diff options
author | Joseph Myers <joseph@codesourcery.com> | 2016-12-15 23:52:18 +0000 |
---|---|---|
committer | Joseph Myers <joseph@codesourcery.com> | 2016-12-15 23:52:18 +0000 |
commit | 0a2546cdaa4284cc6ed6af06835fd858ba341838 (patch) | |
tree | 7396415495f5f470f72cf4e73b77e01baebc1635 | |
parent | 67f779f1ead0bf38d1ca0e882e20c7ce0c03c204 (diff) | |
download | glibc-0a2546cdaa4284cc6ed6af06835fd858ba341838.tar.gz glibc-0a2546cdaa4284cc6ed6af06835fd858ba341838.tar.xz glibc-0a2546cdaa4284cc6ed6af06835fd858ba341838.zip |
Fix x86, x86_64 fmax, fmin sNaN handling, add tests (bug 20947).
Various fmax and fmin function implementations mishandle sNaN arguments: (a) When both arguments are NaNs, the return value should be a qNaN, but sometimes it is an sNaN if at least one argument is an sNaN. (b) Under TS 18661-1 semantics, if either argument is an sNaN then the result should be a qNaN (whereas if one argument is a qNaN and the other is not a NaN, the result should be the non-NaN argument). Various implementations treat sNaNs like qNaNs here. This patch fixes the x86 and x86_64 versions (ignoring float and double for 32-bit x86 given the inability to reliably avoid the sNaN turning into a qNaN before it gets to the called function). Tests of sNaN inputs to these functions are added. Note on architecture versions I haven't changed for this issue: AArch64 already gets this right (it uses a hardware instruction with the correct semantics for both quiet and signaling NaNs) and does not need changes. It's possible Alpha, IA64, SPARC might need changes (this would be shown by the testsuite if so). Tested for x86_64 and x86 (both i686 and i586 builds, to cover the different x86 implementations). [BZ #20947] * sysdeps/i386/fpu/s_fmaxl.S (__fmaxl): Add the arguments when either is a signaling NaN. * sysdeps/i386/fpu/s_fminl.S (__fminl): Likewise. Make code follow fmaxl more closely. * sysdeps/i386/i686/fpu/s_fmaxl.S (__fmaxl): Add the arguments when either is a signaling NaN. * sysdeps/i386/i686/fpu/s_fminl.S (__fminl): Likewise. * sysdeps/x86_64/fpu/s_fmax.S (__fmax): Likewise. * sysdeps/x86_64/fpu/s_fmaxf.S (__fmaxf): Likewise. * sysdeps/x86_64/fpu/s_fmaxl.S (__fmaxl): Likewise. * sysdeps/x86_64/fpu/s_fmin.S (__fmin): Likewise. * sysdeps/x86_64/fpu/s_fminf.S (__fminf): Likewise. * sysdeps/x86_64/fpu/s_fminl.S (__fminl): Likewise. * math/libm-test.inc (fmax_test_data): Add tests of sNaN inputs. (fmin_test_data): Likewise.
-rw-r--r-- | ChangeLog | 19 | ||||
-rw-r--r-- | math/libm-test.inc | 72 | ||||
-rw-r--r-- | sysdeps/i386/fpu/s_fmaxl.S | 30 | ||||
-rw-r--r-- | sysdeps/i386/fpu/s_fminl.S | 40 | ||||
-rw-r--r-- | sysdeps/i386/i686/fpu/s_fmaxl.S | 29 | ||||
-rw-r--r-- | sysdeps/i386/i686/fpu/s_fminl.S | 27 | ||||
-rw-r--r-- | sysdeps/x86_64/fpu/s_fmax.S | 19 | ||||
-rw-r--r-- | sysdeps/x86_64/fpu/s_fmaxf.S | 19 | ||||
-rw-r--r-- | sysdeps/x86_64/fpu/s_fmaxl.S | 29 | ||||
-rw-r--r-- | sysdeps/x86_64/fpu/s_fmin.S | 19 | ||||
-rw-r--r-- | sysdeps/x86_64/fpu/s_fminf.S | 19 | ||||
-rw-r--r-- | sysdeps/x86_64/fpu/s_fminl.S | 27 |
12 files changed, 322 insertions, 27 deletions
diff --git a/ChangeLog b/ChangeLog index 523028da4a..3f7e5d6e20 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,22 @@ +2016-12-15 Joseph Myers <joseph@codesourcery.com> + + [BZ #20947] + * sysdeps/i386/fpu/s_fmaxl.S (__fmaxl): Add the arguments when + either is a signaling NaN. + * sysdeps/i386/fpu/s_fminl.S (__fminl): Likewise. Make code + follow fmaxl more closely. + * sysdeps/i386/i686/fpu/s_fmaxl.S (__fmaxl): Add the arguments + when either is a signaling NaN. + * sysdeps/i386/i686/fpu/s_fminl.S (__fminl): Likewise. + * sysdeps/x86_64/fpu/s_fmax.S (__fmax): Likewise. + * sysdeps/x86_64/fpu/s_fmaxf.S (__fmaxf): Likewise. + * sysdeps/x86_64/fpu/s_fmaxl.S (__fmaxl): Likewise. + * sysdeps/x86_64/fpu/s_fmin.S (__fmin): Likewise. + * sysdeps/x86_64/fpu/s_fminf.S (__fminf): Likewise. + * sysdeps/x86_64/fpu/s_fminl.S (__fminl): Likewise. + * math/libm-test.inc (fmax_test_data): Add tests of sNaN inputs. + (fmin_test_data): Likewise. + 2016-12-15 Andreas Schwab <schwab@suse.de> * support/support_test_main.c (support_test_main): Don't shadow diff --git a/math/libm-test.inc