x2.c5d37700c6bb03a6c24b6c9b494cp+12 no-test-inline xfail-rounding:ldbl-128ibm
 cosh -0x2.c5d37700c6bb03a6c24b6c9b494ep+12 no-test-inline xfail-rounding:ldbl-128ibm
 
-cpow 1 0 0 0
-cpow 2 0 10 0
+cpow 1 0 0 0 ignore-zero-inf-sign
+cpow 2 0 10 0 ignore-zero-inf-sign
 # Bug 14473: cpow results inaccurate.
 cpow e 0 0 2pi xfail
-cpow 2 3 4 0
+cpow 2 3 4 0 xfail-rounding
 
-cpow 0.75 1.25 0.75 1.25
-cpow 0.75 1.25 1.0 1.0
+cpow 0.75 1.25 0.75 1.25 xfail-rounding
+cpow 0.75 1.25 1.0 1.0 xfail-rounding
 cpow 0.75 1.25 1.0 0.0
 cpow 0.75 1.25 0.0 1.0
 
diff --git a/math/gen-auto-libm-tests.c b/math/gen-auto-libm-tests.c
index 132cca2..12c530a 100644
--- a/math/gen-auto-libm-tests.c
+++ b/math/gen-auto-libm-tests.c
@@ -84,15 +84,16 @@ 
    permitted), errno is expected to be left unchanged.
 
    The flag "no-test-inline" indicates a test is disabled for inline
-   function testing; "xfail" indicates the test is disabled as
-   expected to produce incorrect results, "xfail-rounding" indicates
-   the test is disabled only in rounding modes other than
-   round-to-nearest.  Otherwise, test flags are of the form
-   "spurious-<exception>" and "missing-<exception>", for any exception
-   ("overflow", "underflow", "inexact", "invalid", "divbyzero"),
-   "spurious-errno" and "missing-errno", to indicate when tests are
-   expected to deviate from the exception and errno settings
-   corresponding to the mathematical results.  "xfail",
+   function testing; "ignore-zero-inf-sign" indicates the the signs of
+   zero and infinite results should be ignored; "xfail" indicates the
+   test is disabled as expected to produce incorrect results,
+   "xfail-rounding" indicates the test is disabled only in rounding
+   modes other than round-to-nearest.  Otherwise, test flags are of
+   the form "spurious-<exception>" and "missing-<exception>", for any
+   exception ("overflow", "underflow", "inexact", "invalid",
+   "divbyzero"), "spurious-errno" and "missing-errno", to indicate
+   when tests are expected to deviate from the exception and errno
+   settings corresponding to the mathematical results.  "xfail",
    "xfail-rounding", "spurious-" and "missing-" flags should be
    accompanied by a comment referring to an open bug in glibc
    Bugzilla.
@@ -109,17 +110,17 @@ 
    plus_infty or minus_infty for infinite expected results, or as
    integer constant expressions (not necessarily with the right type)
    or IGNORE for integer inputs and outputs.  Flags are
-   "no-test-inline", "xfail", "<exception>", "<exception>-ok",
-   "errno-<value>", "errno-<value>-ok", which may be unconditional or
-   conditional.  "<exception>" indicates that a correct result means
-   the given exception should be raised.  "errno-<value>" indicates
-   that a correct result means errno should be set to the given value.
-   "-ok" means not to test for the given exception or errno value
-   (whether because it was marked as possibly missing or spurious, or
-   because the calculation of correct results indicated it was
-   optional).  Conditions "before-rounding" and "after-rounding"
-   indicate tests where expectations for underflow exceptions depend
-   on how the architecture detects tininess.  */
+   "no-test-inline", "ignore-zero-info-sign", "xfail", "<exception>",
+   "<exception>-ok", "errno-<value>", "errno-<value>-ok", which may be
+   unconditional or conditional.  "<exception>" indicates that a
+   correct result means the given exception should be raised.
+   "errno-<value>" indicates that a correct result means errno should
+   be set to the given value.  "-ok" means not to test for the given
+   exception or errno value (whether because it was marked as possibly
+   missing or spurious, or because the calculation of correct results
+   indicated it was optional).  Conditions "before-rounding" and
+   "after-rounding" indicate tests where expectations for underflow
+   exceptions depend on how the architecture detects tininess.  */
 
