[PATCH 00/12] Improve rounding to interger function for C23

public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed

* [PATCH 00/12] Improve rounding to interger function for C23
@ 2023-12-28 17:20 Adhemerval Zanella
  2023-12-28 17:20 ` [PATCH 01/12] math: Reformat Makefile Adhemerval Zanella
                   ` (12 more replies)
  0 siblings, 13 replies; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

As indicated by GCC documentation [1], ISO C23 does not allow that C
bindings ceil, floor, round, and trunc (in all floating point formats)
to raise 
inexact exceptions (different than ISO C99/C11 where this is allowed).

A recent MIPS patch to used some arch-specific instructions raised this
issue [1] and it was not caught because there was no proper testing. By
adding the missing tests, some implementations do indeed raise inexact
exceptions. 

The generic implementation all uses integer operation, so they are not
subject to this issue. The powerpc (for power4 and lower) and the riscv
avoid the inexact exception by disabling/enabling exceptions. The x86
uses some arch-specific implementation for long double and on i386 (due
to the use of x87 instruction).

Instead of adding newer symbols depending on the required standard
version, the patchset adapts the faulty ones to avoid raising the
inexact exception. The x86 version already saves/restore the floating
point status, so I think it is unlikely the patch would yield much
performance difference (I did not do any performance analysis on whether
a generic implementation would yield better performance).

I checked on powerpc, powerpc64, aarch64, armhf, x86, and did some
regression checks on riscv.

[1] https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fno-fp-int-builtin-inexact
[2] https://sourceware.org/pipermail/libc-alpha/2023-December/153528.html

Adhemerval Zanella (12):
  math: Reformat Makefile.
  powerpc: Add missing arch flags on rounding ifunc variants
  math: Add test to check if ceil raise inexact floating-point exception
  math: Add test to check if floor raise inexact floating-point
    exception
  math: Add test to check if trunc raise inexact floating-point
    exception
  math: Add test to check if round raise inexact floating-point
    exception
  x86: Do not raise inexact exception on ceill
  x86: Do not raise inexact exception on floorl
  x86: Do not raise inexact exception on truncl
  x86: Do not raise inexact exception on floor/floorf
  i386: Do not raise inexact exception on ceil/ceilf
  i386: Do not raise inexact exception on trunc/truncf

 math/Makefile                                 | 861 ++++++++++++++----
 math/test-ceil-except-2.c                     |  67 ++
 math/test-ceil-except.c                       |  85 ++
 math/test-floor-except-2.c                    |  67 ++
 math/test-floor-except.c                      |  85 ++
 math/test-round-except-2.c                    |  67 ++
 math/test-round-except.c                      |  85 ++
 math/test-trunc-except-2.c                    |  67 ++
 math/test-trunc-except.c                      |  85 ++
 sysdeps/i386/fpu/s_ceil.S                     |  34 -
 sysdeps/i386/fpu/s_ceil.c                     |  38 +
 sysdeps/i386/fpu/s_ceilf.S                    |  34 -
 sysdeps/i386/fpu/s_ceilf.c                    |  38 +
 sysdeps/i386/fpu/s_ceill.S                    |  39 -
 sysdeps/i386/fpu/s_floor.S                    |  34 -
 sysdeps/i386/fpu/s_floor.c                    |  38 +
 sysdeps/i386/fpu/s_floorf.S                   |  34 -
 sysdeps/i386/fpu/s_floorf.c                   |  38 +
 sysdeps/i386/fpu/s_floorl.S                   |  39 -
 sysdeps/i386/fpu/{s_trunc.S => s_trunc.c}     |  37 +-
 sysdeps/i386/fpu/{s_truncf.S => s_truncf.c}   |  37 +-
 .../powerpc32/power4/fpu/multiarch/Makefile   |   6 +
 .../fpu/s_truncl.S => x86/fpu/s_ceill.c}      |  38 +-
 sysdeps/x86/fpu/s_floorl.c                    |  38 +
 .../fpu/s_truncl.S => x86/fpu/s_truncl.c}     |  40 +-
 sysdeps/x86_64/fpu/s_ceill.S                  |  34 -
 sysdeps/x86_64/fpu/s_floorl.S                 |  33 -
 27 files changed, 1583 insertions(+), 515 deletions(-)
 create mode 100644 math/test-ceil-except-2.c
 create mode 100644 math/test-ceil-except.c
 create mode 100644 math/test-floor-except-2.c
 create mode 100644 math/test-floor-except.c
 create mode 100644 math/test-round-except-2.c
 create mode 100644 math/test-round-except.c
 create mode 100644 math/test-trunc-except-2.c
 create mode 100644 math/test-trunc-except.c
 delete mode 100644 sysdeps/i386/fpu/s_ceil.S
 create mode 100644 sysdeps/i386/fpu/s_ceil.c
 delete mode 100644 sysdeps/i386/fpu/s_ceilf.S
 create mode 100644 sysdeps/i386/fpu/s_ceilf.c
 delete mode 100644 sysdeps/i386/fpu/s_ceill.S
 delete mode 100644 sysdeps/i386/fpu/s_floor.S
 create mode 100644 sysdeps/i386/fpu/s_floor.c
 delete mode 100644 sysdeps/i386/fpu/s_floorf.S
 create mode 100644 sysdeps/i386/fpu/s_floorf.c
 delete mode 100644 sysdeps/i386/fpu/s_floorl.S
 rename sysdeps/i386/fpu/{s_trunc.S => s_trunc.c} (61%)
 rename sysdeps/i386/fpu/{s_truncf.S => s_truncf.c} (61%)
 rename sysdeps/{x86_64/fpu/s_truncl.S => x86/fpu/s_ceill.c} (57%)
 create mode 100644 sysdeps/x86/fpu/s_floorl.c
 rename sysdeps/{i386/fpu/s_truncl.S => x86/fpu/s_truncl.c} (61%)
 delete mode 100644 sysdeps/x86_64/fpu/s_ceill.S
 delete mode 100644 sysdeps/x86_64/fpu/s_floorl.S

-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 01/12] math: Reformat Makefile.
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 13:57   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 02/12] powerpc: Add missing arch flags on rounding ifunc variants Adhemerval Zanella
                   ` (11 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

Reflow all long lines adding comment terminators.
Sort all reflowed text using scripts/sort-makefile-lines.py.

No code generation changes observed in binary artifacts.
No regressions on x86_64 and i686.
---
 math/Makefile | 844 ++++++++++++++++++++++++++++++++++++++++----------
 1 file changed, 685 insertions(+), 159 deletions(-)

diff --git a/math/Makefile b/math/Makefile
index a9daae09de..5a912c55bd 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -22,87 +22,279 @@ subdir		:= math
 include ../Makeconfig
 
 # Installed header files.
-headers		:= math.h bits/mathcalls.h \
-		   fpu_control.h complex.h bits/cmathcalls.h fenv.h \
-		   bits/fenv.h bits/mathdef.h tgmath.h \
-		   bits/math-vector.h finclude/math-vector-fortran.h \
-		   bits/libm-simd-decl-stubs.h bits/iscanonical.h \
-		   bits/flt-eval-method.h bits/fp-fast.h bits/fp-logb.h \
-		   bits/long-double.h bits/mathcalls-helper-functions.h \
-		   bits/floatn.h bits/floatn-common.h bits/mathcalls-narrow.h
+headers	:= \
+  math.h \
+  bits/mathcalls.h \
+  fpu_control.h \
+  complex.h \
+  bits/cmathcalls.h \
+  fenv.h \
+  bits/fenv.h \
+  bits/mathdef.h \
+  tgmath.h \
+  bits/math-vector.h \
+  finclude/math-vector-fortran.h \
+  bits/libm-simd-decl-stubs.h \
+  bits/iscanonical.h \
+  bits/flt-eval-method.h \
+  bits/fp-fast.h \
+  bits/fp-logb.h \
+  bits/long-double.h \
+  bits/mathcalls-helper-functions.h \
+  bits/floatn.h \
+  bits/floatn-common.h \
+  bits/mathcalls-narrow.h
+  # headers
 
 # FPU support code.
-aux		:= setfpucw fpu_control
+aux := \
+  fpu_control \
+  setfpucw \
+  # aux
 
 # Build the -lm library.
 
 extra-libs	:= libm
 extra-libs-others = $(extra-libs)
 
-libm-support = s_lib_version s_matherr s_signgam			\
-	       fclrexcpt fgetexcptflg fraiseexcpt fsetexcptflg		\
-	       ftestexcept fegetround fesetround fegetenv feholdexcpt	\
-	       fesetenv feupdateenv fedisblxcpt feenablxcpt	\
-	       fegetexcept fesetexcept fetestexceptflag fegetmode	\
-	       fesetmode
+libm-support = \
+  fclrexcpt \
+  fedisblxcpt \
+  feenablxcpt \
+  fegetenv \
+  fegetexcept \
+  fegetmode \
+  fegetround \
+  feholdexcpt \
+  fesetenv \
+  fesetexcept \
+  fesetmode\
+  fesetround \
+  fetestexceptflag \
+  feupdateenv \
+  fgetexcptflg \
+  fraiseexcpt \
+  fsetexcptflg \
+  ftestexcept \
+  s_lib_version \
+  s_matherr \
+  s_signgam \
+  # libm-support
 
 # Wrappers for these functions generated per type using a file named
 # <func>_template.c and the appropriate math-type-macros-<TYPE>.h.
-gen-libm-calls = cargF conjF cimagF crealF cabsF e_scalbF s_cacosF	  \
-	         s_cacoshF s_ccosF s_ccoshF s_casinF s_csinF s_casinhF	  \
-		 k_casinhF s_csinhF k_casinhF s_csinhF s_catanhF s_catanF \
-		 s_ctanF s_ctanhF s_cexpF s_clogF s_cprojF s_csqrtF	  \
-		 s_cpowF s_clog10F s_fdimF s_nextdownF s_fmaxF s_fminF	  \
-		 s_nanF s_iseqsigF s_canonicalizeF s_significandF	  \
-		 w_ilogbF w_llogbF					  \
-		 w_log1pF w_scalblnF s_fmaxmagF s_fminmagF w_acosF	  \
-		 w_acoshF w_asinF w_atan2F w_atanhF w_coshF w_exp10F	  \
-		 w_exp2F w_fmodF w_hypotF w_j0F w_j1F w_jnF w_logF	  \
-		 w_log10F w_log2F w_powF w_remainderF w_scalbF		  \
-		 w_sinhF w_sqrtF					  \
-		 w_tgammaF w_lgammaF w_lgammaF_r w_expF e_exp2F		  \
-		 s_fmaximumF s_fmaximum_magF s_fmaximum_numF		  \
-		 s_fmaximum_mag_numF s_fminimumF s_fminimum_magF	  \
-		 s_fminimum_numF s_fminimum_mag_numF
-
-libm-calls =								  \
-	e_acosF e_acoshF e_asinF e_atan2F e_atanhF e_coshF e_expF e_fmodF \
-	e_hypotF e_j0F e_j1F e_jnF e_lgammaF_r e_logF e_log10F e_powF	  \
-	e_remainderF e_sinhF e_sqrtF e_gammaF_r				  \
-	e_ilogbF							  \
-	k_tanF s_asinhF s_atanF s_cbrtF					  \
-	s_ceilF s_cosF s_erfF s_expm1F s_fabsF				  \
-	s_floorF s_log1pF s_logbF				  \
-	s_nextafterF s_nexttowardF s_rintF s_scalblnF			  \
-	s_sinF s_tanF s_tanhF						  \
-	s_fpclassifyF s_truncF						  \
-	s_remquoF e_log2F s_roundF s_nearbyintF s_sincosF		  \
-	s_fmaF s_lrintF s_llrintF s_lroundF s_llroundF e_exp10F		  \
-	s_issignalingF $(calls:s_%=m_%) x2y2m1F				  \
-	gamma_productF lgamma_negF lgamma_productF			  \
-	s_nextupF s_totalorderF s_totalordermagF s_getpayloadF		  \
-	s_setpayloadF s_setpayloadsigF s_roundevenF s_fromfpF s_ufromfpF  \
-	s_fromfpxF s_ufromfpxF $(gen-libm-calls)
-
-libm-compat-calls =							\
-	w_acosF_compat w_acoshF_compat w_asinF_compat w_atan2F_compat	\
-	w_atanhF_compat w_coshF_compat w_exp2F_compat w_exp10F_compat	\
-	w_fmodF_compat w_hypotF_compat w_j0F_compat w_j1F_compat	\
-	w_jnF_compat w_log2F_compat w_log10F_compat w_logF_compat	\
-	w_powF_compat w_remainderF_compat w_scalbF_compat		\
-	w_sinhF_compat w_sqrtF_compat w_tgammaF_compat			\
-	w_lgammaF_r_compat w_lgammaF_compat2 w_expF_compat		\
-	w_lgamma_compatF k_standardF
-
-libm-narrow-fns = add div fma mul sqrt sub
-libm-narrow-types-basic = s_fF s_f32xFf64
-libm-narrow-types-ldouble-yes = s_fFl s_dFl
-libm-narrow-types-float128-yes = s_f32Ff128 s_f64Ff128 s_f64xFf128
-libm-narrow-types-float128-alias-yes = s_f64xFf128
-libm-narrow-types = $(libm-narrow-types-basic) \
-		    $(libm-narrow-types-ldouble-$(long-double-fcts)) \
-		    $(libm-narrow-types-float128-$(float128-fcts)) \
-		    $(libm-narrow-types-float128-alias-$(float128-alias-fcts))
+gen-libm-calls = \
+  cabsF \
+  cargF \
+  cimagF \
+  conjF \
+  crealF \
+  e_exp2F \
+  e_scalbF \
+  k_casinhF \
+  k_casinhF \
+  s_cacosF  \
+  s_cacoshF \
+  s_canonicalizeF \
+  s_casinF \
+  s_casinhF \
+  s_catanF \
+  s_catanhF \
+  s_ccosF \
+  s_ccoshF \
+  s_cexpF \
+  s_clog10F \
+  s_clogF \
+  s_cpowF \
+  s_cprojF \
+  s_csinF \
+  s_csinhF \
+  s_csinhF \
+  s_csqrtF \
+  s_ctanF \
+  s_ctanhF \
+  s_fdimF \
+  s_fmaxF \
+  s_fmaximumF \
+  s_fmaximum_magF \
+  s_fmaximum_mag_numF \
+  s_fmaximum_numF \
+  s_fmaxmagF \
+  s_fminF \
+  s_fminimumF \
+  s_fminimum_magF \
+  s_fminimum_mag_numF \
+  s_fminimum_numF \
+  s_fminmagF \
+  s_iseqsigF \
+  s_nanF \
+  s_nextdownF \
+  s_significandF \
+  w_acosF \
+  w_acoshF \
+  w_asinF \
+  w_atan2F \
+  w_atanhF \
+  w_coshF \
+  w_exp10F \
+  w_exp2F \
+  w_expF \
+  w_fmodF \
+  w_hypotF \
+  w_ilogbF \
+  w_j0F \
+  w_j1F \
+  w_jnF \
+  w_lgammaF \
+  w_lgammaF_r \
+  w_llogbF \
+  w_log10F \
+  w_log1pF \
+  w_log2F \
+  w_logF  \
+  w_powF \
+  w_remainderF \
+  w_scalbF \
+  w_scalblnF \
+  w_sinhF \
+  w_sqrtF \
+  w_tgammaF \
+  # gen-libm-calls
+
+libm-calls = \
+  $(calls:s_%=m_%) \
+  $(gen-libm-calls) \
+  e_acosF \
+  e_acoshF \
+  e_asinF \
+  e_atan2F \
+  e_atanhF \
+  e_coshF \
+  e_exp10F \
+  e_expF \
+  e_fmodF \
+  e_gammaF_r \
+  e_hypotF \
+  e_ilogbF \
+  e_j0F \
+  e_j1F \
+  e_jnF \
+  e_lgammaF_r \
+  e_log10F \
+  e_log2F \
+  e_logF \
+  e_powF \
+  e_remainderF \
+  e_sinhF \
+  e_sqrtF \
+  gamma_productF \
+  k_tanF \
+  lgamma_negF \
+  lgamma_productF \
+  s_asinhF \
+  s_atanF \
+  s_cbrtF \
+  s_ceilF \
+  s_cosF \
+  s_erfF \
+  s_expm1F \
+  s_fabsF \
+  s_floorF \
+  s_fmaF \
+  s_fpclassifyF \
+  s_fromfpF \
+  s_fromfpxF \
+  s_getpayloadF \
+  s_issignalingF \
+  s_llrintF \
+  s_llroundF \
+  s_log1pF \
+  s_logbF \
+  s_lrintF \
+  s_lroundF \
+  s_nearbyintF \
+  s_nextafterF \
+  s_nexttowardF \
+  s_nextupF \
+  s_remquoF \
+  s_rintF \
+  s_roundF \
+  s_roundevenF \
+  s_scalblnF \
+  s_setpayloadF \
+  s_setpayloadsigF \
+  s_sinF \
+  s_sincosF \
+  s_tanF \
+  s_tanhF \
+  s_totalorderF \
+  s_totalordermagF \
+  s_truncF \
+  s_ufromfpF \
+  s_ufromfpxF \
+  x2y2m1F \
+  # libm-calls
+
+libm-compat-calls = \
+  k_standardF \
+  w_acosF_compat \
+  w_acoshF_compat \
+  w_asinF_compat \
+  w_atan2F_compat \
+  w_atanhF_compat \
+  w_coshF_compat \
+  w_exp10F_compat \
+  w_exp2F_compat \
+  w_expF_compat	\
+  w_fmodF_compat \
+  w_hypotF_compat \
+  w_j0F_compat \
+  w_j1F_compat\
+  w_jnF_compat \
+  w_lgammaF_compat2 \
+  w_lgammaF_r_compat \
+  w_lgamma_compatF \
+  w_log10F_compat \
+  w_log2F_compat \
+  w_logF_compat	\
+  w_powF_compat \
+  w_remainderF_compat \
+  w_scalbF_compat \
+  w_sinhF_compat \
+  w_sqrtF_compat \
+  w_tgammaF_compat \
+  # libm-compat-calls
+
+libm-narrow-fns = \
+  add \
+  div \
+  fma \
+  mul \
+  sqrt \
+  sub \
+  # libm-narrow-fns
+libm-narrow-types-basic = \
+  s_f32xFf64 \
+  s_fF \
+  # libm-narrow-types-basic
+libm-narrow-types-ldouble-yes = \
+  s_dFl \
+  s_fFl \
+  # libm-narrow-types-ldouble-yes
+libm-narrow-types-float128-yes = \
+  s_f32Ff128 \
+  s_f64Ff128 \
+  s_f64xFf128 \
+  # libm-narrow-types-float128-yes
+libm-narrow-types-float128-alias-yes = \
+  s_f64xFf128 \
+  # libm-narrow-types-float128-alias-yes
+libm-narrow-types = \
+  $(libm-narrow-types-basic) \
+  $(libm-narrow-types-float128-$(float128-fcts)) \
+  $(libm-narrow-types-float128-alias-$(float128-alias-fcts)) \
+  $(libm-narrow-types-ldouble-$(long-double-fcts)) \
+  # libm-narrow-types
 
 # Type specific routine support.
 #
@@ -115,33 +307,65 @@ libm-narrow-types = $(libm-narrow-types-basic) \
 # Finally, note that types is an intentionally recursive variable.
 # We only know the full set of supported types for the target machine
 # after the Rules makefile has been parsed.
-types-basic = $(type-ldouble-$(long-double-fcts)) double float
+types-basic = \
+  $(type-ldouble-$(long-double-fcts)) \
+  double \
+  float \
+  # types-basic
 
 # Like types, but includes types whose functions alias those for
 # another type.
-test-types-basic = ldouble double float
+test-types-basic = \
+  double \
+  float \
+  ldouble \
+  # test-types-basic
 