b/math/libm-test.inc index e973a3f6ae..110b4215d5 100644 --- a/math/libm-test.inc +++ b/math/libm-test.inc @@ -7799,6 +7799,14 @@ static const struct test_ff_f_data fmax_test_data[] = TEST_ff_f (fmax, 9, -qnan_value, 9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, -9, qnan_value, -9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, -9, -qnan_value, -9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmax, 0, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, 0, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, minus_zero, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, minus_zero, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, 9, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, 9, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -9, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -9, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), TEST_ff_f (fmax, qnan_value, 0, 0, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, -qnan_value, 0, 0, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, qnan_value, minus_zero, minus_zero, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), @@ -7807,18 +7815,46 @@ static const struct test_ff_f_data fmax_test_data[] = TEST_ff_f (fmax, -qnan_value, 9, 9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, qnan_value, -9, -9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, -qnan_value, -9, -9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmax, snan_value, 0, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, 0, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, snan_value, minus_zero, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, minus_zero, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, snan_value, 9, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, 9, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, snan_value, -9, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, -9, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), TEST_ff_f (fmax, plus_infty, qnan_value, plus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, plus_infty, -qnan_value, plus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, minus_infty, qnan_value, minus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, minus_infty, -qnan_value, minus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmax, plus_infty, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, plus_infty, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, minus_infty, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, minus_infty, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), TEST_ff_f (fmax, qnan_value, plus_infty, plus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, -qnan_value, plus_infty, plus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, qnan_value, minus_infty, minus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, -qnan_value, minus_infty, minus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmax, snan_value, plus_infty, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, plus_infty, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, snan_value, minus_infty, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, minus_infty, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), TEST_ff_f (fmax, qnan_value, qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, qnan_value, -qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, -qnan_value, qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmax, -qnan_value, -qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmax, qnan_value, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, qnan_value, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -qnan_value, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -qnan_value, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, snan_value, qnan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, snan_value, -qnan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, qnan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, -qnan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, snan_value, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, snan_value, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmax, -snan_value, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), }; static