 #define _GNU_SOURCE
 
@@ -315,6 +316,7 @@  typedef struct
 typedef enum
   {
     flag_no_test_inline,
+    flag_ignore_zero_inf_sign,
     flag_xfail,
     flag_xfail_rounding,
     /* The "spurious" and "missing" flags must be in the same order as
@@ -342,6 +344,7 @@  typedef enum
 static const char *const input_flags[num_input_flag_types] =
   {
     "no-test-inline",
+    "ignore-zero-inf-sign",
     "xfail",
     "xfail-rounding",
     "spurious-divbyzero",
@@ -1950,6 +1953,7 @@  output_for_one_input_case (FILE *fp, const char *filename, test_function *tf,
 		switch (it->flags[i].type)
 		  {
 		  case flag_no_test_inline:
+		  case flag_ignore_zero_inf_sign:
 		  case flag_xfail:
 		    if (fprintf (fp, " %s%s",
 				 input_flags[it->flags[i].type],
diff --git a/math/gen-libm-test.pl b/math/gen-libm-test.pl
index 5f62872..b5d599f 100755
--- a/math/gen-libm-test.pl
+++ b/math/gen-libm-test.pl
@@ -442,6 +442,10 @@  sub generate_testfile {
 	    }
 	  }
 	  $flags_conv = "";
+	  if (defined ($flag_cond{"ignore-zero-inf-sign"})) {
+	    $flags_conv .= or_cond_value ($flag_cond{"ignore-zero-inf-sign"},
+					  "IGNORE_ZERO_INF_SIGN", "0");
+	  }
 	  if (defined ($flag_cond{"no-test-inline"})) {
 	    $flags_conv .= or_cond_value ($flag_cond{"no-test-inline"},
 					  "NO_TEST_INLINE", "0");
diff --git a/math/libm-test.inc b/math/libm-test.inc
index fa8e238..05febc5 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -6288,9 +6288,7 @@  static const struct test_cc_c_data cpow_test_data[] =
 static void
 cpow_test (void)
 {
-  START (cpow, 0);
-  RUN_TEST_LOOP_cc_c (cpow, cpow_test_data, );
-  END_COMPLEX;
+  ALL_RM_TEST (cpow, 0, cpow_test_data, RUN_TEST_LOOP_cc_c, END_COMPLEX);
 }
 
 
diff --git a/sysdeps/i386/fpu/libm-test-ulps b/sysdeps/i386/fpu/libm-test-ulps
index d7424a6..d0dc821 100644
--- a/sysdeps/i386/fpu/libm-test-ulps
+++ b/sysdeps/i386/fpu/libm-test-ulps
@@ -916,6 +916,54 @@  ifloat: 1
 ildouble: 4
 ldouble: 4
 
+Function: Real part of "cpow_downward":
+double: 5
+float: 8
+idouble: 5
+ifloat: 8
+ildouble: 7
+ldouble: 7
+
+Function: Imaginary part of "cpow_downward":
+double: 1
+float: 1
+idouble: 1
+ifloat: 1
+ildouble: 2
+ldouble: 2
+
+Function: Real part of "cpow_towardzero":
+double: 5
+float: 8
+idouble: 5
+ifloat: 8
+ildouble: 7
+ldouble: 7
+
+Function: Imaginary part of "cpow_towardzero":
+double: 1
+float: 1
+idouble: 1
+ifloat: 1
+ildouble: 1
+ldouble: 1
+
+Function: Real part of "cpow_upward":
+double: 4
+float: 1
+idouble: 4
+ifloat: 1
+ildouble: 2
+ldouble: 2
+
+Function: Imaginary part of "cpow_upward":
+double: 1
+float: 1
+idouble: 1
+ifloat: 1
+ildouble: 2
+ldouble: 2
+
 Function: Real part of "csin":
 double: 1
 float: 1
diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps
index 92fef5a..dbe9c46 100644
--- a/sysdeps/x86_64/fpu/libm-test-ulps
+++ b/sysdeps/x86_64/fpu/libm-test-ulps
@@ -959,6 +959,54 @@  ifloat: 2
 ildouble: 4
 ldouble: 4
 
+Function: Real part of "cpow_downward":
+double: 4
+float: 8
+idouble: 4
+ifloat: 8
+ildouble: 7
+ldouble: 7
+
+Function: Imaginary part of "cpow_downward":
+double: 2
+float: 2
+idouble: 2
+ifloat: 2
+ildouble: 2
+ldouble: 2
+
+Function: Real part of "cpow_towardzero":
+double: 4
+float: 8
+idouble: 4
+ifloat: 8
+ildouble: 7
+ldouble: 7
+
+Function: Imaginary part of "cpow_towardzero":
+double: 2
+float: 2
+idouble: 2
+ifloat: 2
+ildouble: 1
+ldouble: 1
+
+Function: Real part of "cpow_upward":
+double: 4
+float: 1
+idouble: 4
+ifloat: 1
+ildouble: 2
+ldouble: 2
+
+Function: Imaginary part of "cpow_upward":
+double: 1
+float: 2
+idouble: 1
+ifloat: 2
+ildouble: 2
+ldouble: 2
+
 Function: Real part of "csin":
 double: 1
 float: 1