 # long double support
 type-ldouble-suffix := l
-type-ldouble-routines := t_sincosl k_sinl k_cosl k_sincosl s_iscanonicall \
-			 e_rem_pio2l
+type-ldouble-routines := \
+  e_rem_pio2l \
+  k_cosl \
+  k_sincosl \
+  k_sinl \
+  s_iscanonicall \
+  t_sincosl \
+  # type-ldouble-routines
 type-ldouble-yes := ldouble
 
 # double support
 type-double-suffix :=
-type-double-routines := branred k_rem_pio2 	\
-		       sincostab math_err e_exp_data e_log_data	\
-		       e_log2_data e_pow_log_data
+type-double-routines := \
+  branred \
+  e_exp_data \
+  e_log2_data \
+  e_log_data \
+  e_pow_log_data \
+  k_rem_pio2 \
+  math_err \
+  sincostab \
+  # type-double-routines
 
 # float support
 type-float-suffix := f
-type-float-routines := math_errf e_exp2f_data e_logf_data	\
-		       e_log2f_data e_powf_log2_data s_sincosf_data
+type-float-routines := \
+  e_exp2f_data \
+  e_log2f_data \
+  e_logf_data \
+  e_powf_log2_data \
+  math_errf \
+  s_sincosf_data \
+  # type-float-routines
 
 # _Float128 support
 type-float128-suffix := f128
-type-float128-routines := t_sincosf128 k_sinf128 k_cosf128 k_sincosf128 \
-			  e_rem_pio2f128
+type-float128-routines := \
+  e_rem_pio2f128 \
+  k_cosf128 \
+  k_sincosf128 \
+  k_sinf128 \
+  t_sincosf128 \
+  # type-float128-routines
 type-float128-yes := float128
 
 # _Float64x may be supported, only as an alias type.
@@ -151,11 +375,20 @@ type-float64x-yes := float64x
 type-ibm128-suffix := l
 type-ibm128-yes := ibm128
 
-types = $(types-basic) $(type-float128-$(float128-fcts))
-test-types = $(test-types-basic) $(type-float128-$(float128-fcts)) \
-	     float32 float64 $(type-float128-$(float128-alias-fcts)) \
-	     float32x $(type-float64x-$(float64x-alias-fcts)) \
-	     $(type-ibm128-$(ibm128-fcts))
+types = \
+  $(type-float128-$(float128-fcts)) \
+  $(types-basic) \
+  # types
+test-types = \
+  $(test-types-basic) \
+  $(type-float128-$(float128-alias-fcts)) \
+  $(type-float128-$(float128-fcts)) \
+  $(type-float64x-$(float64x-alias-fcts)) \
+  $(type-ibm128-$(ibm128-fcts)) \
+  float32 \
+  float32x \
+  float64 \
+  # test-types
 
 # Pairs of types for which narrowing functions should be tested (this
 # variable has more entries than libm-narrow-types because it includes
@@ -163,14 +396,25 @@ test-types = $(test-types-basic) $(type-float128-$(float128-fcts)) \
 # for other types). This definition embeds the assumption that if
 # _Float64x is supported, so is _Float128, and vice versa (they may or
 # may not have the same format).
-test-type-pairs = float-double float-ldouble double-ldouble \
-		  float32-float64 float32-float32x float32x-float64 \
-		  $(test-type-pairs-f64xf128-$(float128-fcts)) \
-		  $(test-type-pairs-f64xf128-$(float128-alias-fcts))
-test-type-pairs-f64xf128-yes = float32-float64x float32-float128 \
-			       float64-float64x float64-float128 \
-			       float32x-float64x float32x-float128 \
-			       float64x-float128
+test-type-pairs = \
+  $(test-type-pairs-f64xf128-$(float128-alias-fcts)) \
+  $(test-type-pairs-f64xf128-$(float128-fcts)) \
+  double-ldouble \
+  float-double \
+  float-ldouble \
+  float32-float32x \
+  float32-float64 \
+  float32x-float64 \
+  # test-type-pairs
+test-type-pairs-f64xf128-yes = \
+  float32-float128 \
+  float32-float64x \
+  float32x-float128 \
+  float32x-float64x \
+  float64-float128 \
+  float64-float64x \
+  float64x-float128 \
+  # test-type-pairs-f64xf128-yes
 
 # For each of the basic types (float, double, long double), replace the
 # occurrences of 'F' in arg 1 with the appropriate suffix for the type.
@@ -193,8 +437,17 @@ libm-routines = $(strip $(libm-support)					\
 # and we don't want to have to link every program with -lm.
 # In libm-calls (above), list m_foo in place of s_foo for any
 # routine that should be compiled separately for its libc and libm versions.
-calls = s_isinfF s_isnanF s_finiteF s_copysignF s_modfF s_scalbnF s_frexpF \
-	s_signbitF $(gen-calls)
+calls = \
+  $(gen-calls) \
+  s_copysignF \
+  s_finiteF \
+  s_frexpF \
+  s_isinfF \
+  s_isnanF \
+  s_modfF \
+  s_scalbnF \
+  s_signbitF \
+  # calls
 gen-calls = s_ldexpF
 generated += $(foreach s,.c .S,$(call type-foreach, $(calls:s_%=m_%$(s))))
 routines = $(call type-foreach, $(calls))
@@ -205,8 +458,14 @@ libm-shared-only-routines = $(call type-foreach, $(calls:s_%=m_%))
 ifeq ($(build-mathvec),yes)
 # We need to install libm.so and libm.a as linker scripts
 # for transparent use of vector math library.
-install-lib-ldscripts := libm.so libm.a
-install-others = $(inst_libdir)/libm.so $(inst_libdir)/libm.a
+install-lib-ldscripts := \
+  libm.a \
+  libm.so \
+  # install-lib-ldscripts
+install-others = \
+  $(inst_libdir)/libm.a \
+  $(inst_libdir)/libm.so \
+  # install-others
 $(inst_libdir)/libm.so: $(common-objpfx)format.lds \
 			$(libm) \
 			$(libmvec) \
@@ -234,42 +493,95 @@ $(inst_libdir)/libm.a: $(common-objpfx)format.lds \
 endif
 
 # Rules for the test suite.
-tests = test-matherr-3 test-fenv basic-test \
-	test-misc test-fpucw test-fpucw-ieee tst-definitions test-tgmath \
-	test-tgmath-ret bug-nextafter bug-nexttoward bug-tgmath1 \
-	test-tgmath-int test-tgmath2 test-powl tst-CMPLX tst-CMPLX2 test-snan \
-	test-fenv-tls test-fenv-preserve test-fenv-return \
-	test-nearbyint-except test-fenv-clear \
-	test-nearbyint-except-2 test-signgam-uchar test-signgam-uchar-init \
-	test-signgam-uint test-signgam-uint-init test-signgam-ullong \
-	test-signgam-ullong-init test-nan-overflow test-nan-payload \
-	test-fexcept test-fexcept-traps test-fesetexcept \
-	test-fesetexcept-traps test-fetestexceptflag test-femode \
-	test-femode-traps test-iszero-excess-precision \
-	test-iseqsig-excess-precision test-flt-eval-method \
-	test-fp-ilogb-constants test-fp-llogb-constants \
-	test-narrow-macros \
-	test-nan-const $(tests-static)
+tests = \
+  $(tests-static) \
+  bug-nextafter \
+  bug-nexttoward \
+  bug-tgmath1 \
+  test-femode \
+  test-femode-traps \
+  test-fenv basic-test \
+  test-fenv-clear \
+  test-fenv-preserve \
+  test-fenv-return \
+  test-fenv-tls \
+  test-fesetexcept \
+  test-fesetexcept-traps \
+  test-fetestexceptflag \
+  test-fexcept \
+  test-fexcept-traps \
+  test-flt-eval-method \
+  test-fp-ilogb-constants \
+  test-fp-llogb-constants \
+  test-fpucw \
+  test-fpucw-ieee \
+  test-iseqsig-excess-precision \
+  test-iszero-excess-precision \
+  test-matherr-3 \
+  test-misc \
+  test-nan-const \
+  test-nan-overflow \
+  test-nan-payload \
+  test-narrow-macros \
+  test-nearbyint-except \
+  test-nearbyint-except-2 \
+  test-powl \
+  test-signgam-uchar \
+  test-signgam-uchar-init \
+  test-signgam-uint \
+  test-signgam-uint-init \
+  test-signgam-ullong \
+  test-signgam-ullong-init \
+  test-snan \
+  test-tgmath \
+  test-tgmath-int \
+  test-tgmath-ret \
+  test-tgmath2 \
+  tst-CMPLX \
+  tst-CMPLX2 \
+  tst-definitions \
+  # tests
 ifneq ($(config-cflags-signaling-nans),)
-tests += test-fe-snans-always-signal
+tests += \
+  test-fe-snans-always-signal \
+  # tests
 endif
-tests-static = test-fpucw-static test-fpucw-ieee-static \
-	       test-signgam-uchar-static test-signgam-uchar-init-static \
-	       test-signgam-uint-static test-signgam-uint-init-static \
-	       test-signgam-ullong-static test-signgam-ullong-init-static
+tests-static = \
+  test-fpucw-ieee-static \
+  test-fpucw-static \
+  test-signgam-uchar-init-static \
+  test-signgam-uchar-static \
+  test-signgam-uint-init-static \
+  test-signgam-uint-static \
+  test-signgam-ullong-init-static \
+  test-signgam-ullong-static \
+  # tests-static
 
 # The tested symbols matherr, _LIB_VERSION have been removed in glibc 2.27.
 ifeq ($(have-GLIBC_2.26)$(build-shared),yesyes)
-tests += test-matherr test-matherr-2
+tests += \
+  test-matherr \
+  test-matherr-2 \
+  # tests
 endif
 
 # These tests use internal (unexported) GMP functions and are linked
 # statically to obtain access to these functions.
-tests-static += atest-exp atest-sincos atest-exp2
+tests-static += \
+  atest-exp \
+  atest-exp2 \
+  atest-sincos \
+  # tests-static
 
 ifneq (,$(CXX))
-tests += test-math-isinff test-math-iszero test-math-issignaling \
-	 test-math-iscanonical test-math-cxx11 test-math-iseqsig
+tests += \
+  test-math-cxx11 \
+  test-math-iscanonical \
+  test-math-iseqsig \
+  test-math-isinff \
+  test-math-issignaling \
+  test-math-iszero \
+  # tests
 endif
 
 libm-vec-tests = $(addprefix test-,$(libmvec-tests))
@@ -286,35 +598,167 @@ $(objpfx)libm-test-ulps.h: $(ulps-file) gen-libm-test.py
 	$(make-target-directory)
 	$(PYTHON) gen-libm-test.py -u $< -H $@
 
-libm-test-funcs-auto = acos acosh asin asinh atan atan2 atanh cabs cacos \
-		       cacosh carg casin casinh catan catanh cbrt ccos ccosh \
-		       cexp clog clog10 cos cosh cpow csin csinh csqrt ctan \
-		       ctanh erf erfc exp exp10 exp2 expm1 fma hypot j0 j1 jn \
-		       lgamma log log10 log1p log2 pow sin sincos sinh sqrt \
-		       tan tanh tgamma y0 y1 yn
-libm-test-funcs-noauto = canonicalize ceil cimag conj copysign cproj creal \
-			 fabs fdim floor fmax fmaximum fmaximum_mag \
-			 fmaximum_mag_num fmaximum_num fmaxmag fmin fminimum \
-			 fminimum_mag fminimum_mag_num fminimum_num fminmag \
-			 fmod fpclassify frexp fromfp fromfpx getpayload \
-			 ilogb iscanonical iseqsig isfinite isgreater \
-			 isgreaterequal isinf isless islessequal \
-			 islessgreater isnan isnormal issignaling issubnormal \
-			 isunordered iszero llogb llrint llround logb lrint \
-			 lround modf nearbyint nextafter nextdown nexttoward \
-			 nextup remainder remquo rint round roundeven scalb \
-			 scalbln scalbn setpayload setpayloadsig signbit \
-			 significand totalorder totalordermag trunc ufromfp \
-			 ufromfpx compat_totalorder compat_totalordermag
-libm-test-funcs-compat = compat_totalorder compat_totalordermag
-libm-test-funcs-narrow = add div fma mul sqrt sub
-libm-test-funcs-all = $(libm-test-funcs-auto) $(libm-test-funcs-noauto)
+libm-test-funcs-auto = \
+  acos \
+  acosh \
+  asin \
+  asinh \
+  atan \
+  atan2 \
+  atanh \
+  cabs \
+  cacos \
+  cacosh \
+  carg \
+  casin \
+  casinh \
+  catan \
+  catanh \
+  cbrt \
+  ccos \
+  ccosh \
+  cexp \
+  clog \
+  clog10 \
+  cos \
+  cosh \
+  cpow \
+  csin \
+  csinh \
+  csqrt \
+  ctan \
+  ctanh \
+  erf \
+  erfc \
+  exp \
+  exp2 \
+  exp10 \
+  expm1 \
+  fma \
+  hypot \
+  j0 \
+  j1 \
+  jn \
+  lgamma \
+  log \
+  log10 \
+  log1p \
+  log2 \
+  pow \
+  sin \
+  sincos \
+  sinh \
+  sqrt \
+  tan \
+  tanh \
+  tgamma \
+  y0 \
+  y1 \
+  yn \
+  # libm-test-funcs-auto
+libm-test-funcs-noauto = \
+  canonicalize \
+  ceil \
+  cimag \
+  compat_totalorder \
+  compat_totalordermag \
+  conj \
+  copysign \
+  cproj \
+  creal \
+  fabs \
+  fdim \
+  floor \
+  fmax \
+  fmaximum \
+  fmaximum_mag \
+  fmaximum_mag_num \
+  fmaximum_num \
+  fmaxmag \
+  fmin \
+  fminimum \
+  fminimum_mag \
+  fminimum_mag_num \
+  fminimum_num \
+  fminmag \
+  fmod \
+  fpclassify \
+  frexp \
+  fromfp \
+  fromfpx \
+  getpayload \
+  ilogb \
+  iscanonical \
+  iseqsig \
+  isfinite \
+  isgreater \
+  isgreaterequal \
+  isinf \
+  isless \
+  islessequal \
+  islessgreater \
+  isnan \
+  isnormal \
+  issignaling \
+  issubnormal \
+  isunordered \
+  iszero \
+  llogb \
+  llrint \
+  llround \
+  logb \
+  lrint \
+  lround \
+  modf \
+  nearbyint \
+  nextafter \
+  nextdown \
+  nexttoward \
+  nextup \
+  remainder \
+  remquo \
+  rint \
+  round \
+  roundeven \
+  scalb \
+  scalbln \
+  scalbn \
+  setpayload \
+  setpayloadsig \
+  signbit \
+  significand \
+  totalorder \
+  totalordermag \
+  trunc \
+  ufromfp \
+  ufromfpx \
+  # libm-test-funcs-noauto
+libm-test-funcs-compat = \
+  compat_totalorder \
+  compat_totalordermag \
+  # libm-test-funcs-compat
+libm-test-funcs-narrow = \
+  add \
+  div \
+  fma \
+  mul \
+  sqrt \
+  sub \
+  # libm-test-funcs-narrow
+libm-test-funcs-all = \
+  $(libm-test-funcs-auto) \
+  $(libm-test-funcs-noauto) \
+  # libm-test-funcs-all
 libm-test-c-auto = $(foreach f,$(libm-test-funcs-auto),libm-test-$(f).c)
 libm-test-c-noauto = $(foreach f,$(libm-test-funcs-noauto),libm-test-$(f).c)
 libm-test-c-narrow = $(foreach f,$(libm-test-funcs-narrow),\
 				 libm-test-narrow-$(f).c)
-generated += libm-test-ulps.h $(libm-test-c-auto) $(libm-test-c-noauto) \
-	     $(libm-test-c-narrow)
+generated += \
+  $(libm-test-c-auto) \
+  $(libm-test-c-narrow) \
+  $(libm-test-c-noauto) \
+  libm-test-ulps.h \
+  # generated
 
 libm-tests-base-normal = $(foreach t,$(test-types),test-$(t))
 libm-tests-base-narrow = $(foreach t,$(test-type-pairs),test-$(t))
@@ -374,19 +818,101 @@ $(foreach t,$(libm-tests-base),\
 	    $(objpfx)$(t)-compat_totalordermag.o): $(objpfx)libm-test-totalordermag.c
 
 # _Float128x omitted as not supported by gen-tgmath-tests.py.
-tgmath3-narrow-types = f d f16 f32 f64 f128 f32x f64x
+tgmath3-narrow-types = \
+  d \
+  f \
+  f16 \
+  f32 \
+  f128 \
+  f32x \
+  f64 \
+  f64x \
+  # tgmath3-narrow-types
 tgmath3-narrow-macros = $(foreach t,$(tgmath3-narrow-types), \
 				    $(foreach f,$(libm-narrow-fns),$(t)$(f)))
-tgmath3-macros = atan2 cbrt ceil copysign erf erfc exp10 exp2 expm1 fdim \
-		 floor fma fmax fmin fmod frexp hypot ilogb ldexp lgamma \
-		 llrint llround log10 log1p log2 logb lrint lround nearbyint \
-		 nextafter nexttoward remainder remquo rint round scalbn \
-		 scalbln tgamma trunc acos asin atan acosh asinh atanh cos \
-		 sin tan cosh sinh tanh exp log pow sqrt fabs carg cimag conj \
-		 cproj creal roundeven nextup nextdown fminmag fmaxmag \
-		 fmaximum fmaximum_mag fmaximum_num fmaximum_mag_num \
-		 fminimum fminimum_mag fminimum_num fminimum_mag_num llogb \
-		 fromfp fromfpx ufromfp ufromfpx scalb $(tgmath3-narrow-macros)
+tgmath3-macros = \
+  $(tgmath3-narrow-macros) \
+  acos \
+  acosh \
+  asin \
+  asinh \
+  atan \
+  atan2 \
+  atanh \
+  carg \
+  cbrt \
+  ceil \
+  cimag \
+  conj \
+  copysign \
+  cos \
+  cosh \
+  cproj \
+  creal \
+  erf \
+  erfc \
+  exp \
+  exp2 \
+  exp10 \
+  expm1 \
+  fabs \
+  fdim \
+  floor \
+  fma \
+  fmax \
+  fmaximum \
+  fmaximum_mag \
+  fmaximum_mag_num \
+  fmaximum_num \
+  fmaxmag \
+  fmin \
+  fminimum \
+  fminimum_mag \
+  fminimum_mag_num \
+  fminimum_num \
+  fminmag \
+  fmod \
+  frexp \
+  fromfp \
+  fromfpx \
+  hypot \
+  ilogb \
+  ldexp \
+  lgamma \
+  llogb \
+  llrint \
+  llround \
+  log \
+  log10 \
+  log1p \
+  log2 \
+  logb \
+  lrint \
+  lround \
+  nearbyint \
+  nextafter \
+  nextdown \
+  nexttoward \
+  nextup \
+  pow \
+  remainder \
+  remquo \
+  rint \
+  round \
+  roundeven \
+  scalb \
+  scalbln \
+  scalbn \
+  sin \
+  sinh \
+  sqrt \
+  tan \
+  tanh \
+  tgamma \
+  trunc \
+  ufromfp \
+  ufromfpx \
+  # tgmath3-macros
 tgmath3-macro-tests = $(addprefix test-tgmath3-,$(tgmath3-macros))
 tests += $(tgmath3-macro-tests)
 generated += $(addsuffix .c,$(tgmath3-macro-tests))
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 02/12] powerpc: Add missing arch flags on rounding ifunc variants
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
  2023-12-28 17:20 ` [PATCH 01/12] math: Reformat Makefile Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 13:59   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 03/12] math: Add test to check if ceil raise inexact floating-point exception Adhemerval Zanella
                   ` (10 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

The ifunc variants now uses the powerpc implementation which in turn
uses the compiler builtin.  Without the proper -mcpu switch the builtin
does not generate the expected optimization.