void @@ -7862,6 +7898,14 @@ static const struct test_ff_f_data fmin_test_data[] = TEST_ff_f (fmin, 9, -qnan_value, 9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, -9, qnan_value, -9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, -9, -qnan_value, -9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmin, 0, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, 0, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, minus_zero, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, minus_zero, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, 9, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, 9, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -9, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -9, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), TEST_ff_f (fmin, qnan_value, 0, 0, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, -qnan_value, 0, 0, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, qnan_value, minus_zero, minus_zero, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), @@ -7870,18 +7914,46 @@ static const struct test_ff_f_data fmin_test_data[] = TEST_ff_f (fmin, -qnan_value, 9, 9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, qnan_value, -9, -9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, -qnan_value, -9, -9, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmin, snan_value, 0, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, 0, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, snan_value, minus_zero, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, minus_zero, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, snan_value, 9, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, 9, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, snan_value, -9, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, -9, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), TEST_ff_f (fmin, plus_infty, qnan_value, plus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, plus_infty, -qnan_value, plus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, minus_infty, qnan_value, minus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, minus_infty, -qnan_value, minus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmin, plus_infty, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, plus_infty, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, minus_infty, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, minus_infty, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), TEST_ff_f (fmin, qnan_value, plus_infty, plus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, -qnan_value, plus_infty, plus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, qnan_value, minus_infty, minus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, -qnan_value, minus_infty, minus_infty, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmin, snan_value, plus_infty, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, plus_infty, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, snan_value, minus_infty, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, minus_infty, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), TEST_ff_f (fmin, qnan_value, qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, qnan_value, -qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, -qnan_value, qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmin, -qnan_value, -qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmin, qnan_value, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, qnan_value, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -qnan_value, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -qnan_value, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, snan_value, qnan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, snan_value, -qnan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, qnan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, -qnan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, snan_value, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, snan_value, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), + TEST_ff_f (fmin, -snan_value, -snan_value, qnan_value, NO_INEXACT_EXCEPTION|INVALID_EXCEPTION), }; static void diff --git a/sysdeps/i386/fpu/s_fmaxl.S b/sysdeps/i386/fpu/s_fmaxl.S index a38a1946bc..a30401a4db 100644 --- a/sysdeps/i386/fpu/s_fmaxl.S +++ b/sysdeps/i386/fpu/s_fmaxl.S @@ -28,7 +28,13 @@ ENTRY(__fmaxl) andb $0x45, %ah cmpb $0x01, %ah - je 1f // y == NaN + je 2f // y == NaN + + fxam + fnstsw + andb $0x45, %ah + cmpb $0x01, %ah + je 3f // x == NaN fucom %st(1) fnstsw @@ -39,5 +45,27 @@ ENTRY(__fmaxl) 1: fstp %st(1) ret + +2: // st(1) is a NaN; st(0) may or may not be. + fxam + fnstsw + andb $0x45, %ah + cmpb $0x01, %ah + je 4f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 23(%esp) + jz 4f + fstp %st(1) + ret + +3: // st(0) is a NaN; st(1) is not. Test if st(0) is signaling. + testb $0x40, 11(%esp) + jz 4f + fstp %st(0) + ret + +4: // Both arguments are NaNs, or one is a signaling NaN. + faddp + ret END(__fmaxl) weak_alias (__fmaxl, fmaxl) diff --git a/sysdeps/i386/fpu/s_fminl.S b/sysdeps/i386/fpu/s_fminl.S index fb5169b8f2..a617e460dc 100644 --- a/sysdeps/i386/fpu/s_fminl.S +++ b/sysdeps/i386/fpu/s_fminl.S @@ -21,23 +21,51 @@ .text ENTRY(__fminl) - fldt 4(%esp) // x - fldt 16(%esp) // x : y + fldt 16(%esp) // y + fxam + fnstsw + fldt 4(%esp) // y : x + + andb $0x45, %ah + cmpb $0x01, %ah + je 2f // y == NaN fxam fnstsw andb $0x45, %ah cmpb $0x01, %ah - je 1f // y == NaN + je 3f // x == NaN fucom %st(1) fnstsw sahf - jc 2f + jc 1f + + fxch %st(1) +1: fstp %st(1) + + ret -1: fxch %st(1) -2: fstp %st(1) +2: // st(1) is a NaN; st(0) may or may not be. + fxam + fnstsw + andb $0x45, %ah + cmpb $0x01, %ah + je 4f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 23(%esp) + jz 4f + fstp %st(1) + ret + +3: // st(0) is a NaN; st(1) is not. Test if st(0) is signaling. + testb $0x40, 11(%esp) + jz 4f + fstp %st(0) + ret +4: // Both arguments are NaNs, or one is a signaling NaN. + faddp ret END(__fminl) weak_alias (__fminl, fminl) diff --git a/sysdeps/i386/i686/fpu/s_fmaxl.S b/sysdeps/i386/i686/fpu/s_fmaxl.S index e5dcd26923..5bade5b5e4 100644 --- a/sysdeps/i386/i686/fpu/s_fmaxl.S +++ b/sysdeps/i386/i686/fpu/s_fmaxl.S @@ -24,16 +24,35 @@ ENTRY(__fmaxl) fldt 4(%esp) // x fldt 16(%esp) // x : y - fucomi %st(0), %st - fcmovu %st(1), %st // now %st contains y if not NaN, x otherwise - - fxch - fucomi %st(1), %st + jp 2f fcmovb %st(1), %st fstp %st(1) ret + +2: // Unordered. + fucomi %st(0), %st + jp 3f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 11(%esp) + jz 4f + fstp %st(1) + ret + +3: // st(0) is a NaN; st(1) may or may not be. + fxch + fucomi %st(0), %st + jp 4f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 23(%esp) + jz 4f + fstp %st(1) + ret + +4: // Both arguments are NaNs, or one is a signaling NaN. + faddp + ret END(__fmaxl) weak_alias (__fmaxl, fmaxl) diff --git a/sysdeps/i386/i686/fpu/s_fminl.S b/sysdeps/i386/i686/fpu/s_fminl.S index ddbd81115e..d586f52b79 100644 --- a/sysdeps/i386/i686/fpu/s_fminl.S +++ b/sysdeps/i386/i686/fpu/s_fminl.S @@ -24,14 +24,35 @@ ENTRY(__fminl) fldt 4(%esp) // x fldt 16(%esp) // x : y - fucomi %st(0), %st - fcmovu %st(1), %st // now %st contains y if not NaN, x otherwise - fucomi %st(1), %st + jp 2f fcmovnb %st(1), %st fstp %st(1) ret + +2: // Unordered. + fucomi %st(0), %st + jp 3f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 11(%esp) + jz 4f + fstp %st(1) + ret + +3: // st(0) is a NaN; st(1) may or may not be. + fxch + fucomi %st(0), %st + jp 4f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 23(%esp) + jz 4f + fstp %st(1) + ret + +4: // Both arguments are NaNs, or one is a signaling NaN. + faddp + ret END(__fminl) weak_alias (__fminl, fminl) diff --git a/sysdeps/x86_64/fpu/s_fmax.S b/sysdeps/x86_64/fpu/s_fmax.S index 02096c0aea..0ff326fb66 100644 --- a/sysdeps/x86_64/fpu/s_fmax.S +++ b/sysdeps/x86_64/fpu/s_fmax.S @@ -27,8 +27,25 @@ ENTRY(__fmax) jmp 2f 1: ucomisd %xmm1, %xmm1 // Is xmm1 a NaN? - jp 2f // then return xmm0 + jp 3f + // xmm0 is a NaN; xmm1 is not. Test if xmm0 is signaling. + movsd %xmm0, -8(%rsp) + testb $0x8, -2(%rsp) + jz 4f movsd %xmm1, %xmm0 // otherwise return xmm1 + ret + +3: // xmm1 is a NaN; xmm0 may or may not be. + ucomisd %xmm0, %xmm0 + jp 4f + // xmm1 