Checked on powerpc-linux-gnu.
---
 sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile b/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile
index 64317506c6..2f5c0eded6 100644
--- a/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile
+++ b/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile
@@ -34,6 +34,12 @@ CFLAGS-s_modff-power5+.c = -mcpu=power5+
 CFLAGS-s_logbl-power7.c = -mcpu=power7
 CFLAGS-s_logb-power7.c = -mcpu=power7
 CFLAGS-s_logbf-power7.c = -mcpu=power7
+CFLAGS-s_round-power5+.c += -mcpu=power5+
+CFLAGS-s_roundf-power5+.c += -mcpu=power5+
+CFLAGS-s_floor-power5+.c += -mcpu=power5+
+CFLAGS-s_floorf-power5+.c += -mcpu=power5+
+CFLAGS-s_trunc-power5+.c += -mcpu=power5+
+CFLAGS-s_truncf-power5+.c += -mcpu=power5+
 
 # These files quiet sNaNs in a way that is optimized away without
 # -fsignaling-nans.
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 03/12] math: Add test to check if ceil raise inexact floating-point exception
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
  2023-12-28 17:20 ` [PATCH 01/12] math: Reformat Makefile Adhemerval Zanella
  2023-12-28 17:20 ` [PATCH 02/12] powerpc: Add missing arch flags on rounding ifunc variants Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:01   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 04/12] math: Add test to check if floor " Adhemerval Zanella
                   ` (9 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.
---
 math/Makefile             |  5 +++
 math/test-ceil-except-2.c | 67 ++++++++++++++++++++++++++++++
 math/test-ceil-except.c   | 85 +++++++++++++++++++++++++++++++++++++++
 3 files changed, 157 insertions(+)
 create mode 100644 math/test-ceil-except-2.c
 create mode 100644 math/test-ceil-except.c

diff --git a/math/Makefile b/math/Makefile
index 5a912c55bd..4fc7842313 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -498,6 +498,8 @@ tests = \
   bug-nextafter \
   bug-nexttoward \
   bug-tgmath1 \
+  test-ceil-except \
+  test-ceil-except-2 \
   test-femode \
   test-femode-traps \
   test-fenv basic-test \
@@ -989,6 +991,9 @@ CFLAGS-test-fe-snans-always-signal.c += $(config-cflags-signaling-nans)
 
 CFLAGS-test-nan-const.c += -fno-builtin
 
+CFLAGS-test-ceil-except.c += -fno-builtin
+CFLAGS-test-ceil-except-2.c += -fno-builtin
+
 include ../Rules
 
 gen-all-calls = $(gen-libm-calls) $(gen-calls)
diff --git a/math/test-ceil-except-2.c b/math/test-ceil-except-2.c
new file mode 100644
index 0000000000..c25956a402
--- /dev/null
+++ b/math/test-ceil-except-2.c
@@ -0,0 +1,67 @@
+/* Test ceil functions do not disable exception traps.
+   Copyright (C) 2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#include <fenv.h>
+#include <math.h>
+#include <stdio.h>
+
+#ifndef FE_INEXACT
+# define FE_INEXACT 0
+#endif
+
+#define TEST_FUNC(NAME, FLOAT, SUFFIX)					\
+static int								\
+NAME (void)								\
+{									\
+  int result = 0;							\
+  volatile FLOAT a, b __attribute__ ((unused));				\
+  a = 1.5;								\
+  /* ceil must work when traps on "inexact" are enabled.  */	\
+  b = ceil ## SUFFIX (a);						\
+  /* And it must have left those traps enabled.  */			\
+  if (fegetexcept () == FE_INEXACT)					\
+    puts ("PASS: " #FLOAT);						\
+  else									\
+    {									\
+      puts ("FAIL: " #FLOAT);						\
+      result = 1;							\
+    }									\
+  return result;							\
+}
+
+TEST_FUNC (float_test, float, f)
+TEST_FUNC (double_test, double, )
+TEST_FUNC (ldouble_test, long double, l)
+
+static int
+do_test (void)
+{
+  if (feenableexcept (FE_INEXACT) == -1)
+    {
+      puts ("enabling FE_INEXACT traps failed, cannot test");
+      return 77;
+    }
+  int result = float_test ();
+  feenableexcept (FE_INEXACT);
+  result |= double_test ();
+  feenableexcept (FE_INEXACT);
+  result |= ldouble_test ();
+  return result;
+}
+
+#include <support/test-driver.c>
diff --git a/math/test-ceil-except.c b/math/test-ceil-except.c
new file mode 100644
index 0000000000..cd0717a96f
--- /dev/null
+++ b/math/test-ceil-except.c
@@ -0,0 +1,85 @@
+/* Test ceil functions do not clear exceptions.
+   Copyright (C) 2015-2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#include <fenv.h>
+#include <math.h>
+#include <stdbool.h>
+#include <stdio.h>
+
+#include <math-tests.h>
+
+#ifndef FE_INVALID
+# define FE_INVALID 0
+#endif
+
+static bool any_supported = false;
+
+#define TEST_FUNC(NAME, FLOAT, SUFFIX)					\
+static int								\
+NAME (void)								\
+{									\
+  int result = 0;							\
+  if (!EXCEPTION_TESTS (FLOAT))						\
+    return 0;								\
+  any_supported = true;							\
+  volatile FLOAT a, b __attribute__ ((unused));				\
+  a = 1.0;								\
+  /* ceil must not clear already-raised exceptions.  */		\
+  feraiseexcept (FE_ALL_EXCEPT);					\
+  b = ceil ## SUFFIX (a);						\
+  if (fetestexcept (FE_ALL_EXCEPT) == FE_ALL_EXCEPT)			\
+    puts ("PASS: " #FLOAT);						\
+  else									\
+    {									\
+      puts ("FAIL: " #FLOAT);						\
+      result = 1;							\
+    }									\
+  /* But it mustn't lose exceptions from sNaN arguments.  */		\
+  if (SNAN_TESTS (FLOAT))						\
+    {									\
+      static volatile FLOAT snan = __builtin_nans ## SUFFIX ("");	\
+      volatile FLOAT c __attribute__ ((unused));			\
+      feclearexcept (FE_ALL_EXCEPT);					\
+      c = ceil ## SUFFIX (snan);					\
+      if (fetestexcept (FE_INVALID) == FE_INVALID)			\
+	puts ("PASS: " #FLOAT " sNaN");					\
+      else								\
+	{								\
+	  puts ("FAIL: " #FLOAT " sNaN");				\
+	  result = 1;							\
+	}								\
+    }									\
+  return result;							\
+}
+
+TEST_FUNC (float_test, float, f)
+TEST_FUNC (double_test, double, )
+TEST_FUNC (ldouble_test, long double, l)
+
+static int
+do_test (void)
+{
+  int result = float_test ();
+  result |= double_test ();
+  result |= ldouble_test ();
+  if (!any_supported)
+    return 77;
+  return result;
+}
+
+#include <support/test-driver.c>
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 04/12] math: Add test to check if floor raise inexact floating-point exception
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (2 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 03/12] math: Add test to check if ceil raise inexact floating-point exception Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:02   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 05/12] math: Add test to check if trunc " Adhemerval Zanella
                   ` (8 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.
---
 math/Makefile              |  4 ++
 math/test-floor-except-2.c | 67 ++++++++++++++++++++++++++++++
 math/test-floor-except.c   | 85 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 156 insertions(+)
 create mode 100644 math/test-floor-except-2.c
 create mode 100644 math/test-floor-except.c

diff --git a/math/Makefile b/math/Makefile
index 4fc7842313..19e234874a 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -512,6 +512,8 @@ tests = \
   test-fetestexceptflag \
   test-fexcept \
   test-fexcept-traps \
+  test-floor-except \
+  test-floor-except-2 \
   test-flt-eval-method \
   test-fp-ilogb-constants \
   test-fp-llogb-constants \
@@ -993,6 +995,8 @@ CFLAGS-test-nan-const.c += -fno-builtin
 
 CFLAGS-test-ceil-except.c += -fno-builtin
 CFLAGS-test-ceil-except-2.c += -fno-builtin
+CFLAGS-test-floor-except.c += -fno-builtin
+CFLAGS-test-floor-except-2.c += -fno-builtin
 
 include ../Rules
 
diff --git a/math/test-floor-except-2.c b/math/test-floor-except-2.c
new file mode 100644
index 0000000000..01b22f984b
--- /dev/null
+++ b/math/test-floor-except-2.c
@@ -0,0 +1,67 @@
+/* Test floor functions do not disable exception traps.
+   Copyright (C) 2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#include <fenv.h>
+#include <math.h>
+#include <stdio.h>
+
+#ifndef FE_INEXACT
+# define FE_INEXACT 0
+#endif
+
+#define TEST_FUNC(NAME, FLOAT, SUFFIX)					\
+static int								\
+NAME (void)								\
+{									\
+  int result = 0;							\
+  volatile FLOAT a, b __attribute__ ((unused));				\
+  a = 1.5;								\
+  /* floor must work when traps on "inexact" are enabled.  */	\
+  b = floor ## SUFFIX (a);						\
+  /* And it must have left those traps enabled.  */			\
+  if (fegetexcept () == FE_INEXACT)					\
+    puts ("PASS: " #FLOAT);						\
+  else									\
+    {									\
+      puts ("FAIL: " #FLOAT);						\
+      result = 1;							\
+    }									\
+  return result;							\
+}
+
+TEST_FUNC (float_test, float, f)
+TEST_FUNC (double_test, double, )
+TEST_FUNC (ldouble_test, long double, l)
+
+static int
+do_test (void)
+{
+  if (feenableexcept (FE_INEXACT) == -1)
+    {
+      puts ("enabling FE_INEXACT traps failed, cannot test");
+      return 77;
+    }
+  int result = float_test ();
+  feenableexcept (FE_INEXACT);
+  result |= double_test ();
+  feenableexcept (FE_INEXACT);
+  result |= ldouble_test ();
+  return result;
+}
+
+#include <support/test-driver.c>
diff --git a/math/test-floor-except.c b/math/test-floor-except.c
new file mode 100644
index 0000000000..fe77866ddc
--- /dev/null
+++ b/math/test-floor-except.c
@@ -0,0 +1,85 @@
+/* Test floor functions do not clear exceptions.
+   Copyright (C) 2015-2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#include <fenv.h>
+#include <math.h>
+#include <stdbool.h>
+#include <stdio.h>
+
+#include <math-tests.h>
+
+#ifndef FE_INVALID
+# define FE_INVALID 0
+#endif
+
+static bool any_supported = false;
+
+#define TEST_FUNC(NAME, FLOAT, SUFFIX)					\
+static int								\
+NAME (void)								\
+{									\
+  int result = 0;							\
+  if (!EXCEPTION_TESTS (FLOAT))						\
+    return 0;								\
+  any_supported = true;							\
+  volatile FLOAT a, b __attribute__ ((unused));				\
+  a = 1.0;								\
+  /* floor must not clear already-raised exceptions.  */		\
+  feraiseexcept (FE_ALL_EXCEPT);					\
+  b = floor ## SUFFIX (a);						\
+  if (fetestexcept (FE_ALL_EXCEPT) == FE_ALL_EXCEPT)			\
+    puts ("PASS: " #FLOAT);						\
+  else									\
+    {									\
+      puts ("FAIL: " #FLOAT);						\
+      result = 1;							\
+    }									\
+  /* But it mustn't lose exceptions from sNaN arguments.  */		\
+  if (SNAN_TESTS (FLOAT))						\
+    {									\
+      static volatile FLOAT snan = __builtin_nans ## SUFFIX ("");	\
+      volatile FLOAT c __attribute__ ((unused));			\
+      feclearexcept (FE_ALL_EXCEPT);					\
+      c = floor ## SUFFIX (snan);					\
+      if (fetestexcept (FE_INVALID) == FE_INVALID)			\
+	puts ("PASS: " #FLOAT " sNaN");					\
+      else								\
+	{								\
+	  puts ("FAIL: " #FLOAT " sNaN");				\
+	  result = 1;							\
+	}								\
+    }									\
+  return result;							\
+}
+
+TEST_FUNC (float_test, float, f)
+TEST_FUNC (double_test, double, )
+TEST_FUNC (ldouble_test, long double, l)
+
+static int
+do_test (void)
+{
+  int result = float_test ();
+  result |= double_test ();
+  result |= ldouble_test ();
+  if (!any_supported)
+    return 77;
+  return result;
+}
+
+#include <support/test-driver.c>
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 05/12] math: Add test to check if trunc raise inexact floating-point exception
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (3 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 04/12] math: Add test to check if floor " Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:03   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 06/12] math: Add test to check if round " Adhemerval Zanella
                   ` (7 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.
---
 math/Makefile              |  4 ++
 math/test-trunc-except-2.c | 67 ++++++++++++++++++++++++++++++
 math/test-trunc-except.c   | 85 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 156 insertions(+)
 create mode 100644 math/test-trunc-except-2.c
 create mode 100644 math/test-trunc-except.c

diff --git a/math/Makefile b/math/Makefile
index 19e234874a..7ae107170c 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -541,6 +541,8 @@ tests = \
   test-tgmath-int \
   test-tgmath-ret \
   test-tgmath2 \
+  test-trunc-except \
+  test-trunc-except-2 \
   tst-CMPLX \
   tst-CMPLX2 \
   tst-definitions \
@@ -997,6 +999,8 @@ CFLAGS-test-ceil-except.c += -fno-builtin
 CFLAGS-test-ceil-except-2.c += -fno-builtin
 CFLAGS-test-floor-except.c += -fno-builtin
 CFLAGS-test-floor-except-2.c += -fno-builtin
+CFLAGS-test-trunc-except.c += -fno-builtin
+CFLAGS-test-trunc-except-2.c += -fno-builtin
 
 include ../Rules
 
diff --git a/math/test-trunc-except-2.c b/math/test-trunc-except-2.c
new file mode 100644
index 0000000000..29716efc0b
--- /dev/null
+++ b/math/test-trunc-except-2.c
@@ -0,0 +1,67 @@
+/* Test trunc functions do not disable exception traps.
+   Copyright (C) 2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#include <fenv.h>
+#include <math.h>
+#include <stdio.h>
+
+#ifndef FE_INEXACT
+# define FE_INEXACT 0
+#endif
+
+#define TEST_FUNC(NAME, FLOAT, SUFFIX)					\
+static int								\
+NAME (void)								\
+{									\
+  int result = 0;							\
+  volatile FLOAT a, b __attribute__ ((unused));				\
+  a = 1.5;								\
+  /* trunc must work when traps on "inexact" are enabled.  */	\
+  b = trunc ## SUFFIX (a);						\
+  /* And it must have left those traps enabled.  */			\
+  if (fegetexcept () == FE_INEXACT)					\
+    puts ("PASS: " #FLOAT);						\
+  else									\
+    {									\
+      puts ("FAIL: " #FLOAT);						\
+      result = 1;							\
+    }									\
+  return result;							\
+}
+
+TEST_FUNC (float_test, float, f)
+TEST_FUNC (double_test, double, )
+TEST_FUNC (ldouble_test, long double, l)
+
+static int
+do_test (void)
+{
+  if (feenableexcept (FE_INEXACT) == -1)
+    {
+      puts ("enabling FE_INEXACT traps failed, cannot test");
+      return 77;
+    }
+  int result = float_test ();
+  feenableexcept (FE_INEXACT);
+  result |= double_test ();
+  feenableexcept (FE_INEXACT);
+  result |= ldouble_test ();
+  return result;
+}
+
+#include <support/test-driver.c>
diff --git a/math/test-trunc-except.c b/math/test-trunc-except.c
new file mode 100644
index 0000000000..c9e341de0d
--- /dev/null
+++ b/math/test-trunc-except.c
@@ -0,0 +1,85 @@
+/* Test trunc functions do not clear exceptions.
+   Copyright (C) 2015-2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#include <fenv.h>
+#include <math.h>
+#include <stdbool.h>
+#include <stdio.h>
+
+#include <math-tests.h>
+
+#ifndef FE_INVALID
+# define FE_INVALID 0
+#endif
+
+static bool any_supported = false;
+
+#define TEST_FUNC(NAME, FLOAT, SUFFIX)					\
+static int								\
+NAME (void)								\
+{									\
+  int result = 0;							\
+  if (!EXCEPTION_TESTS (FLOAT))						\
+    return 0;								\
+  any_supported = true;							\
+  volatile FLOAT a, b __attribute__ ((unused));				\
+  a = 1.0;								\
+  /* trunc must not clear already-raised exceptions.  */		\
+  feraiseexcept (FE_ALL_EXCEPT);					\
+  b = trunc ## SUFFIX (a);						\
+  if (fetestexcept (FE_ALL_EXCEPT) == FE_ALL_EXCEPT)			\
+    puts ("PASS: " #FLOAT);						\
+  else									\
+    {									\
+      puts ("FAIL: " #FLOAT);						\
+      result = 1;							\
+    }									\
+  /* But it mustn't lose exceptions from sNaN arguments.  */		\
+  if (SNAN_TESTS (FLOAT))						\
+    {									\
+      static volatile FLOAT snan = __builtin_nans ## SUFFIX ("");	\
+      volatile FLOAT c __attribute__ ((unused));			\
+      feclearexcept (FE_ALL_EXCEPT);					\
+      c = trunc ## SUFFIX (snan);					\
+      if (fetestexcept (FE_INVALID) == FE_INVALID)			\
+	puts ("PASS: " #FLOAT " sNaN");					\
+      else								\
+	{								\
+	  puts ("FAIL: " #FLOAT " sNaN");				\
+	  result = 1;							\
+	}								\
+    }									\
+  return result;							\
+}
+
+TEST_FUNC (float_test, float, f)
+TEST_FUNC (double_test, double, )
+TEST_FUNC (ldouble_test, long double, l)
+
+static int
+do_test (void)
+{
+  int result = float_test ();
+  result |= double_test ();
+  result |= ldouble_test ();
+  if (!any_supported)
+    return 77;
+  return result;
+}
+
+#include <support/test-driver.c>
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 06/12] math: Add test to check if round raise inexact floating-point exception
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (4 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 05/12] math: Add test to check if trunc " Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:02   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 07/12] x86: Do not raise inexact exception on ceill Adhemerval Zanella
                   ` (6 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.
---
 math/Makefile              |  4 ++
 math/test-round-except-2.c | 67 ++++++++++++++++++++++++++++++
 math/test-round-except.c   | 85 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 156 insertions(+)
 create mode 100644 math/test-round-except-2.c
 create mode 100644 math/test-round-except.c

diff --git a/math/Makefile b/math/Makefile
index 7ae107170c..dfe24a8c03 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -530,6 +530,8 @@ tests = \
   test-nearbyint-except \
   test-nearbyint-except-2 \
   test-powl \
+  test-round-except \
+  test-round-except-2 \
   test-signgam-uchar \
   test-signgam-uchar-init \
   test-signgam-uint \
@@ -1001,6 +1003,8 @@ CFLAGS-test-floor-except.c += -fno-builtin
 CFLAGS-test-floor-except-2.c += -fno-builtin
 CFLAGS-test-trunc-except.c += -fno-builtin
 CFLAGS-test-trunc-except-2.c += -fno-builtin
+CFLAGS-test-round-except.c += -fno-builtin
+CFLAGS-test-round-except-2.c += -fno-builtin
 
 include ../Rules
 
diff --git a/math/test-round-except-2.c b/math/test-round-except-2.c
new file mode 100644
index 0000000000..dffc02e91f
--- /dev/null
+++ b/math/test-round-except-2.c
@@ -0,0 +1,67 @@
+/* Test round functions do not disable exception traps.
+   Copyright (C) 2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#include <fenv.h>
+#include <math.h>
+#include <stdio.h>
+
+#ifndef FE_INEXACT
+# define FE_INEXACT 0
+#endif
+
+#define TEST_FUNC(NAME, FLOAT, SUFFIX)					\
+static int								\
+NAME (void)								\
+{									\
+  int result = 0;							\
+  volatile FLOAT a, b __attribute__ ((unused));				\
+  a = 1.5;								\
+  /* round must work when traps on "inexact" are enabled.  */	\
+  b = round ## SUFFIX (a);						\
+  /* And it must have left those traps enabled.  */			\
+  if (fegetexcept () == FE_INEXACT)					\
+    puts ("PASS: " #FLOAT);						\
+  else									\
+    {									\
+      puts ("FAIL: " #FLOAT);						\
+      result = 1;							\
+    }									\
+  return result;							\
+}
+
+TEST_FUNC (float_test, float, f)
+TEST_FUNC (double_test, double, )
+TEST_FUNC (ldouble_test, long double, l)
+
+static int
+do_test (void)
+{
+  if (feenableexcept (FE_INEXACT) == -1)
+    {
+      puts ("enabling FE_INEXACT traps failed, cannot test");
+      return 77;
+    }
+  int result = float_test ();
+  feenableexcept (FE_INEXACT);
+  result |= double_test ();
+  feenableexcept (FE_INEXACT);
+  result |= ldouble_test ();
+  return result;
+}
+
+#include <support/test-driver.c>
diff --git a/math/test-round-except.c b/math/test-round-except.c
new file mode 100644
index 0000000000..cd43c964ab
--- /dev/null
+++ b/math/test-round-except.c
@@ -0,0 +1,85 @@
+/* Test round functions do not clear exceptions.
+   Copyright (C) 2015-2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#include <fenv.h>
+#include <math.h>
+#include <stdbool.h>
+#include <stdio.h>
+
+#include <math-tests.h>
+
+#ifndef FE_INVALID
+# define FE_INVALID 0
+#endif
+
+static bool any_supported = false;
+
+#define TEST_FUNC(NAME, FLOAT, SUFFIX)					\
+static int								\
+NAME (void)								\
+{									\
+  int result = 0;							\
+  if (!EXCEPTION_TESTS (FLOAT))						\
+    return 0;								\
+  any_supported = true;							\
+  volatile FLOAT a, b __attribute__ ((unused));				\
+  a = 1.0;								\
+  /* round must not clear already-raised exceptions.  */		\
+  feraiseexcept (FE_ALL_EXCEPT);					\
+  b = round ## SUFFIX (a);						\
+  if (fetestexcept (FE_ALL_EXCEPT) == FE_ALL_EXCEPT)			\
+    puts ("PASS: " #FLOAT);						\
+  else									\
+    {									\
+      puts ("FAIL: " #FLOAT);						\
+      result = 1;							\
+    }									\
+  /* But it mustn't lose exceptions from sNaN arguments.  */		\
+  if (SNAN_TESTS (FLOAT))						\
+    {									\
+      static volatile FLOAT snan = __builtin_nans ## SUFFIX ("");	\
+      volatile FLOAT c __attribute__ ((unused));			\
+      feclearexcept (FE_ALL_EXCEPT);					\
+      c = round ## SUFFIX (snan);					\
+      if (fetestexcept (FE_INVALID) == FE_INVALID)			\
+	puts ("PASS: " #FLOAT " sNaN");					\
+      else								\
+	{								\
+	  puts ("FAIL: " #FLOAT " sNaN");				\
+	  result = 1;							\
+	}								\
+    }									\
+  return result;							\
+}
+
+TEST_FUNC (float_test, float, f)
+TEST_FUNC (double_test, double, )
+TEST_FUNC (ldouble_test, long double, l)
+
+static int
+do_test (void)
+{
+  int result = float_test ();
+  result |= double_test ();
+  result |= ldouble_test ();
+  if (!any_supported)
+    return 77;
+  return result;
+}
+
+#include <support/test-driver.c>
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 07/12] x86: Do not raise inexact exception on ceill
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (5 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 06/12] math: Add test to check if round " Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:12   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 08/12] x86: Do not raise inexact exception on floorl Adhemerval Zanella
                   ` (5 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.

Checked on x86_64-linux-gnu and i686-linux-gnu.
---
 sysdeps/i386/fpu/s_ceill.S   | 39 ------------------------------------
 sysdeps/x86/fpu/s_ceill.c    | 38 +++++++++++++++++++++++++++++++++++
 sysdeps/x86_64/fpu/s_ceill.S | 34 -------------------------------
 3 files changed, 38 insertions(+), 73 deletions(-)
 delete mode 100644 sysdeps/i386/fpu/s_ceill.S
 create mode 100644 sysdeps/x86/fpu/s_ceill.c
 delete mode 100644 sysdeps/x86_64/fpu/s_ceill.S

diff --git a/sysdeps/i386/fpu/s_ceill.S b/sysdeps/i386/fpu/s_ceill.S
deleted file mode 100644
index a551fce7f9..0000000000
--- a/sysdeps/i386/fpu/s_ceill.S
+++ /dev/null
@@ -1,39 +0,0 @@
-/*
- * Public domain.
- */
-
-#include <libm-alias-ldouble.h>
-#include <machine/asm.h>
-
-RCSID("$NetBSD: $")
-
-ENTRY(__ceill)
-	fldt	4(%esp)
-	subl	$32,%esp
-	cfi_adjust_cfa_offset (32)
-
-	fnstenv	4(%esp)			/* store fpu environment */
-
-	/* We use here %edx although only the low 1 bits are defined.
-	   But none of the operations should care and they are faster
-	   than the 16 bit operations.  */
-	movl	$0x0800,%edx		/* round towards +oo */
-	orl	4(%esp),%edx
-	andl	$0xfbff,%edx
-	movl	%edx,(%esp)
-	fldcw	(%esp)			/* load modified control word */
-
-	frndint				/* round */
-
-	/* Preserve "invalid" exceptions from sNaN input.  */
-	fnstsw
-	andl	$0x1, %eax
-	orl	%eax, 8(%esp)
-
-	fldenv	4(%esp)			/* restore original environment */
-
-	addl	$32,%esp
-	cfi_adjust_cfa_offset (-32)
-	ret
-END (__ceill)
-libm_alias_ldouble (__ceil, ceil)
diff --git a/sysdeps/x86/fpu/s_ceill.c b/sysdeps/x86/fpu/s_ceill.c
new file mode 100644
index 0000000000..96c901e383
--- /dev/null
+++ b/sysdeps/x86/fpu/s_ceill.c
@@ -0,0 +1,38 @@
+/* Return smallest integral value not less than argument.  x86 version.
+   Copyright (C) 2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <fenv_private.h>
+#include <libm-alias-ldouble.h>
+
+long double
+__ceill (long double x)
+{
+  fenv_t fenv;
+  long double r;
+
+  libc_feholdexcept_setround_387 (&fenv, FE_UPWARD);
+  asm volatile ("frndint" : "=t" (r) : "0" (x));
+  /* Preserve "invalid" exceptions from sNaN input.  */
+  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
+  libc_fesetenv_387 (&fenv);
+
+  return r;
+}
+libm_alias_ldouble (__ceil, ceil)
diff --git a/sysdeps/x86_64/fpu/s_ceill.S b/sysdeps/x86_64/fpu/s_ceill.S
deleted file mode 100644
index 16dbecd56d..0000000000
--- a/sysdeps/x86_64/fpu/s_ceill.S
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Public domain.
- */
-
-#include <libm-alias-ldouble.h>
-#include <machine/asm.h>
-
-
-ENTRY(__ceill)
-	fldt	8(%rsp)
-
-	fnstenv	-28(%rsp)		/* store fpu environment */
-
-	/* We use here %edx although only the low 1 bits are defined.
-	   But none of the operations should care and they are faster
-	   than the 16 bit operations.  */
-	movl	$0x0800,%edx		/* round towards +oo */
-	orl	-28(%rsp),%edx
-	andl	$0xfbff,%edx
-	movl	%edx,-32(%rsp)
-	fldcw	-32(%rsp)		/* load modified control word */
-
-	frndint				/* round */
-
-	/* Preserve "invalid" exceptions from sNaN input.  */
-	fnstsw
-	andl	$0x1, %eax
-	orl	%eax, -24(%rsp)
-
-	fldenv	-28(%rsp)		/* restore original environment */
-
-	ret
-END (__ceill)
-libm_alias_ldouble (__ceil, ceil)
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 08/12] x86: Do not raise inexact exception on floorl
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (6 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 07/12] x86: Do not raise inexact exception on ceill Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:13   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 09/12] x86: Do not raise inexact exception on truncl Adhemerval Zanella
                   ` (4 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.

Checked on x86_64-linux-gnu and i686-linux-gnu.
---
 sysdeps/i386/fpu/s_floorl.S   | 39 -----------------------------------
 sysdeps/x86/fpu/s_floorl.c    | 38 ++++++++++++++++++++++++++++++++++
 sysdeps/x86_64/fpu/s_floorl.S | 33 -----------------------------
 3 files changed, 38 insertions(+), 72 deletions(-)
 delete mode 100644 sysdeps/i386/fpu/s_floorl.S
 create mode 100644 sysdeps/x86/fpu/s_floorl.c
 delete mode 100644 sysdeps/x86_64/fpu/s_floorl.S

diff --git a/sysdeps/i386/fpu/s_floorl.S b/sysdeps/i386/fpu/s_floorl.S
deleted file mode 100644
index 3ec28b477b..0000000000
--- a/sysdeps/i386/fpu/s_floorl.S
+++ /dev/null
@@ -1,39 +0,0 @@
-/*
- * Public domain.
- */
-
-#include <libm-alias-ldouble.h>
-#include <machine/asm.h>
-
-RCSID("$NetBSD: $")
-
-ENTRY(__floorl)
-	fldt	4(%esp)
-	subl	$32,%esp
-	cfi_adjust_cfa_offset (32)
-
-	fnstenv	4(%esp)			/* store fpu environment */
-
-	/* We use here %edx although only the low 1 bits are defined.
-	   But none of the operations should care and they are faster
-	   than the 16 bit operations.  */
-	movl	$0x400,%edx		/* round towards -oo */
-	orl	4(%esp),%edx
-	andl	$0xf7ff,%edx
-	movl	%edx,(%esp)
-	fldcw	(%esp)			/* load modified control word */
-
-	frndint				/* round */
-
-	/* Preserve "invalid" exceptions from sNaN input.  */
-	fnstsw
-	andl	$0x1, %eax
-	orl	%eax, 8(%esp)
-
-	fldenv	4(%esp)			/* restore original environment */
-
-	addl	$32,%esp
-	cfi_adjust_cfa_offset (-32)
-	ret
-END (__floorl)
-libm_alias_ldouble (__floor, floor)
diff --git a/sysdeps/x86/fpu/s_floorl.c b/sysdeps/x86/fpu/s_floorl.c
new file mode 100644
index 0000000000..edaf5e954f
--- /dev/null
+++ b/sysdeps/x86/fpu/s_floorl.c
@@ -0,0 +1,38 @@
+/* Return largest integral value not less than argument.  x86 version.
+   Copyright (C) 2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <fenv_private.h>
+#include <libm-alias-ldouble.h>
+
+long double
+__floorl (long double x)
+{
+  fenv_t fenv;
+  long double r;
+
+  libc_feholdexcept_setround_387 (&fenv, FE_DOWNWARD);
+  asm volatile ("frndint" : "=t" (r) : "0" (x));
+  /* Preserve "invalid" exceptions from sNaN input.  */
+  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
+  libc_fesetenv_387 (&fenv);
+
+  return r;
+}
+libm_alias_ldouble (__floor, floor)
diff --git a/sysdeps/x86_64/fpu/s_floorl.S b/sysdeps/x86_64/fpu/s_floorl.S
deleted file mode 100644
index b74d1a4d6b..0000000000
--- a/sysdeps/x86_64/fpu/s_floorl.S
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Public domain.
- */
-
-#include <libm-alias-ldouble.h>
-#include <machine/asm.h>
-
-ENTRY(__floorl)
-	fldt	8(%rsp)
-
-	fnstenv	-28(%rsp)		/* store fpu environment */
-
-	/* We use here %edx although only the low 1 bits are defined.
-	   But none of the operations should care and they are faster
-	   than the 16 bit operations.  */
-	movl	$0x400,%edx		/* round towards -oo */
-	orl	-28(%rsp),%edx
-	andl	$0xf7ff,%edx
-	movl	%edx,-32(%rsp)
-	fldcw	-32(%rsp)		/* load modified control word */
-
-	frndint				/* round */
-
-	/* Preserve "invalid" exceptions from sNaN input.  */
-	fnstsw
-	andl	$0x1, %eax
-	orl	%eax, -24(%rsp)
-
-	fldenv	-28(%rsp)		/* restore original environment */
-
-	ret
-END (__floorl)
-libm_alias_ldouble (__floor, floor)
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 09/12] x86: Do not raise inexact exception on truncl
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (7 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 08/12] x86: Do not raise inexact exception on floorl Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:14   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 10/12] x86: Do not raise inexact exception on floor/floorf Adhemerval Zanella
                   ` (3 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.

Checked on x86_64-linux-gnu and i686-linux-gnu.
---
 .../fpu/s_truncl.S => x86/fpu/s_truncl.c}     | 40 +++++++++----------
 sysdeps/x86_64/fpu/s_truncl.S                 | 36 -----------------
 2 files changed, 19 insertions(+), 57 deletions(-)
 rename sysdeps/{i386/fpu/s_truncl.S => x86/fpu/s_truncl.c} (61%)
 delete mode 100644 sysdeps/x86_64/fpu/s_truncl.S

diff --git a/sysdeps/i386/fpu/s_truncl.S b/sysdeps/x86/fpu/s_truncl.c
similarity index 61%
rename from sysdeps/i386/fpu/s_truncl.S
rename to sysdeps/x86/fpu/s_truncl.c
index cfd4e75252..249b307004 100644
--- a/sysdeps/i386/fpu/s_truncl.S
+++ b/sysdeps/x86/fpu/s_truncl.c
@@ -1,5 +1,5 @@
-/* Truncate long double value.
-   Copyright (C) 1997-2023 Free Software Foundation, Inc.
+/* Round to integer, toward zero.  x86 version.
+   Copyright (C) 2023 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
@@ -16,25 +16,23 @@
    License along with the GNU C Library; if not, see
    <https://www.gnu.org/licenses/>.  */
 
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <fenv_private.h>
 #include <libm-alias-ldouble.h>
-#include <machine/asm.h>
 
-ENTRY(__truncl)
-	fldt	4(%esp)
-	subl	$32, %esp
-	cfi_adjust_cfa_offset (32)
-	fnstenv	4(%esp)
-	movl	$0xc00, %edx
-	orl	4(%esp), %edx
-	movl	%edx, (%esp)
-	fldcw	(%esp)
-	frndint
-	fnstsw
-	andl	$0x1, %eax
-	orl	%eax, 8(%esp)
-	fldenv	4(%esp)
-	addl	$32, %esp
-	cfi_adjust_cfa_offset (-32)
-	ret
-END(__truncl)
+long double
+__truncl (long double x)
+{
+  fenv_t fenv;
+  long double r;
+
+  libc_feholdexcept_setround_387 (&fenv, FE_TOWARDZERO);
+  asm volatile ("frndint" : "=t" (r) : "0" (x));
+  /* Preserve "invalid" exceptions from sNaN input.  */
+  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
+  libc_fesetenv_387 (&fenv);
+
+  return r;
+}
 libm_alias_ldouble (__trunc, trunc)
diff --git a/sysdeps/x86_64/fpu/s_truncl.S b/sysdeps/x86_64/fpu/s_truncl.S
deleted file mode 100644
index 8d319e68b0..0000000000
--- a/sysdeps/x86_64/fpu/s_truncl.S
+++ /dev/null
@@ -1,36 +0,0 @@
-/* Truncate long double value.
-   Copyright (C) 1997-2023 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
-   <https://www.gnu.org/licenses/>.  */
-
-#include <libm-alias-ldouble.h>
-#include <machine/asm.h>
-
-ENTRY(__truncl)
-	fldt	8(%rsp)
-	fnstenv	-28(%rsp)
-	movl	$0xc00, %edx
-	orl	-28(%rsp), %edx
-	movl	%edx, -32(%rsp)
-	fldcw	-32(%rsp)
-	frndint
-	fnstsw
-	andl	$0x1, %eax
-	orl	%eax, -24(%rsp)
-	fldenv	-28(%rsp)
-	ret
-END(__truncl)
-libm_alias_ldouble (__trunc, trunc)
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 10/12] x86: Do not raise inexact exception on floor/floorf
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (8 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 09/12] x86: Do not raise inexact exception on truncl Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:15   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 11/12] i386: Do not raise inexact exception on ceil/ceilf Adhemerval Zanella
                   ` (2 subsequent siblings)
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.

Checked on x86_64-linux-gnu and i686-linux-gnu.
---
 sysdeps/i386/fpu/s_floor.S  | 34 ---------------------------------
 sysdeps/i386/fpu/s_floor.c  | 38 +++++++++++++++++++++++++++++++++++++
 sysdeps/i386/fpu/s_floorf.S | 34 ---------------------------------
 sysdeps/i386/fpu/s_floorf.c | 38 +++++++++++++++++++++++++++++++++++++
 4 files changed, 76 insertions(+), 68 deletions(-)
 delete mode 100644 sysdeps/i386/fpu/s_floor.S
 create mode 100644 sysdeps/i386/fpu/s_floor.c
 delete mode 100644 sysdeps/i386/fpu/s_floorf.S
 create mode 100644 sysdeps/i386/fpu/s_floorf.c

diff --git a/sysdeps/i386/fpu/s_floor.S b/sysdeps/i386/fpu/s_floor.S
deleted file mode 100644
index 7143fdcc9a..0000000000
--- a/sysdeps/i386/fpu/s_floor.S
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Public domain.
- */
-
-#include <machine/asm.h>
-#include <libm-alias-double.h>
-
-RCSID("$NetBSD: s_floor.S,v 1.4 1995/05/09 00:01:59 jtc Exp $")
-
-ENTRY(__floor)
-	fldl	4(%esp)
-	subl	$32,%esp
-	cfi_adjust_cfa_offset (32)
-
-	fnstenv	4(%esp)			/* store fpu environment */
-
-	/* We use here %edx although only the low 1 bits are defined.
-	   But none of the operations should care and they are faster
-	   than the 16 bit operations.  */
-	movl	$0x400,%edx		/* round towards -oo */
-	orl	4(%esp),%edx
-	andl	$0xf7ff,%edx
-	movl	%edx,(%esp)
-	fldcw	(%esp)			/* load modified control word */
-
-	frndint				/* round */
-
-	fldenv	4(%esp)			/* restore original environment */
-
-	addl	$32,%esp
-	cfi_adjust_cfa_offset (-32)
-	ret
-END (__floor)
-libm_alias_double (__floor, floor)
diff --git a/sysdeps/i386/fpu/s_floor.c b/sysdeps/i386/fpu/s_floor.c
new file mode 100644
index 0000000000..90ba61da92
--- /dev/null
+++ b/sysdeps/i386/fpu/s_floor.c
@@ -0,0 +1,38 @@
+/* Return largest integral value not less than argument.  i386 version.
+   Copyright (C) 2023 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <fenv_private.h>
+#include <libm-alias-double.h>
+
+double
+__floor (double x)
+{
+  fenv_t fenv;
+  double r;
+
+  libc_feholdexcept_setround_387 (&fenv, FE_DOWNWARD);
+  asm volatile ("frndint" : "=t" (r) : "0" (x));
+  /* Preserve "invalid" exceptions from sNaN input.  */
+  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
+  libc_fesetenv_387 (&fenv);
+
+  return r;
+}
+libm_alias_double (__floor, floor)
diff --git a/sysdeps/i386/fpu/s_floorf.S b/sysdeps/i386/fpu/s_floorf.S
deleted file mode 100644
index 8fad9c0698..0000000000
--- a/sysdeps/i386/fpu/s_floorf.S
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Public domain.
- */
-
-#include <machine/asm.h>
-#include <libm-alias-float.h>
-
-RCSID("$NetBSD: s_floorf.S,v 1.3 1995/05/09 00:04:32 jtc Exp $")
-
-ENTRY(__floorf)
-	flds	4(%esp)
-	subl	$32,%esp
-	cfi_adjust_cfa_offset (32)
-
-	fnstenv	4(%esp)			/* store fpu environment */
-
-	/* We use here %edx although only the low 1 bits are defined.
-	   But none of the operations should care and they are faster
-	   than the 16 bit operations.  */
-	movl	$0x400,%edx		/* round towards -oo */
-	orl	4(%esp),%edx
-	andl	$0xf7ff,%edx
-	movl	%edx,(%esp)
-	fldcw	(%esp)			/* load modified control word */
-
-	frndint				/* round */
-
-	fldenv	4(%esp)			/* restore original environment */
-
-	addl	$32,%esp
-	cfi_adjust_cfa_offset (-32)
-	ret
-END (__floorf)
-libm_alias_float (__floor, floor)
diff --git a/sysdeps/i386/fpu/s_floorf.c b/sysdeps/i386/fpu/s_floorf.c
new file mode 100644
index 0000000000..0e7e8223a8
--- /dev/null
+++ b/sysdeps/i386/fpu/s_floorf.c
@@ -0,0 +1,38 @@
+/* Return largest integral value not less than argument.  i386 version.
+   Copyright (C) 2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <fenv_private.h>
+#include <libm-alias-float.h>
+
+float
+__floorf (float x)
+{
+  fenv_t fenv;
+  float r;
+
+  libc_feholdexcept_setround_387 (&fenv, FE_DOWNWARD);
+  asm volatile ("frndint" : "=t" (r) : "0" (x));
+  /* Preserve "invalid" exceptions from sNaN input.  */
+  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
+  libc_fesetenv_387 (&fenv);
+
+  return r;
+}
+libm_alias_float (__floor, floor)
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 11/12] i386: Do not raise inexact exception on ceil/ceilf
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (9 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 10/12] x86: Do not raise inexact exception on floor/floorf Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:16   ` H.J. Lu
  2023-12-28 17:20 ` [PATCH 12/12] i386: Do not raise inexact exception on trunc/truncf Adhemerval Zanella
  2024-02-16 17:47 ` [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella Netto
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.