is a NaN; xmm0 is not. Test if xmm1 is signaling. + movsd %xmm1, -8(%rsp) + testb $0x8, -2(%rsp) + jz 4f + ret + +4: // Both arguments are NaNs, or one is a signaling NaN. + addsd %xmm1, %xmm0 2: ret END(__fmax) diff --git a/sysdeps/x86_64/fpu/s_fmaxf.S b/sysdeps/x86_64/fpu/s_fmaxf.S index 28e129701e..0f36ee084c 100644 --- a/sysdeps/x86_64/fpu/s_fmaxf.S +++ b/sysdeps/x86_64/fpu/s_fmaxf.S @@ -27,8 +27,25 @@ ENTRY(__fmaxf) jmp 2f 1: ucomiss %xmm1, %xmm1 // Is xmm1 a NaN? - jp 2f // then return xmm0 + jp 3f + // xmm0 is a NaN; xmm1 is not. Test if xmm0 is signaling. + movss %xmm0, -4(%rsp) + testb $0x40, -2(%rsp) + jz 4f movss %xmm1, %xmm0 // otherwise return xmm1 + ret + +3: // xmm1 is a NaN; xmm0 may or may not be. + ucomiss %xmm0, %xmm0 + jp 4f + // xmm1 is a NaN; xmm0 is not. Test if xmm1 is signaling. + movss %xmm1, -4(%rsp) + testb $0x40, -2(%rsp) + jz 4f + ret + +4: // Both arguments are NaNs, or one is a signaling NaN. + addss %xmm1, %xmm0 2: ret END(__fmaxf) diff --git a/sysdeps/x86_64/fpu/s_fmaxl.S b/sysdeps/x86_64/fpu/s_fmaxl.S index f0c2bc0d56..5f0b1e0860 100644 --- a/sysdeps/x86_64/fpu/s_fmaxl.S +++ b/sysdeps/x86_64/fpu/s_fmaxl.S @@ -24,16 +24,35 @@ ENTRY(__fmaxl) fldt 8(%rsp) // x fldt 24(%rsp) // x : y - fucomi %st(0), %st - fcmovu %st(1), %st // now %st contains y if not NaN, x otherwise - - fxch - fucomi %st(1), %st + jp 2f fcmovb %st(1), %st fstp %st(1) ret + +2: // Unordered. + fucomi %st(0), %st + jp 3f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 15(%rsp) + jz 4f + fstp %st(1) + ret + +3: // st(0) is a NaN; st(1) may or may not be. + fxch + fucomi %st(0), %st + jp 4f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 31(%rsp) + jz 4f + fstp %st(1) + ret + +4: // Both arguments are NaNs, or one is a signaling NaN. + faddp + ret END(__fmaxl) weak_alias (__fmaxl, fmaxl) diff --git a/sysdeps/x86_64/fpu/s_fmin.S b/sysdeps/x86_64/fpu/s_fmin.S index fb14e2f3ed..db89befb6b 100644 --- a/sysdeps/x86_64/fpu/s_fmin.S +++ b/sysdeps/x86_64/fpu/s_fmin.S @@ -27,8 +27,25 @@ ENTRY(__fmin) jmp 2f 1: ucomisd %xmm1, %xmm1 // Is xmm1 a NaN? - jp 2f // then return xmm0 + jp 3f + // xmm0 is a NaN; xmm1 is not. Test if xmm0 is signaling. + movsd %xmm0, -8(%rsp) + testb $0x8, -2(%rsp) + jz 4f movsd %xmm1, %xmm0 // otherwise return xmm1 + ret + +3: // xmm1 is a NaN; xmm0 may or may not be. + ucomisd %xmm0, %xmm0 + jp 4f + // xmm1 is a NaN; xmm0 is not. Test if xmm1 is signaling. + movsd %xmm1, -8(%rsp) + testb $0x8, -2(%rsp) + jz 4f + ret + +4: // Both arguments are NaNs, or one is a signaling NaN. + addsd %xmm1, %xmm0 2: ret END(__fmin) diff --git a/sysdeps/x86_64/fpu/s_fminf.S b/sysdeps/x86_64/fpu/s_fminf.S index c8d6d0fd33..41a99787d3 100644 --- a/sysdeps/x86_64/fpu/s_fminf.S +++ b/sysdeps/x86_64/fpu/s_fminf.S @@ -27,8 +27,25 @@ ENTRY(__fminf) jmp 2f 1: ucomiss %xmm1, %xmm1 // Is xmm1 a NaN? - jp 2f // then return xmm0 + jp 3f + // xmm0 is a NaN; xmm1 is not. Test if xmm0 is signaling. + movss %xmm0, -4(%rsp) + testb $0x40, -2(%rsp) + jz 4f movss %xmm1, %xmm0 // otherwise return xmm1 + ret + +3: // xmm1 is a NaN; xmm0 may or may not be. + ucomiss %xmm0, %xmm0 + jp 4f + // xmm1 is a NaN; xmm0 is not. Test if xmm1 is signaling. + movss %xmm1, -4(%rsp) + testb $0x40, -2(%rsp) + jz 4f + ret + +4: // Both arguments are NaNs, or one is a signaling NaN. + addss %xmm1, %xmm0 2: ret END(__fminf) diff --git a/sysdeps/x86_64/fpu/s_fminl.S b/sysdeps/x86_64/fpu/s_fminl.S index f1a06d29d7..12fc3fb06c 100644 --- a/sysdeps/x86_64/fpu/s_fminl.S +++ b/sysdeps/x86_64/fpu/s_fminl.S @@ -24,14 +24,35 @@ ENTRY(__fminl) fldt 8(%rsp) // x fldt 24(%rsp) // x : y - fucomi %st(0), %st - fcmovu %st(1), %st // now %st contains y if not NaN, x otherwise - fucomi %st(1), %st + jp 2f fcmovnb %st(1), %st fstp %st(1) ret + +2: // Unordered. + fucomi %st(0), %st + jp 3f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 15(%rsp) + jz 4f + fstp %st(1) + ret + +3: // st(0) is a NaN; st(1) may or may not be. + fxch + fucomi %st(0), %st + jp 4f + // st(1) is a NaN; st(0) is not. Test if st(1) is signaling. + testb $0x40, 31(%rsp) + jz 4f + fstp %st(1) + ret + +4: // Both arguments are NaNs, or one is a signaling NaN. + faddp + ret END(__fminl) weak_alias (__fminl, fminl) |