Checked on x86_64-linux-gnu and i686-linux-gnu.
---
 sysdeps/i386/fpu/s_ceil.S  | 34 ----------------------------------
 sysdeps/i386/fpu/s_ceil.c  | 38 ++++++++++++++++++++++++++++++++++++++
 sysdeps/i386/fpu/s_ceilf.S | 34 ----------------------------------
 sysdeps/i386/fpu/s_ceilf.c | 38 ++++++++++++++++++++++++++++++++++++++
 4 files changed, 76 insertions(+), 68 deletions(-)
 delete mode 100644 sysdeps/i386/fpu/s_ceil.S
 create mode 100644 sysdeps/i386/fpu/s_ceil.c
 delete mode 100644 sysdeps/i386/fpu/s_ceilf.S
 create mode 100644 sysdeps/i386/fpu/s_ceilf.c

diff --git a/sysdeps/i386/fpu/s_ceil.S b/sysdeps/i386/fpu/s_ceil.S
deleted file mode 100644
index 99984f9b8d..0000000000
--- a/sysdeps/i386/fpu/s_ceil.S
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Public domain.
- */
-
-#include <machine/asm.h>
-#include <libm-alias-double.h>
-
-RCSID("$NetBSD: s_ceil.S,v 1.4 1995/05/08 23:52:13 jtc Exp $")
-
-ENTRY(__ceil)
-	fldl	4(%esp)
-	subl	$32,%esp
-	cfi_adjust_cfa_offset (32)
-
-	fnstenv	4(%esp)			/* store fpu environment */
-
-	/* We use here %edx although only the low 1 bits are defined.
-	   But none of the operations should care and they are faster
-	   than the 16 bit operations.  */
-	movl	$0x0800,%edx		/* round towards +oo */
-	orl	4(%esp),%edx
-	andl	$0xfbff,%edx
-	movl	%edx,(%esp)
-	fldcw	(%esp)			/* load modified control word */
-
-	frndint				/* round */
-
-	fldenv	4(%esp)			/* restore original environment */
-
-	addl	$32,%esp
-	cfi_adjust_cfa_offset (-32)
-	ret
-END (__ceil)
-libm_alias_double (__ceil, ceil)
diff --git a/sysdeps/i386/fpu/s_ceil.c b/sysdeps/i386/fpu/s_ceil.c
new file mode 100644
index 0000000000..ec948ef83f
--- /dev/null
+++ b/sysdeps/i386/fpu/s_ceil.c
@@ -0,0 +1,38 @@
+/* Return smallest integral value not less than argument.  i386 version.
+   Copyright (C) 2023 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <fenv_private.h>
+#include <libm-alias-double.h>
+
+double
+__ceil (double x)
+{
+  fenv_t fenv;
+  double r;
+
+  libc_feholdexcept_setround_387 (&fenv, FE_UPWARD);
+  asm volatile ("frndint" : "=t" (r) : "0" (x));
+  /* Preserve "invalid" exceptions from sNaN input.  */
+  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
+  libc_fesetenv_387 (&fenv);
+
+  return r;
+}
+libm_alias_double (__ceil, ceil)
diff --git a/sysdeps/i386/fpu/s_ceilf.S b/sysdeps/i386/fpu/s_ceilf.S
deleted file mode 100644
index 03e8e22609..0000000000
--- a/sysdeps/i386/fpu/s_ceilf.S
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Public domain.
- */
-
-#include <machine/asm.h>
-#include <libm-alias-float.h>
-
-RCSID("$NetBSD: s_ceilf.S,v 1.3 1995/05/08 23:52:44 jtc Exp $")
-
-ENTRY(__ceilf)
-	flds	4(%esp)
-	subl	$32,%esp
-	cfi_adjust_cfa_offset (32)
-
-	fnstenv	4(%esp)			/* store fpu environment */
-
-	/* We use here %edx although only the low 1 bits are defined.
-	   But none of the operations should care and they are faster
-	   than the 16 bit operations.  */
-	movl	$0x0800,%edx		/* round towards +oo */
-	orl	4(%esp),%edx
-	andl	$0xfbff,%edx
-	movl	%edx,(%esp)
-	fldcw	(%esp)			/* load modified control word */
-
-	frndint				/* round */
-
-	fldenv	4(%esp)			/* restore original environment */
-
-	addl	$32,%esp
-	cfi_adjust_cfa_offset (-32)
-	ret
-END (__ceilf)
-libm_alias_float (__ceil, ceil)
diff --git a/sysdeps/i386/fpu/s_ceilf.c b/sysdeps/i386/fpu/s_ceilf.c
new file mode 100644
index 0000000000..36159610b2
--- /dev/null
+++ b/sysdeps/i386/fpu/s_ceilf.c
@@ -0,0 +1,38 @@
+/* Return largest integral value not less than argument.  i386 version.
+   Copyright (C) 2023 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
+   <https://www.gnu.org/licenses/>.  */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <fenv_private.h>
+#include <libm-alias-float.h>
+
+float
+__ceilf (float x)
+{
+  fenv_t fenv;
+  float r;
+
+  libc_feholdexcept_setround_387 (&fenv, FE_UPWARD);
+  asm volatile ("frndint" : "=t" (r) : "0" (x));
+  /* Preserve "invalid" exceptions from sNaN input.  */
+  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
+  libc_fesetenv_387 (&fenv);
+
+  return r;
+}
+libm_alias_float (__ceil, ceil)
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* [PATCH 12/12] i386: Do not raise inexact exception on trunc/truncf
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (10 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 11/12] i386: Do not raise inexact exception on ceil/ceilf Adhemerval Zanella
@ 2023-12-28 17:20 ` Adhemerval Zanella
  2024-04-01 14:17   ` H.J. Lu
  2024-02-16 17:47 ` [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella Netto
  12 siblings, 1 reply; 27+ messages in thread
From: Adhemerval Zanella @ 2023-12-28 17:20 UTC (permalink / raw)
  To: libc-alpha

It is not allowed anymore on ISO C23.

Checked on x86_64-linux-gnu and i686-linux-gnu.
---
 sysdeps/i386/fpu/{s_trunc.S => s_trunc.c}   | 37 +++++++++++----------
 sysdeps/i386/fpu/{s_truncf.S => s_truncf.c} | 37 +++++++++++----------
 2 files changed, 38 insertions(+), 36 deletions(-)
 rename sysdeps/i386/fpu/{s_trunc.S => s_trunc.c} (61%)
 rename sysdeps/i386/fpu/{s_truncf.S => s_truncf.c} (61%)

diff --git a/sysdeps/i386/fpu/s_trunc.S b/sysdeps/i386/fpu/s_trunc.c
similarity index 61%
rename from sysdeps/i386/fpu/s_trunc.S
rename to sysdeps/i386/fpu/s_trunc.c
index 145288bc8e..b071ad9750 100644
--- a/sysdeps/i386/fpu/s_trunc.S
+++ b/sysdeps/i386/fpu/s_trunc.c
@@ -1,5 +1,5 @@
-/* Truncate double value.
-   Copyright (C) 1997-2023 Free Software Foundation, Inc.
+/* Round to integer, toward zero.  i386 version.
+   Copyright (C) 2023 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
@@ -16,22 +16,23 @@
    License along with the GNU C Library; if not, see
    <https://www.gnu.org/licenses/>.  */
 
-#include <machine/asm.h>
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <fenv_private.h>
 #include <libm-alias-double.h>
 
-ENTRY(__trunc)
-	fldl	4(%esp)
-	subl	$32, %esp
-	cfi_adjust_cfa_offset (32)
-	fnstenv	4(%esp)
-	movl	$0xc00, %edx
-	orl	4(%esp), %edx
-	movl	%edx, (%esp)
-	fldcw	(%esp)
-	frndint
-	fldenv	4(%esp)
-	addl	$32, %esp
-	cfi_adjust_cfa_offset (-32)
-	ret
-END(__trunc)
+double
+__trunc (double x)
+{
+  fenv_t fenv;
+  double r;
+
+  libc_feholdexcept_setround_387 (&fenv, FE_TOWARDZERO);
+  asm volatile ("frndint" : "=t" (r) : "0" (x));
+  /* Preserve "invalid" exceptions from sNaN input.  */
+  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
+  libc_fesetenv_387 (&fenv);
+
+  return r;
+}
 libm_alias_double (__trunc, trunc)
diff --git a/sysdeps/i386/fpu/s_truncf.S b/sysdeps/i386/fpu/s_truncf.c
similarity index 61%
rename from sysdeps/i386/fpu/s_truncf.S
rename to sysdeps/i386/fpu/s_truncf.c
index 7bdf0e5111..86d46ff660 100644
--- a/sysdeps/i386/fpu/s_truncf.S
+++ b/sysdeps/i386/fpu/s_truncf.c
@@ -1,5 +1,5 @@
-/* Truncate float value.
-   Copyright (C) 1997-2023 Free Software Foundation, Inc.
+/* Round to integer, toward zero.  i386 version.
+   Copyright (C) 2023 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
@@ -16,22 +16,23 @@
    License along with the GNU C Library; if not, see
    <https://www.gnu.org/licenses/>.  */
 
-#include <machine/asm.h>
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <fenv_private.h>
 #include <libm-alias-float.h>
 
-ENTRY(__truncf)
-	flds	4(%esp)
-	subl	$32, %esp
-	cfi_adjust_cfa_offset (32)
-	fnstenv	4(%esp)
-	movl	$0xc00, %edx
-	orl	4(%esp), %edx
-	movl	%edx, (%esp)
-	fldcw	(%esp)
-	frndint
-	fldenv	4(%esp)
-	addl	$32, %esp
-	cfi_adjust_cfa_offset (-32)
-	ret
-END(__truncf)
+float
+__truncf (float x)
+{
+  fenv_t fenv;
+  float r;
+
+  libc_feholdexcept_setround_387 (&fenv, FE_TOWARDZERO);
+  asm volatile ("frndint" : "=t" (r) : "0" (x));
+  /* Preserve "invalid" exceptions from sNaN input.  */
+  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
+  libc_fesetenv_387 (&fenv);
+
+  return r;
+}
 libm_alias_float (__trunc, trunc)
-- 
2.34.1


^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 00/12] Improve rounding to interger function for C23
  2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
                   ` (11 preceding siblings ...)
  2023-12-28 17:20 ` [PATCH 12/12] i386: Do not raise inexact exception on trunc/truncf Adhemerval Zanella
@ 2024-02-16 17:47 ` Adhemerval Zanella Netto
  12 siblings, 0 replies; 27+ messages in thread
From: Adhemerval Zanella Netto @ 2024-02-16 17:47 UTC (permalink / raw)
  To: libc-alpha

Ping.

On 28/12/23 14:20, Adhemerval Zanella wrote:
> As indicated by GCC documentation [1], ISO C23 does not allow that C
> bindings ceil, floor, round, and trunc (in all floating point formats)
> to raise 
> inexact exceptions (different than ISO C99/C11 where this is allowed).
> 
> A recent MIPS patch to used some arch-specific instructions raised this
> issue [1] and it was not caught because there was no proper testing. By
> adding the missing tests, some implementations do indeed raise inexact
> exceptions. 
> 
> The generic implementation all uses integer operation, so they are not
> subject to this issue. The powerpc (for power4 and lower) and the riscv
> avoid the inexact exception by disabling/enabling exceptions. The x86
> uses some arch-specific implementation for long double and on i386 (due
> to the use of x87 instruction).
> 
> Instead of adding newer symbols depending on the required standard
> version, the patchset adapts the faulty ones to avoid raising the
> inexact exception. The x86 version already saves/restore the floating
> point status, so I think it is unlikely the patch would yield much
> performance difference (I did not do any performance analysis on whether
> a generic implementation would yield better performance).
> 
> I checked on powerpc, powerpc64, aarch64, armhf, x86, and did some
> regression checks on riscv.
> 
> [1] https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fno-fp-int-builtin-inexact
> [2] https://sourceware.org/pipermail/libc-alpha/2023-December/153528.html
> 
> Adhemerval Zanella (12):
>   math: Reformat Makefile.
>   powerpc: Add missing arch flags on rounding ifunc variants
>   math: Add test to check if ceil raise inexact floating-point exception
>   math: Add test to check if floor raise inexact floating-point
>     exception
>   math: Add test to check if trunc raise inexact floating-point
>     exception
>   math: Add test to check if round raise inexact floating-point
>     exception
>   x86: Do not raise inexact exception on ceill
>   x86: Do not raise inexact exception on floorl
>   x86: Do not raise inexact exception on truncl
>   x86: Do not raise inexact exception on floor/floorf
>   i386: Do not raise inexact exception on ceil/ceilf
>   i386: Do not raise inexact exception on trunc/truncf
> 
>  math/Makefile                                 | 861 ++++++++++++++----
>  math/test-ceil-except-2.c                     |  67 ++
>  math/test-ceil-except.c                       |  85 ++
>  math/test-floor-except-2.c                    |  67 ++
>  math/test-floor-except.c                      |  85 ++
>  math/test-round-except-2.c                    |  67 ++
>  math/test-round-except.c                      |  85 ++
>  math/test-trunc-except-2.c                    |  67 ++
>  math/test-trunc-except.c                      |  85 ++
>  sysdeps/i386/fpu/s_ceil.S                     |  34 -
>  sysdeps/i386/fpu/s_ceil.c                     |  38 +
>  sysdeps/i386/fpu/s_ceilf.S                    |  34 -
>  sysdeps/i386/fpu/s_ceilf.c                    |  38 +
>  sysdeps/i386/fpu/s_ceill.S                    |  39 -
>  sysdeps/i386/fpu/s_floor.S                    |  34 -
>  sysdeps/i386/fpu/s_floor.c                    |  38 +
>  sysdeps/i386/fpu/s_floorf.S                   |  34 -
>  sysdeps/i386/fpu/s_floorf.c                   |  38 +
>  sysdeps/i386/fpu/s_floorl.S                   |  39 -
>  sysdeps/i386/fpu/{s_trunc.S => s_trunc.c}     |  37 +-
>  sysdeps/i386/fpu/{s_truncf.S => s_truncf.c}   |  37 +-
>  .../powerpc32/power4/fpu/multiarch/Makefile   |   6 +
>  .../fpu/s_truncl.S => x86/fpu/s_ceill.c}      |  38 +-
>  sysdeps/x86/fpu/s_floorl.c                    |  38 +
>  .../fpu/s_truncl.S => x86/fpu/s_truncl.c}     |  40 +-
>  sysdeps/x86_64/fpu/s_ceill.S                  |  34 -
>  sysdeps/x86_64/fpu/s_floorl.S                 |  33 -
>  27 files changed, 1583 insertions(+), 515 deletions(-)
>  create mode 100644 math/test-ceil-except-2.c
>  create mode 100644 math/test-ceil-except.c
>  create mode 100644 math/test-floor-except-2.c
>  create mode 100644 math/test-floor-except.c
>  create mode 100644 math/test-round-except-2.c
>  create mode 100644 math/test-round-except.c
>  create mode 100644 math/test-trunc-except-2.c
>  create mode 100644 math/test-trunc-except.c
>  delete mode 100644 sysdeps/i386/fpu/s_ceil.S
>  create mode 100644 sysdeps/i386/fpu/s_ceil.c
>  delete mode 100644 sysdeps/i386/fpu/s_ceilf.S
>  create mode 100644 sysdeps/i386/fpu/s_ceilf.c
>  delete mode 100644 sysdeps/i386/fpu/s_ceill.S
>  delete mode 100644 sysdeps/i386/fpu/s_floor.S
>  create mode 100644 sysdeps/i386/fpu/s_floor.c
>  delete mode 100644 sysdeps/i386/fpu/s_floorf.S
>  create mode 100644 sysdeps/i386/fpu/s_floorf.c
>  delete mode 100644 sysdeps/i386/fpu/s_floorl.S
>  rename sysdeps/i386/fpu/{s_trunc.S => s_trunc.c} (61%)
>  rename sysdeps/i386/fpu/{s_truncf.S => s_truncf.c} (61%)
>  rename sysdeps/{x86_64/fpu/s_truncl.S => x86/fpu/s_ceill.c} (57%)
>  create mode 100644 sysdeps/x86/fpu/s_floorl.c
>  rename sysdeps/{i386/fpu/s_truncl.S => x86/fpu/s_truncl.c} (61%)
>  delete mode 100644 sysdeps/x86_64/fpu/s_ceill.S
>  delete mode 100644 sysdeps/x86_64/fpu/s_floorl.S
> 

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 01/12] math: Reformat Makefile.
  2023-12-28 17:20 ` [PATCH 01/12] math: Reformat Makefile Adhemerval Zanella
@ 2024-04-01 13:57   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 13:57 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:21 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> Reflow all long lines adding comment terminators.
> Sort all reflowed text using scripts/sort-makefile-lines.py.
>
> No code generation changes observed in binary artifacts.
> No regressions on x86_64 and i686.
> ---
>  math/Makefile | 844 ++++++++++++++++++++++++++++++++++++++++----------
>  1 file changed, 685 insertions(+), 159 deletions(-)
>
> diff --git a/math/Makefile b/math/Makefile
> index a9daae09de..5a912c55bd 100644
> --- a/math/Makefile
> +++ b/math/Makefile
> @@ -22,87 +22,279 @@ subdir             := math
>  include ../Makeconfig
>
>  # Installed header files.
> -headers                := math.h bits/mathcalls.h \
> -                  fpu_control.h complex.h bits/cmathcalls.h fenv.h \
> -                  bits/fenv.h bits/mathdef.h tgmath.h \
> -                  bits/math-vector.h finclude/math-vector-fortran.h \
> -                  bits/libm-simd-decl-stubs.h bits/iscanonical.h \
> -                  bits/flt-eval-method.h bits/fp-fast.h bits/fp-logb.h \
> -                  bits/long-double.h bits/mathcalls-helper-functions.h \
> -                  bits/floatn.h bits/floatn-common.h bits/mathcalls-narrow.h
> +headers        := \
> +  math.h \
> +  bits/mathcalls.h \
> +  fpu_control.h \
> +  complex.h \
> +  bits/cmathcalls.h \
> +  fenv.h \
> +  bits/fenv.h \
> +  bits/mathdef.h \
> +  tgmath.h \
> +  bits/math-vector.h \
> +  finclude/math-vector-fortran.h \
> +  bits/libm-simd-decl-stubs.h \
> +  bits/iscanonical.h \
> +  bits/flt-eval-method.h \
> +  bits/fp-fast.h \
> +  bits/fp-logb.h \
> +  bits/long-double.h \
> +  bits/mathcalls-helper-functions.h \
> +  bits/floatn.h \
> +  bits/floatn-common.h \
> +  bits/mathcalls-narrow.h
> +  # headers
>
>  # FPU support code.
> -aux            := setfpucw fpu_control
> +aux := \
> +  fpu_control \
> +  setfpucw \
> +  # aux
>
>  # Build the -lm library.
>
>  extra-libs     := libm
>  extra-libs-others = $(extra-libs)
>
> -libm-support = s_lib_version s_matherr s_signgam                       \
> -              fclrexcpt fgetexcptflg fraiseexcpt fsetexcptflg          \
> -              ftestexcept fegetround fesetround fegetenv feholdexcpt   \
> -              fesetenv feupdateenv fedisblxcpt feenablxcpt     \
> -              fegetexcept fesetexcept fetestexceptflag fegetmode       \
> -              fesetmode
> +libm-support = \
> +  fclrexcpt \
> +  fedisblxcpt \
> +  feenablxcpt \
> +  fegetenv \
> +  fegetexcept \
> +  fegetmode \
> +  fegetround \
> +  feholdexcpt \
> +  fesetenv \
> +  fesetexcept \
> +  fesetmode\
> +  fesetround \
> +  fetestexceptflag \
> +  feupdateenv \
> +  fgetexcptflg \
> +  fraiseexcpt \
> +  fsetexcptflg \
> +  ftestexcept \
> +  s_lib_version \
> +  s_matherr \
> +  s_signgam \
> +  # libm-support
>
>  # Wrappers for these functions generated per type using a file named
>  # <func>_template.c and the appropriate math-type-macros-<TYPE>.h.
> -gen-libm-calls = cargF conjF cimagF crealF cabsF e_scalbF s_cacosF       \
> -                s_cacoshF s_ccosF s_ccoshF s_casinF s_csinF s_casinhF    \
> -                k_casinhF s_csinhF k_casinhF s_csinhF s_catanhF s_catanF \
> -                s_ctanF s_ctanhF s_cexpF s_clogF s_cprojF s_csqrtF       \
> -                s_cpowF s_clog10F s_fdimF s_nextdownF s_fmaxF s_fminF    \
> -                s_nanF s_iseqsigF s_canonicalizeF s_significandF         \
> -                w_ilogbF w_llogbF                                        \
> -                w_log1pF w_scalblnF s_fmaxmagF s_fminmagF w_acosF        \
> -                w_acoshF w_asinF w_atan2F w_atanhF w_coshF w_exp10F      \
> -                w_exp2F w_fmodF w_hypotF w_j0F w_j1F w_jnF w_logF        \
> -                w_log10F w_log2F w_powF w_remainderF w_scalbF            \
> -                w_sinhF w_sqrtF                                          \
> -                w_tgammaF w_lgammaF w_lgammaF_r w_expF e_exp2F           \
> -                s_fmaximumF s_fmaximum_magF s_fmaximum_numF              \
> -                s_fmaximum_mag_numF s_fminimumF s_fminimum_magF          \
> -                s_fminimum_numF s_fminimum_mag_numF
> -
> -libm-calls =                                                             \
> -       e_acosF e_acoshF e_asinF e_atan2F e_atanhF e_coshF e_expF e_fmodF \
> -       e_hypotF e_j0F e_j1F e_jnF e_lgammaF_r e_logF e_log10F e_powF     \
> -       e_remainderF e_sinhF e_sqrtF e_gammaF_r                           \
> -       e_ilogbF                                                          \
> -       k_tanF s_asinhF s_atanF s_cbrtF                                   \
> -       s_ceilF s_cosF s_erfF s_expm1F s_fabsF                            \
> -       s_floorF s_log1pF s_logbF                                 \
> -       s_nextafterF s_nexttowardF s_rintF s_scalblnF                     \
> -       s_sinF s_tanF s_tanhF                                             \
> -       s_fpclassifyF s_truncF                                            \
> -       s_remquoF e_log2F s_roundF s_nearbyintF s_sincosF                 \
> -       s_fmaF s_lrintF s_llrintF s_lroundF s_llroundF e_exp10F           \
> -       s_issignalingF $(calls:s_%=m_%) x2y2m1F                           \
> -       gamma_productF lgamma_negF lgamma_productF                        \
> -       s_nextupF s_totalorderF s_totalordermagF s_getpayloadF            \
> -       s_setpayloadF s_setpayloadsigF s_roundevenF s_fromfpF s_ufromfpF  \
> -       s_fromfpxF s_ufromfpxF $(gen-libm-calls)
> -
> -libm-compat-calls =                                                    \
> -       w_acosF_compat w_acoshF_compat w_asinF_compat w_atan2F_compat   \
> -       w_atanhF_compat w_coshF_compat w_exp2F_compat w_exp10F_compat   \
> -       w_fmodF_compat w_hypotF_compat w_j0F_compat w_j1F_compat        \
> -       w_jnF_compat w_log2F_compat w_log10F_compat w_logF_compat       \
> -       w_powF_compat w_remainderF_compat w_scalbF_compat               \
> -       w_sinhF_compat w_sqrtF_compat w_tgammaF_compat                  \
> -       w_lgammaF_r_compat w_lgammaF_compat2 w_expF_compat              \
> -       w_lgamma_compatF k_standardF
> -
> -libm-narrow-fns = add div fma mul sqrt sub
> -libm-narrow-types-basic = s_fF s_f32xFf64
> -libm-narrow-types-ldouble-yes = s_fFl s_dFl
> -libm-narrow-types-float128-yes = s_f32Ff128 s_f64Ff128 s_f64xFf128
> -libm-narrow-types-float128-alias-yes = s_f64xFf128
> -libm-narrow-types = $(libm-narrow-types-basic) \
> -                   $(libm-narrow-types-ldouble-$(long-double-fcts)) \
> -                   $(libm-narrow-types-float128-$(float128-fcts)) \
> -                   $(libm-narrow-types-float128-alias-$(float128-alias-fcts))
> +gen-libm-calls = \
> +  cabsF \
> +  cargF \
> +  cimagF \
> +  conjF \
> +  crealF \
> +  e_exp2F \
> +  e_scalbF \
> +  k_casinhF \
> +  k_casinhF \
> +  s_cacosF  \
> +  s_cacoshF \
> +  s_canonicalizeF \
> +  s_casinF \
> +  s_casinhF \
> +  s_catanF \
> +  s_catanhF \
> +  s_ccosF \
> +  s_ccoshF \
> +  s_cexpF \
> +  s_clog10F \
> +  s_clogF \
> +  s_cpowF \
> +  s_cprojF \
> +  s_csinF \
> +  s_csinhF \
> +  s_csinhF \
> +  s_csqrtF \
> +  s_ctanF \
> +  s_ctanhF \
> +  s_fdimF \
> +  s_fmaxF \
> +  s_fmaximumF \
> +  s_fmaximum_magF \
> +  s_fmaximum_mag_numF \
> +  s_fmaximum_numF \
> +  s_fmaxmagF \
> +  s_fminF \
> +  s_fminimumF \
> +  s_fminimum_magF \
> +  s_fminimum_mag_numF \
> +  s_fminimum_numF \
> +  s_fminmagF \
> +  s_iseqsigF \
> +  s_nanF \
> +  s_nextdownF \
> +  s_significandF \
> +  w_acosF \
> +  w_acoshF \
> +  w_asinF \
> +  w_atan2F \
> +  w_atanhF \
> +  w_coshF \
> +  w_exp10F \
> +  w_exp2F \
> +  w_expF \
> +  w_fmodF \
> +  w_hypotF \
> +  w_ilogbF \
> +  w_j0F \
> +  w_j1F \
> +  w_jnF \
> +  w_lgammaF \
> +  w_lgammaF_r \
> +  w_llogbF \
> +  w_log10F \
> +  w_log1pF \
> +  w_log2F \
> +  w_logF  \
> +  w_powF \
> +  w_remainderF \
> +  w_scalbF \
> +  w_scalblnF \
> +  w_sinhF \
> +  w_sqrtF \
> +  w_tgammaF \
> +  # gen-libm-calls
> +
> +libm-calls = \
> +  $(calls:s_%=m_%) \
> +  $(gen-libm-calls) \
> +  e_acosF \
> +  e_acoshF \
> +  e_asinF \
> +  e_atan2F \
> +  e_atanhF \
> +  e_coshF \
> +  e_exp10F \
> +  e_expF \
> +  e_fmodF \
> +  e_gammaF_r \
> +  e_hypotF \
> +  e_ilogbF \
> +  e_j0F \
> +  e_j1F \
> +  e_jnF \
> +  e_lgammaF_r \
> +  e_log10F \
> +  e_log2F \
> +  e_logF \
> +  e_powF \
> +  e_remainderF \
> +  e_sinhF \
> +  e_sqrtF \
> +  gamma_productF \
> +  k_tanF \
> +  lgamma_negF \
> +  lgamma_productF \
> +  s_asinhF \
> +  s_atanF \
> +  s_cbrtF \
> +  s_ceilF \
> +  s_cosF \
> +  s_erfF \
> +  s_expm1F \
> +  s_fabsF \
> +  s_floorF \
> +  s_fmaF \
> +  s_fpclassifyF \
> +  s_fromfpF \
> +  s_fromfpxF \
> +  s_getpayloadF \
> +  s_issignalingF \
> +  s_llrintF \
> +  s_llroundF \
> +  s_log1pF \
> +  s_logbF \
> +  s_lrintF \
> +  s_lroundF \
> +  s_nearbyintF \
> +  s_nextafterF \
> +  s_nexttowardF \
> +  s_nextupF \
> +  s_remquoF \
> +  s_rintF \
> +  s_roundF \
> +  s_roundevenF \
> +  s_scalblnF \
> +  s_setpayloadF \
> +  s_setpayloadsigF \
> +  s_sinF \
> +  s_sincosF \
> +  s_tanF \
> +  s_tanhF \
> +  s_totalorderF \
> +  s_totalordermagF \
> +  s_truncF \
> +  s_ufromfpF \
> +  s_ufromfpxF \
> +  x2y2m1F \
> +  # libm-calls
> +
> +libm-compat-calls = \
> +  k_standardF \
> +  w_acosF_compat \
> +  w_acoshF_compat \
> +  w_asinF_compat \
> +  w_atan2F_compat \
> +  w_atanhF_compat \
> +  w_coshF_compat \
> +  w_exp10F_compat \
> +  w_exp2F_compat \
> +  w_expF_compat        \
> +  w_fmodF_compat \
> +  w_hypotF_compat \
> +  w_j0F_compat \
> +  w_j1F_compat\
> +  w_jnF_compat \
> +  w_lgammaF_compat2 \
> +  w_lgammaF_r_compat \
> +  w_lgamma_compatF \
> +  w_log10F_compat \
> +  w_log2F_compat \
> +  w_logF_compat        \
> +  w_powF_compat \
> +  w_remainderF_compat \
> +  w_scalbF_compat \
> +  w_sinhF_compat \
> +  w_sqrtF_compat \
> +  w_tgammaF_compat \
> +  # libm-compat-calls
> +
> +libm-narrow-fns = \
> +  add \
> +  div \
> +  fma \
> +  mul \
> +  sqrt \
> +  sub \
> +  # libm-narrow-fns
> +libm-narrow-types-basic = \
> +  s_f32xFf64 \
> +  s_fF \
> +  # libm-narrow-types-basic
> +libm-narrow-types-ldouble-yes = \
> +  s_dFl \
> +  s_fFl \
> +  # libm-narrow-types-ldouble-yes
> +libm-narrow-types-float128-yes = \
> +  s_f32Ff128 \
> +  s_f64Ff128 \
> +  s_f64xFf128 \
> +  # libm-narrow-types-float128-yes
> +libm-narrow-types-float128-alias-yes = \
> +  s_f64xFf128 \
> +  # libm-narrow-types-float128-alias-yes
> +libm-narrow-types = \
> +  $(libm-narrow-types-basic) \
> +  $(libm-narrow-types-float128-$(float128-fcts)) \
> +  $(libm-narrow-types-float128-alias-$(float128-alias-fcts)) \
> +  $(libm-narrow-types-ldouble-$(long-double-fcts)) \
> +  # libm-narrow-types
>
>  # Type specific routine support.
>  #
> @@ -115,33 +307,65 @@ libm-narrow-types = $(libm-narrow-types-basic) \
>  # Finally, note that types is an intentionally recursive variable.
>  # We only know the full set of supported types for the target machine
>  # after the Rules makefile has been parsed.
> -types-basic = $(type-ldouble-$(long-double-fcts)) double float
> +types-basic = \
> +  $(type-ldouble-$(long-double-fcts)) \
> +  double \
> +  float \
> +  # types-basic
>
>  # Like types, but includes types whose functions alias those for
>  # another type.
> -test-types-basic = ldouble double float
> +test-types-basic = \
> +  double \
> +  float \
> +  ldouble \
> +  # test-types-basic
>
>  # long double support
>  type-ldouble-suffix := l
> -type-ldouble-routines := t_sincosl k_sinl k_cosl k_sincosl s_iscanonicall \
> -                        e_rem_pio2l
> +type-ldouble-routines := \
> +  e_rem_pio2l \
> +  k_cosl \
> +  k_sincosl \
> +  k_sinl \
> +  s_iscanonicall \
> +  t_sincosl \
> +  # type-ldouble-routines
>  type-ldouble-yes := ldouble
>
>  # double support
>  type-double-suffix :=
> -type-double-routines := branred k_rem_pio2     \
> -                      sincostab math_err e_exp_data e_log_data \
> -                      e_log2_data e_pow_log_data
> +type-double-routines := \
> +  branred \
> +  e_exp_data \
> +  e_log2_data \
> +  e_log_data \
> +  e_pow_log_data \
> +  k_rem_pio2 \
> +  math_err \
> +  sincostab \
> +  # type-double-routines
>
>  # float support
>  type-float-suffix := f
> -type-float-routines := math_errf e_exp2f_data e_logf_data      \
> -                      e_log2f_data e_powf_log2_data s_sincosf_data
> +type-float-routines := \
> +  e_exp2f_data \
> +  e_log2f_data \
> +  e_logf_data \
> +  e_powf_log2_data \
> +  math_errf \
> +  s_sincosf_data \
> +  # type-float-routines
>
>  # _Float128 support
>  type-float128-suffix := f128
> -type-float128-routines := t_sincosf128 k_sinf128 k_cosf128 k_sincosf128 \
> -                         e_rem_pio2f128
> +type-float128-routines := \
> +  e_rem_pio2f128 \
> +  k_cosf128 \
> +  k_sincosf128 \
> +  k_sinf128 \
> +  t_sincosf128 \
> +  # type-float128-routines
>  type-float128-yes := float128
>
>  # _Float64x may be supported, only as an alias type.
> @@ -151,11 +375,20 @@ type-float64x-yes := float64x
>  type-ibm128-suffix := l
>  type-ibm128-yes := ibm128
>
> -types = $(types-basic) $(type-float128-$(float128-fcts))
> -test-types = $(test-types-basic) $(type-float128-$(float128-fcts)) \
> -            float32 float64 $(type-float128-$(float128-alias-fcts)) \
> -            float32x $(type-float64x-$(float64x-alias-fcts)) \
> -            $(type-ibm128-$(ibm128-fcts))
> +types = \
> +  $(type-float128-$(float128-fcts)) \
> +  $(types-basic) \
> +  # types
> +test-types = \
> +  $(test-types-basic) \
> +  $(type-float128-$(float128-alias-fcts)) \
> +  $(type-float128-$(float128-fcts)) \
> +  $(type-float64x-$(float64x-alias-fcts)) \
> +  $(type-ibm128-$(ibm128-fcts)) \
> +  float32 \
> +  float32x \
> +  float64 \
> +  # test-types
>
>  # Pairs of types for which narrowing functions should be tested (this
>  # variable has more entries than libm-narrow-types because it includes
> @@ -163,14 +396,25 @@ test-types = $(test-types-basic) $(type-float128-$(float128-fcts)) \
>  # for other types). This definition embeds the assumption that if
>  # _Float64x is supported, so is _Float128, and vice versa (they may or
>  # may not have the same format).
> -test-type-pairs = float-double float-ldouble double-ldouble \
> -                 float32-float64 float32-float32x float32x-float64 \
> -                 $(test-type-pairs-f64xf128-$(float128-fcts)) \
> -                 $(test-type-pairs-f64xf128-$(float128-alias-fcts))
> -test-type-pairs-f64xf128-yes = float32-float64x float32-float128 \
> -                              float64-float64x float64-float128 \
> -                              float32x-float64x float32x-float128 \
> -                              float64x-float128
> +test-type-pairs = \
> +  $(test-type-pairs-f64xf128-$(float128-alias-fcts)) \
> +  $(test-type-pairs-f64xf128-$(float128-fcts)) \
> +  double-ldouble \
> +  float-double \
> +  float-ldouble \
> +  float32-float32x \
> +  float32-float64 \
> +  float32x-float64 \
> +  # test-type-pairs
> +test-type-pairs-f64xf128-yes = \
> +  float32-float128 \
> +  float32-float64x \
> +  float32x-float128 \
> +  float32x-float64x \
> +  float64-float128 \
> +  float64-float64x \
> +  float64x-float128 \
> +  # test-type-pairs-f64xf128-yes
>
>  # For each of the basic types (float, double, long double), replace the
>  # occurrences of 'F' in arg 1 with the appropriate suffix for the type.
> @@ -193,8 +437,17 @@ libm-routines = $(strip $(libm-support)                                    \
>  # and we don't want to have to link every program with -lm.
>  # In libm-calls (above), list m_foo in place of s_foo for any
>  # routine that should be compiled separately for its libc and libm versions.
> -calls = s_isinfF s_isnanF s_finiteF s_copysignF s_modfF s_scalbnF s_frexpF \
> -       s_signbitF $(gen-calls)
> +calls = \
> +  $(gen-calls) \
> +  s_copysignF \
> +  s_finiteF \
> +  s_frexpF \
> +  s_isinfF \
> +  s_isnanF \
> +  s_modfF \
> +  s_scalbnF \
> +  s_signbitF \
> +  # calls
>  gen-calls = s_ldexpF
>  generated += $(foreach s,.c .S,$(call type-foreach, $(calls:s_%=m_%$(s))))
>  routines = $(call type-foreach, $(calls))
> @@ -205,8 +458,14 @@ libm-shared-only-routines = $(call type-foreach, $(calls:s_%=m_%))
>  ifeq ($(build-mathvec),yes)
>  # We need to install libm.so and libm.a as linker scripts
>  # for transparent use of vector math library.
> -install-lib-ldscripts := libm.so libm.a
> -install-others = $(inst_libdir)/libm.so $(inst_libdir)/libm.a
> +install-lib-ldscripts := \
> +  libm.a \
> +  libm.so \
> +  # install-lib-ldscripts
> +install-others = \
> +  $(inst_libdir)/libm.a \
> +  $(inst_libdir)/libm.so \
> +  # install-others
>  $(inst_libdir)/libm.so: $(common-objpfx)format.lds \
>                         $(libm) \
>                         $(libmvec) \
> @@ -234,42 +493,95 @@ $(inst_libdir)/libm.a: $(common-objpfx)format.lds \
>  endif
>
>  # Rules for the test suite.
> -tests = test-matherr-3 test-fenv basic-test \
> -       test-misc test-fpucw test-fpucw-ieee tst-definitions test-tgmath \
> -       test-tgmath-ret bug-nextafter bug-nexttoward bug-tgmath1 \
> -       test-tgmath-int test-tgmath2 test-powl tst-CMPLX tst-CMPLX2 test-snan \
> -       test-fenv-tls test-fenv-preserve test-fenv-return \
> -       test-nearbyint-except test-fenv-clear \
> -       test-nearbyint-except-2 test-signgam-uchar test-signgam-uchar-init \
> -       test-signgam-uint test-signgam-uint-init test-signgam-ullong \
> -       test-signgam-ullong-init test-nan-overflow test-nan-payload \
> -       test-fexcept test-fexcept-traps test-fesetexcept \
> -       test-fesetexcept-traps test-fetestexceptflag test-femode \
> -       test-femode-traps test-iszero-excess-precision \
> -       test-iseqsig-excess-precision test-flt-eval-method \
> -       test-fp-ilogb-constants test-fp-llogb-constants \
> -       test-narrow-macros \
> -       test-nan-const $(tests-static)
> +tests = \
> +  $(tests-static) \
> +  bug-nextafter \
> +  bug-nexttoward \
> +  bug-tgmath1 \
> +  test-femode \
> +  test-femode-traps \
> +  test-fenv basic-test \
> +  test-fenv-clear \
> +  test-fenv-preserve \
> +  test-fenv-return \
> +  test-fenv-tls \
> +  test-fesetexcept \
> +  test-fesetexcept-traps \
> +  test-fetestexceptflag \
> +  test-fexcept \
> +  test-fexcept-traps \
> +  test-flt-eval-method \
> +  test-fp-ilogb-constants \
> +  test-fp-llogb-constants \
> +  test-fpucw \
> +  test-fpucw-ieee \
> +  test-iseqsig-excess-precision \
> +  test-iszero-excess-precision \
> +  test-matherr-3 \
> +  test-misc \
> +  test-nan-const \
> +  test-nan-overflow \
> +  test-nan-payload \
> +  test-narrow-macros \
> +  test-nearbyint-except \
> +  test-nearbyint-except-2 \
> +  test-powl \
> +  test-signgam-uchar \
> +  test-signgam-uchar-init \
> +  test-signgam-uint \
> +  test-signgam-uint-init \
> +  test-signgam-ullong \
> +  test-signgam-ullong-init \
> +  test-snan \
> +  test-tgmath \
> +  test-tgmath-int \
> +  test-tgmath-ret \
> +  test-tgmath2 \
> +  tst-CMPLX \
> +  tst-CMPLX2 \
> +  tst-definitions \
> +  # tests
>  ifneq ($(config-cflags-signaling-nans),)
> -tests += test-fe-snans-always-signal
> +tests += \
> +  test-fe-snans-always-signal \
> +  # tests
>  endif
> -tests-static = test-fpucw-static test-fpucw-ieee-static \
> -              test-signgam-uchar-static test-signgam-uchar-init-static \
> -              test-signgam-uint-static test-signgam-uint-init-static \
> -              test-signgam-ullong-static test-signgam-ullong-init-static
> +tests-static = \
> +  test-fpucw-ieee-static \
> +  test-fpucw-static \
> +  test-signgam-uchar-init-static \
> +  test-signgam-uchar-static \
> +  test-signgam-uint-init-static \
> +  test-signgam-uint-static \
> +  test-signgam-ullong-init-static \
> +  test-signgam-ullong-static \
> +  # tests-static
>
>  # The tested symbols matherr, _LIB_VERSION have been removed in glibc 2.27.
>  ifeq ($(have-GLIBC_2.26)$(build-shared),yesyes)
> -tests += test-matherr test-matherr-2
> +tests += \
> +  test-matherr \
> +  test-matherr-2 \
> +  # tests
>  endif
>
>  # These tests use internal (unexported) GMP functions and are linked
>  # statically to obtain access to these functions.
> -tests-static += atest-exp atest-sincos atest-exp2
> +tests-static += \
> +  atest-exp \
> +  atest-exp2 \
> +  atest-sincos \
> +  # tests-static
>
>  ifneq (,$(CXX))
> -tests += test-math-isinff test-math-iszero test-math-issignaling \
> -        test-math-iscanonical test-math-cxx11 test-math-iseqsig
> +tests += \
> +  test-math-cxx11 \
> +  test-math-iscanonical \
> +  test-math-iseqsig \
> +  test-math-isinff \
> +  test-math-issignaling \
> +  test-math-iszero \
> +  # tests
>  endif
>
>  libm-vec-tests = $(addprefix test-,$(libmvec-tests))
> @@ -286,35 +598,167 @@ $(objpfx)libm-test-ulps.h: $(ulps-file) gen-libm-test.py
>         $(make-target-directory)
>         $(PYTHON) gen-libm-test.py -u $< -H $@
>
> -libm-test-funcs-auto = acos acosh asin asinh atan atan2 atanh cabs cacos \
> -                      cacosh carg casin casinh catan catanh cbrt ccos ccosh \
> -                      cexp clog clog10 cos cosh cpow csin csinh csqrt ctan \
> -                      ctanh erf erfc exp exp10 exp2 expm1 fma hypot j0 j1 jn \
> -                      lgamma log log10 log1p log2 pow sin sincos sinh sqrt \
> -                      tan tanh tgamma y0 y1 yn
> -libm-test-funcs-noauto = canonicalize ceil cimag conj copysign cproj creal \
> -                        fabs fdim floor fmax fmaximum fmaximum_mag \
> -                        fmaximum_mag_num fmaximum_num fmaxmag fmin fminimum \
> -                        fminimum_mag fminimum_mag_num fminimum_num fminmag \
> -                        fmod fpclassify frexp fromfp fromfpx getpayload \
> -                        ilogb iscanonical iseqsig isfinite isgreater \
> -                        isgreaterequal isinf isless islessequal \
> -                        islessgreater isnan isnormal issignaling issubnormal \
> -                        isunordered iszero llogb llrint llround logb lrint \
> -                        lround modf nearbyint nextafter nextdown nexttoward \
> -                        nextup remainder remquo rint round roundeven scalb \
> -                        scalbln scalbn setpayload setpayloadsig signbit \
> -                        significand totalorder totalordermag trunc ufromfp \
> -                        ufromfpx compat_totalorder compat_totalordermag
> -libm-test-funcs-compat = compat_totalorder compat_totalordermag
> -libm-test-funcs-narrow = add div fma mul sqrt sub
> -libm-test-funcs-all = $(libm-test-funcs-auto) $(libm-test-funcs-noauto)
> +libm-test-funcs-auto = \
> +  acos \
> +  acosh \
> +  asin \
> +  asinh \
> +  atan \
> +  atan2 \
> +  atanh \
> +  cabs \
> +  cacos \
> +  cacosh \
> +  carg \
> +  casin \
> +  casinh \
> +  catan \
> +  catanh \
> +  cbrt \
> +  ccos \
> +  ccosh \
> +  cexp \
> +  clog \
> +  clog10 \
> +  cos \
> +  cosh \
> +  cpow \
> +  csin \
> +  csinh \
> +  csqrt \
> +  ctan \
> +  ctanh \
> +  erf \
> +  erfc \
> +  exp \
> +  exp2 \
> +  exp10 \
> +  expm1 \
> +  fma \
> +  hypot \
> +  j0 \
> +  j1 \
> +  jn \
> +  lgamma \
> +  log \
> +  log10 \
> +  log1p \
> +  log2 \
> +  pow \
> +  sin \
> +  sincos \
> +  sinh \
> +  sqrt \
> +  tan \
> +  tanh \
> +  tgamma \
> +  y0 \
> +  y1 \
> +  yn \
> +  # libm-test-funcs-auto
> +libm-test-funcs-noauto = \
> +  canonicalize \
> +  ceil \
> +  cimag \
> +  compat_totalorder \
> +  compat_totalordermag \
> +  conj \
> +  copysign \
> +  cproj \
> +  creal \
> +  fabs \
> +  fdim \
> +  floor \
> +  fmax \
> +  fmaximum \
> +  fmaximum_mag \
> +  fmaximum_mag_num \
> +  fmaximum_num \
> +  fmaxmag \
> +  fmin \
> +  fminimum \
> +  fminimum_mag \
> +  fminimum_mag_num \
> +  fminimum_num \
> +  fminmag \
> +  fmod \
> +  fpclassify \
> +  frexp \
> +  fromfp \
> +  fromfpx \
> +  getpayload \
> +  ilogb \
> +  iscanonical \
> +  iseqsig \
> +  isfinite \
> +  isgreater \
> +  isgreaterequal \
> +  isinf \
> +  isless \
> +  islessequal \
> +  islessgreater \
> +  isnan \
> +  isnormal \
> +  issignaling \
> +  issubnormal \
> +  isunordered \
> +  iszero \
> +  llogb \
> +  llrint \
> +  llround \
> +  logb \
> +  lrint \
> +  lround \
> +  modf \
> +  nearbyint \
> +  nextafter \
> +  nextdown \
> +  nexttoward \
> +  nextup \
> +  remainder \
> +  remquo \
> +  rint \
> +  round \
> +  roundeven \
> +  scalb \
> +  scalbln \
> +  scalbn \
> +  setpayload \
> +  setpayloadsig \
> +  signbit \
> +  significand \
> +  totalorder \
> +  totalordermag \
> +  trunc \
> +  ufromfp \
> +  ufromfpx \
> +  # libm-test-funcs-noauto
> +libm-test-funcs-compat = \
> +  compat_totalorder \
> +  compat_totalordermag \
> +  # libm-test-funcs-compat
> +libm-test-funcs-narrow = \
> +  add \
> +  div \
> +  fma \
> +  mul \
> +  sqrt \
> +  sub \
> +  # libm-test-funcs-narrow
> +libm-test-funcs-all = \
> +  $(libm-test-funcs-auto) \
> +  $(libm-test-funcs-noauto) \
> +  # libm-test-funcs-all
>  libm-test-c-auto = $(foreach f,$(libm-test-funcs-auto),libm-test-$(f).c)
>  libm-test-c-noauto = $(foreach f,$(libm-test-funcs-noauto),libm-test-$(f).c)
>  libm-test-c-narrow = $(foreach f,$(libm-test-funcs-narrow),\
>                                  libm-test-narrow-$(f).c)
> -generated += libm-test-ulps.h $(libm-test-c-auto) $(libm-test-c-noauto) \
> -            $(libm-test-c-narrow)
> +generated += \
> +  $(libm-test-c-auto) \
> +  $(libm-test-c-narrow) \
> +  $(libm-test-c-noauto) \
> +  libm-test-ulps.h \
> +  # generated
>
>  libm-tests-base-normal = $(foreach t,$(test-types),test-$(t))
>  libm-tests-base-narrow = $(foreach t,$(test-type-pairs),test-$(t))
> @@ -374,19 +818,101 @@ $(foreach t,$(libm-tests-base),\
>             $(objpfx)$(t)-compat_totalordermag.o): $(objpfx)libm-test-totalordermag.c
>
>  # _Float128x omitted as not supported by gen-tgmath-tests.py.
> -tgmath3-narrow-types = f d f16 f32 f64 f128 f32x f64x
> +tgmath3-narrow-types = \
> +  d \
> +  f \
> +  f16 \
> +  f32 \
> +  f128 \
> +  f32x \
> +  f64 \
> +  f64x \
> +  # tgmath3-narrow-types
>  tgmath3-narrow-macros = $(foreach t,$(tgmath3-narrow-types), \
>                                     $(foreach f,$(libm-narrow-fns),$(t)$(f)))
> -tgmath3-macros = atan2 cbrt ceil copysign erf erfc exp10 exp2 expm1 fdim \
> -                floor fma fmax fmin fmod frexp hypot ilogb ldexp lgamma \
> -                llrint llround log10 log1p log2 logb lrint lround nearbyint \
> -                nextafter nexttoward remainder remquo rint round scalbn \
> -                scalbln tgamma trunc acos asin atan acosh asinh atanh cos \
> -                sin tan cosh sinh tanh exp log pow sqrt fabs carg cimag conj \
> -                cproj creal roundeven nextup nextdown fminmag fmaxmag \
> -                fmaximum fmaximum_mag fmaximum_num fmaximum_mag_num \
> -                fminimum fminimum_mag fminimum_num fminimum_mag_num llogb \
> -                fromfp fromfpx ufromfp ufromfpx scalb $(tgmath3-narrow-macros)
> +tgmath3-macros = \
> +  $(tgmath3-narrow-macros) \
> +  acos \
> +  acosh \
> +  asin \
> +  asinh \
> +  atan \
> +  atan2 \
> +  atanh \
> +  carg \
> +  cbrt \
> +  ceil \
> +  cimag \
> +  conj \
> +  copysign \
> +  cos \
> +  cosh \
> +  cproj \
> +  creal \
> +  erf \
> +  erfc \
> +  exp \
> +  exp2 \
> +  exp10 \
> +  expm1 \
> +  fabs \
> +  fdim \
> +  floor \
> +  fma \
> +  fmax \
> +  fmaximum \
> +  fmaximum_mag \
> +  fmaximum_mag_num \
> +  fmaximum_num \
> +  fmaxmag \
> +  fmin \
> +  fminimum \
> +  fminimum_mag \
> +  fminimum_mag_num \
> +  fminimum_num \
> +  fminmag \
> +  fmod \
> +  frexp \
> +  fromfp \
> +  fromfpx \
> +  hypot \
> +  ilogb \
> +  ldexp \
> +  lgamma \
> +  llogb \
> +  llrint \
> +  llround \
> +  log \
> +  log10 \
> +  log1p \
> +  log2 \
> +  logb \
> +  lrint \
> +  lround \
> +  nearbyint \
> +  nextafter \
> +  nextdown \
> +  nexttoward \
> +  nextup \
> +  pow \
> +  remainder \
> +  remquo \
> +  rint \
> +  round \
> +  roundeven \
> +  scalb \
> +  scalbln \
> +  scalbn \
> +  sin \
> +  sinh \
> +  sqrt \
> +  tan \
> +  tanh \
> +  tgamma \
> +  trunc \
> +  ufromfp \
> +  ufromfpx \
> +  # tgmath3-macros
>  tgmath3-macro-tests = $(addprefix test-tgmath3-,$(tgmath3-macros))
>  tests += $(tgmath3-macro-tests)
>  generated += $(addsuffix .c,$(tgmath3-macro-tests))
> --
> 2.34.1
>

LGTM.

Reviewed-by: H.J. Lu <hjl.tools@gmail.com>

Thanks.

-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 02/12] powerpc: Add missing arch flags on rounding ifunc variants
  2023-12-28 17:20 ` [PATCH 02/12] powerpc: Add missing arch flags on rounding ifunc variants Adhemerval Zanella
@ 2024-04-01 13:59   ` H.J. Lu
  2024-04-01 16:50     ` Peter Bergner
  0 siblings, 1 reply; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 13:59 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:21 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> The ifunc variants now uses the powerpc implementation which in turn
> uses the compiler builtin.  Without the proper -mcpu switch the builtin
> does not generate the expected optimization.
>
> Checked on powerpc-linux-gnu.
> ---
>  sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile | 6 ++++++
>  1 file changed, 6 insertions(+)
>
> diff --git a/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile b/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile
> index 64317506c6..2f5c0eded6 100644
> --- a/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile
> +++ b/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile
> @@ -34,6 +34,12 @@ CFLAGS-s_modff-power5+.c = -mcpu=power5+
>  CFLAGS-s_logbl-power7.c = -mcpu=power7
>  CFLAGS-s_logb-power7.c = -mcpu=power7
>  CFLAGS-s_logbf-power7.c = -mcpu=power7
> +CFLAGS-s_round-power5+.c += -mcpu=power5+
> +CFLAGS-s_roundf-power5+.c += -mcpu=power5+
> +CFLAGS-s_floor-power5+.c += -mcpu=power5+
> +CFLAGS-s_floorf-power5+.c += -mcpu=power5+
> +CFLAGS-s_trunc-power5+.c += -mcpu=power5+
> +CFLAGS-s_truncf-power5+.c += -mcpu=power5+
>
>  # These files quiet sNaNs in a way that is optimized away without
>  # -fsignaling-nans.
> --
> 2.34.1
>

LGTM.

Reviewed-by: H.J. Lu <hjl.tools@gmail.com>

Thanks.

-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 03/12] math: Add test to check if ceil raise inexact floating-point exception
  2023-12-28 17:20 ` [PATCH 03/12] math: Add test to check if ceil raise inexact floating-point exception Adhemerval Zanella
@ 2024-04-01 14:01   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:01 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:22 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
> ---
>  math/Makefile             |  5 +++
>  math/test-ceil-except-2.c | 67 ++++++++++++++++++++++++++++++
>  math/test-ceil-except.c   | 85 +++++++++++++++++++++++++++++++++++++++
>  3 files changed, 157 insertions(+)
>  create mode 100644 math/test-ceil-except-2.c
>  create mode 100644 math/test-ceil-except.c
>
> diff --git a/math/Makefile b/math/Makefile
> index 5a912c55bd..4fc7842313 100644
> --- a/math/Makefile
> +++ b/math/Makefile
> @@ -498,6 +498,8 @@ tests = \
>    bug-nextafter \
>    bug-nexttoward \
>    bug-tgmath1 \
> +  test-ceil-except \
> +  test-ceil-except-2 \
>    test-femode \
>    test-femode-traps \
>    test-fenv basic-test \
> @@ -989,6 +991,9 @@ CFLAGS-test-fe-snans-always-signal.c += $(config-cflags-signaling-nans)
>
>  CFLAGS-test-nan-const.c += -fno-builtin
>
> +CFLAGS-test-ceil-except.c += -fno-builtin
> +CFLAGS-test-ceil-except-2.c += -fno-builtin
> +
>  include ../Rules
>
>  gen-all-calls = $(gen-libm-calls) $(gen-calls)
> diff --git a/math/test-ceil-except-2.c b/math/test-ceil-except-2.c
> new file mode 100644
> index 0000000000..c25956a402
> --- /dev/null
> +++ b/math/test-ceil-except-2.c
> @@ -0,0 +1,67 @@
> +/* Test ceil functions do not disable exception traps.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <fenv.h>
> +#include <math.h>
> +#include <stdio.h>
> +
> +#ifndef FE_INEXACT
> +# define FE_INEXACT 0
> +#endif
> +
> +#define TEST_FUNC(NAME, FLOAT, SUFFIX)                                 \
> +static int                                                             \
> +NAME (void)                                                            \
> +{                                                                      \
> +  int result = 0;                                                      \
> +  volatile FLOAT a, b __attribute__ ((unused));                                \
> +  a = 1.5;                                                             \
> +  /* ceil must work when traps on "inexact" are enabled.  */   \
> +  b = ceil ## SUFFIX (a);                                              \
> +  /* And it must have left those traps enabled.  */                    \
> +  if (fegetexcept () == FE_INEXACT)                                    \
> +    puts ("PASS: " #FLOAT);                                            \
> +  else                                                                 \
> +    {                                                                  \
> +      puts ("FAIL: " #FLOAT);                                          \
> +      result = 1;                                                      \
> +    }                                                                  \
> +  return result;                                                       \
> +}
> +
> +TEST_FUNC (float_test, float, f)
> +TEST_FUNC (double_test, double, )
> +TEST_FUNC (ldouble_test, long double, l)
> +
> +static int
> +do_test (void)
> +{
> +  if (feenableexcept (FE_INEXACT) == -1)
> +    {
> +      puts ("enabling FE_INEXACT traps failed, cannot test");
> +      return 77;
> +    }
> +  int result = float_test ();
> +  feenableexcept (FE_INEXACT);
> +  result |= double_test ();
> +  feenableexcept (FE_INEXACT);
> +  result |= ldouble_test ();
> +  return result;
> +}
> +
> +#include <support/test-driver.c>
> diff --git a/math/test-ceil-except.c b/math/test-ceil-except.c
> new file mode 100644
> index 0000000000..cd0717a96f
> --- /dev/null
> +++ b/math/test-ceil-except.c
> @@ -0,0 +1,85 @@
> +/* Test ceil functions do not clear exceptions.
> +   Copyright (C) 2015-2023 Free Software Foundation, Inc.

Just 2024.

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <fenv.h>
> +#include <math.h>
> +#include <stdbool.h>
> +#include <stdio.h>
> +
> +#include <math-tests.h>
> +
> +#ifndef FE_INVALID
> +# define FE_INVALID 0
> +#endif
> +
> +static bool any_supported = false;
> +
> +#define TEST_FUNC(NAME, FLOAT, SUFFIX)                                 \
> +static int                                                             \
> +NAME (void)                                                            \
> +{                                                                      \
> +  int result = 0;                                                      \
> +  if (!EXCEPTION_TESTS (FLOAT))                                                \
> +    return 0;                                                          \
> +  any_supported = true;                                                        \
> +  volatile FLOAT a, b __attribute__ ((unused));                                \
> +  a = 1.0;                                                             \
> +  /* ceil must not clear already-raised exceptions.  */                \
> +  feraiseexcept (FE_ALL_EXCEPT);                                       \
> +  b = ceil ## SUFFIX (a);                                              \
> +  if (fetestexcept (FE_ALL_EXCEPT) == FE_ALL_EXCEPT)                   \
> +    puts ("PASS: " #FLOAT);                                            \
> +  else                                                                 \
> +    {                                                                  \
> +      puts ("FAIL: " #FLOAT);                                          \
> +      result = 1;                                                      \
> +    }                                                                  \
> +  /* But it mustn't lose exceptions from sNaN arguments.  */           \
> +  if (SNAN_TESTS (FLOAT))                                              \
> +    {                                                                  \
> +      static volatile FLOAT snan = __builtin_nans ## SUFFIX ("");      \
> +      volatile FLOAT c __attribute__ ((unused));                       \
> +      feclearexcept (FE_ALL_EXCEPT);                                   \
> +      c = ceil ## SUFFIX (snan);                                       \
> +      if (fetestexcept (FE_INVALID) == FE_INVALID)                     \
> +       puts ("PASS: " #FLOAT " sNaN");                                 \
> +      else                                                             \
> +       {                                                               \
> +         puts ("FAIL: " #FLOAT " sNaN");                               \
> +         result = 1;                                                   \
> +       }                                                               \
> +    }                                                                  \
> +  return result;                                                       \
> +}
> +
> +TEST_FUNC (float_test, float, f)
> +TEST_FUNC (double_test, double, )
> +TEST_FUNC (ldouble_test, long double, l)
> +
> +static int
> +do_test (void)
> +{
> +  int result = float_test ();
> +  result |= double_test ();
> +  result |= ldouble_test ();
> +  if (!any_supported)
> +    return 77;
> +  return result;
> +}
> +
> +#include <support/test-driver.c>
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 04/12] math: Add test to check if floor raise inexact floating-point exception
  2023-12-28 17:20 ` [PATCH 04/12] math: Add test to check if floor " Adhemerval Zanella
@ 2024-04-01 14:02   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:02 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:22 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
> ---
>  math/Makefile              |  4 ++
>  math/test-floor-except-2.c | 67 ++++++++++++++++++++++++++++++
>  math/test-floor-except.c   | 85 ++++++++++++++++++++++++++++++++++++++
>  3 files changed, 156 insertions(+)
>  create mode 100644 math/test-floor-except-2.c
>  create mode 100644 math/test-floor-except.c
>
> diff --git a/math/Makefile b/math/Makefile
> index 4fc7842313..19e234874a 100644
> --- a/math/Makefile
> +++ b/math/Makefile
> @@ -512,6 +512,8 @@ tests = \
>    test-fetestexceptflag \
>    test-fexcept \
>    test-fexcept-traps \
> +  test-floor-except \
> +  test-floor-except-2 \
>    test-flt-eval-method \
>    test-fp-ilogb-constants \
>    test-fp-llogb-constants \
> @@ -993,6 +995,8 @@ CFLAGS-test-nan-const.c += -fno-builtin
>
>  CFLAGS-test-ceil-except.c += -fno-builtin
>  CFLAGS-test-ceil-except-2.c += -fno-builtin
> +CFLAGS-test-floor-except.c += -fno-builtin
> +CFLAGS-test-floor-except-2.c += -fno-builtin
>
>  include ../Rules
>
> diff --git a/math/test-floor-except-2.c b/math/test-floor-except-2.c
> new file mode 100644
> index 0000000000..01b22f984b
> --- /dev/null
> +++ b/math/test-floor-except-2.c
> @@ -0,0 +1,67 @@
> +/* Test floor functions do not disable exception traps.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <fenv.h>
> +#include <math.h>
> +#include <stdio.h>
> +
> +#ifndef FE_INEXACT
> +# define FE_INEXACT 0
> +#endif
> +
> +#define TEST_FUNC(NAME, FLOAT, SUFFIX)                                 \
> +static int                                                             \
> +NAME (void)                                                            \
> +{                                                                      \
> +  int result = 0;                                                      \
> +  volatile FLOAT a, b __attribute__ ((unused));                                \
> +  a = 1.5;                                                             \
> +  /* floor must work when traps on "inexact" are enabled.  */  \
> +  b = floor ## SUFFIX (a);                                             \
> +  /* And it must have left those traps enabled.  */                    \
> +  if (fegetexcept () == FE_INEXACT)                                    \
> +    puts ("PASS: " #FLOAT);                                            \
> +  else                                                                 \
> +    {                                                                  \
> +      puts ("FAIL: " #FLOAT);                                          \
> +      result = 1;                                                      \
> +    }                                                                  \
> +  return result;                                                       \
> +}
> +
> +TEST_FUNC (float_test, float, f)
> +TEST_FUNC (double_test, double, )
> +TEST_FUNC (ldouble_test, long double, l)
> +
> +static int
> +do_test (void)
> +{
> +  if (feenableexcept (FE_INEXACT) == -1)
> +    {
> +      puts ("enabling FE_INEXACT traps failed, cannot test");
> +      return 77;
> +    }
> +  int result = float_test ();
> +  feenableexcept (FE_INEXACT);
> +  result |= double_test ();
> +  feenableexcept (FE_INEXACT);
> +  result |= ldouble_test ();
> +  return result;
> +}
> +
> +#include <support/test-driver.c>
> diff --git a/math/test-floor-except.c b/math/test-floor-except.c
> new file mode 100644
> index 0000000000..fe77866ddc
> --- /dev/null
> +++ b/math/test-floor-except.c
> @@ -0,0 +1,85 @@
> +/* Test floor functions do not clear exceptions.
> +   Copyright (C) 2015-2023 Free Software Foundation, Inc.

Just 2024.

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <fenv.h>
> +#include <math.h>
> +#include <stdbool.h>
> +#include <stdio.h>
> +
> +#include <math-tests.h>
> +
> +#ifndef FE_INVALID
> +# define FE_INVALID 0
> +#endif
> +
> +static bool any_supported = false;
> +
> +#define TEST_FUNC(NAME, FLOAT, SUFFIX)                                 \
> +static int                                                             \
> +NAME (void)                                                            \
> +{                                                                      \
> +  int result = 0;                                                      \
> +  if (!EXCEPTION_TESTS (FLOAT))                                                \
> +    return 0;                                                          \
> +  any_supported = true;                                                        \
> +  volatile FLOAT a, b __attribute__ ((unused));                                \
> +  a = 1.0;                                                             \
> +  /* floor must not clear already-raised exceptions.  */               \
> +  feraiseexcept (FE_ALL_EXCEPT);                                       \
> +  b = floor ## SUFFIX (a);                                             \
> +  if (fetestexcept (FE_ALL_EXCEPT) == FE_ALL_EXCEPT)                   \
> +    puts ("PASS: " #FLOAT);                                            \
> +  else                                                                 \
> +    {                                                                  \
> +      puts ("FAIL: " #FLOAT);                                          \
> +      result = 1;                                                      \
> +    }                                                                  \
> +  /* But it mustn't lose exceptions from sNaN arguments.  */           \
> +  if (SNAN_TESTS (FLOAT))                                              \
> +    {                                                                  \
> +      static volatile FLOAT snan = __builtin_nans ## SUFFIX ("");      \
> +      volatile FLOAT c __attribute__ ((unused));                       \
> +      feclearexcept (FE_ALL_EXCEPT);                                   \
> +      c = floor ## SUFFIX (snan);                                      \
> +      if (fetestexcept (FE_INVALID) == FE_INVALID)                     \
> +       puts ("PASS: " #FLOAT " sNaN");                                 \
> +      else                                                             \
> +       {                                                               \
> +         puts ("FAIL: " #FLOAT " sNaN");                               \
> +         result = 1;                                                   \
> +       }                                                               \
> +    }                                                                  \
> +  return result;                                                       \
> +}
> +
> +TEST_FUNC (float_test, float, f)
> +TEST_FUNC (double_test, double, )
> +TEST_FUNC (ldouble_test, long double, l)
> +
> +static int
> +do_test (void)
> +{
> +  int result = float_test ();
> +  result |= double_test ();
> +  result |= ldouble_test ();
> +  if (!any_supported)
> +    return 77;
> +  return result;
> +}
> +
> +#include <support/test-driver.c>
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 06/12] math: Add test to check if round raise inexact floating-point exception
  2023-12-28 17:20 ` [PATCH 06/12] math: Add test to check if round " Adhemerval Zanella
@ 2024-04-01 14:02   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:02 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:22 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
> ---
>  math/Makefile              |  4 ++
>  math/test-round-except-2.c | 67 ++++++++++++++++++++++++++++++
>  math/test-round-except.c   | 85 ++++++++++++++++++++++++++++++++++++++
>  3 files changed, 156 insertions(+)
>  create mode 100644 math/test-round-except-2.c
>  create mode 100644 math/test-round-except.c
>
> diff --git a/math/Makefile b/math/Makefile
> index 7ae107170c..dfe24a8c03 100644
> --- a/math/Makefile
> +++ b/math/Makefile
> @@ -530,6 +530,8 @@ tests = \
>    test-nearbyint-except \
>    test-nearbyint-except-2 \
>    test-powl \
> +  test-round-except \
> +  test-round-except-2 \
>    test-signgam-uchar \
>    test-signgam-uchar-init \
>    test-signgam-uint \
> @@ -1001,6 +1003,8 @@ CFLAGS-test-floor-except.c += -fno-builtin
>  CFLAGS-test-floor-except-2.c += -fno-builtin
>  CFLAGS-test-trunc-except.c += -fno-builtin
>  CFLAGS-test-trunc-except-2.c += -fno-builtin
> +CFLAGS-test-round-except.c += -fno-builtin
> +CFLAGS-test-round-except-2.c += -fno-builtin
>
>  include ../Rules
>
> diff --git a/math/test-round-except-2.c b/math/test-round-except-2.c
> new file mode 100644
> index 0000000000..dffc02e91f
> --- /dev/null
> +++ b/math/test-round-except-2.c
> @@ -0,0 +1,67 @@
> +/* Test round functions do not disable exception traps.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <fenv.h>
> +#include <math.h>
> +#include <stdio.h>
> +
> +#ifndef FE_INEXACT
> +# define FE_INEXACT 0
> +#endif
> +
> +#define TEST_FUNC(NAME, FLOAT, SUFFIX)                                 \
> +static int                                                             \
> +NAME (void)                                                            \
> +{                                                                      \
> +  int result = 0;                                                      \
> +  volatile FLOAT a, b __attribute__ ((unused));                                \
> +  a = 1.5;                                                             \
> +  /* round must work when traps on "inexact" are enabled.  */  \
> +  b = round ## SUFFIX (a);                                             \
> +  /* And it must have left those traps enabled.  */                    \
> +  if (fegetexcept () == FE_INEXACT)                                    \
> +    puts ("PASS: " #FLOAT);                                            \
> +  else                                                                 \
> +    {                                                                  \
> +      puts ("FAIL: " #FLOAT);                                          \
> +      result = 1;                                                      \
> +    }                                                                  \
> +  return result;                                                       \
> +}
> +
> +TEST_FUNC (float_test, float, f)
> +TEST_FUNC (double_test, double, )
> +TEST_FUNC (ldouble_test, long double, l)
> +
> +static int
> +do_test (void)
> +{
> +  if (feenableexcept (FE_INEXACT) == -1)
> +    {
> +      puts ("enabling FE_INEXACT traps failed, cannot test");
> +      return 77;
> +    }
> +  int result = float_test ();
> +  feenableexcept (FE_INEXACT);
> +  result |= double_test ();
> +  feenableexcept (FE_INEXACT);
> +  result |= ldouble_test ();
> +  return result;
> +}
> +
> +#include <support/test-driver.c>
> diff --git a/math/test-round-except.c b/math/test-round-except.c
> new file mode 100644
> index 0000000000..cd43c964ab
> --- /dev/null
> +++ b/math/test-round-except.c
> @@ -0,0 +1,85 @@
> +/* Test round functions do not clear exceptions.
> +   Copyright (C) 2015-2023 Free Software Foundation, Inc.

Just 2024.

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <fenv.h>
> +#include <math.h>
> +#include <stdbool.h>
> +#include <stdio.h>
> +
> +#include <math-tests.h>
> +
> +#ifndef FE_INVALID
> +# define FE_INVALID 0
> +#endif
> +
> +static bool any_supported = false;
> +
> +#define TEST_FUNC(NAME, FLOAT, SUFFIX)                                 \
> +static int                                                             \
> +NAME (void)                                                            \
> +{                                                                      \
> +  int result = 0;                                                      \
> +  if (!EXCEPTION_TESTS (FLOAT))                                                \
> +    return 0;                                                          \
> +  any_supported = true;                                                        \
> +  volatile FLOAT a, b __attribute__ ((unused));                                \
> +  a = 1.0;                                                             \
> +  /* round must not clear already-raised exceptions.  */               \
> +  feraiseexcept (FE_ALL_EXCEPT);                                       \
> +  b = round ## SUFFIX (a);                                             \
> +  if (fetestexcept (FE_ALL_EXCEPT) == FE_ALL_EXCEPT)                   \
> +    puts ("PASS: " #FLOAT);                                            \
> +  else                                                                 \
> +    {                                                                  \
> +      puts ("FAIL: " #FLOAT);                                          \
> +      result = 1;                                                      \
> +    }                                                                  \
> +  /* But it mustn't lose exceptions from sNaN arguments.  */           \
> +  if (SNAN_TESTS (FLOAT))                                              \
> +    {                                                                  \
> +      static volatile FLOAT snan = __builtin_nans ## SUFFIX ("");      \
> +      volatile FLOAT c __attribute__ ((unused));                       \
> +      feclearexcept (FE_ALL_EXCEPT);                                   \
> +      c = round ## SUFFIX (snan);                                      \
> +      if (fetestexcept (FE_INVALID) == FE_INVALID)                     \
> +       puts ("PASS: " #FLOAT " sNaN");                                 \
> +      else                                                             \
> +       {                                                               \
> +         puts ("FAIL: " #FLOAT " sNaN");                               \
> +         result = 1;                                                   \
> +       }                                                               \
> +    }                                                                  \
> +  return result;                                                       \
> +}
> +
> +TEST_FUNC (float_test, float, f)
> +TEST_FUNC (double_test, double, )
> +TEST_FUNC (ldouble_test, long double, l)
> +
> +static int
> +do_test (void)
> +{
> +  int result = float_test ();
> +  result |= double_test ();
> +  result |= ldouble_test ();
> +  if (!any_supported)
> +    return 77;
> +  return result;
> +}
> +
> +#include <support/test-driver.c>
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 05/12] math: Add test to check if trunc raise inexact floating-point exception
  2023-12-28 17:20 ` [PATCH 05/12] math: Add test to check if trunc " Adhemerval Zanella
@ 2024-04-01 14:03   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:03 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:22 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
> ---
>  math/Makefile              |  4 ++
>  math/test-trunc-except-2.c | 67 ++++++++++++++++++++++++++++++
>  math/test-trunc-except.c   | 85 ++++++++++++++++++++++++++++++++++++++
>  3 files changed, 156 insertions(+)
>  create mode 100644 math/test-trunc-except-2.c
>  create mode 100644 math/test-trunc-except.c
>
> diff --git a/math/Makefile b/math/Makefile
> index 19e234874a..7ae107170c 100644
> --- a/math/Makefile
> +++ b/math/Makefile
> @@ -541,6 +541,8 @@ tests = \
>    test-tgmath-int \
>    test-tgmath-ret \
>    test-tgmath2 \
> +  test-trunc-except \
> +  test-trunc-except-2 \
>    tst-CMPLX \
>    tst-CMPLX2 \
>    tst-definitions \
> @@ -997,6 +999,8 @@ CFLAGS-test-ceil-except.c += -fno-builtin
>  CFLAGS-test-ceil-except-2.c += -fno-builtin
>  CFLAGS-test-floor-except.c += -fno-builtin
>  CFLAGS-test-floor-except-2.c += -fno-builtin
> +CFLAGS-test-trunc-except.c += -fno-builtin
> +CFLAGS-test-trunc-except-2.c += -fno-builtin
>
>  include ../Rules
>
> diff --git a/math/test-trunc-except-2.c b/math/test-trunc-except-2.c
> new file mode 100644
> index 0000000000..29716efc0b
> --- /dev/null
> +++ b/math/test-trunc-except-2.c
> @@ -0,0 +1,67 @@
> +/* Test trunc functions do not disable exception traps.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <fenv.h>
> +#include <math.h>
> +#include <stdio.h>
> +
> +#ifndef FE_INEXACT
> +# define FE_INEXACT 0
> +#endif
> +
> +#define TEST_FUNC(NAME, FLOAT, SUFFIX)                                 \
> +static int                                                             \
> +NAME (void)                                                            \
> +{                                                                      \
> +  int result = 0;                                                      \
> +  volatile FLOAT a, b __attribute__ ((unused));                                \
> +  a = 1.5;                                                             \
> +  /* trunc must work when traps on "inexact" are enabled.  */  \
> +  b = trunc ## SUFFIX (a);                                             \
> +  /* And it must have left those traps enabled.  */                    \
> +  if (fegetexcept () == FE_INEXACT)                                    \
> +    puts ("PASS: " #FLOAT);                                            \
> +  else                                                                 \
> +    {                                                                  \
> +      puts ("FAIL: " #FLOAT);                                          \
> +      result = 1;                                                      \
> +    }                                                                  \
> +  return result;                                                       \
> +}
> +
> +TEST_FUNC (float_test, float, f)
> +TEST_FUNC (double_test, double, )
> +TEST_FUNC (ldouble_test, long double, l)
> +
> +static int
> +do_test (void)
> +{
> +  if (feenableexcept (FE_INEXACT) == -1)
> +    {
> +      puts ("enabling FE_INEXACT traps failed, cannot test");
> +      return 77;
> +    }
> +  int result = float_test ();
> +  feenableexcept (FE_INEXACT);
> +  result |= double_test ();
> +  feenableexcept (FE_INEXACT);
> +  result |= ldouble_test ();
> +  return result;
> +}
> +
> +#include <support/test-driver.c>
> diff --git a/math/test-trunc-except.c b/math/test-trunc-except.c
> new file mode 100644
> index 0000000000..c9e341de0d
> --- /dev/null
> +++ b/math/test-trunc-except.c
> @@ -0,0 +1,85 @@
> +/* Test trunc functions do not clear exceptions.
> +   Copyright (C) 2015-2023 Free Software Foundation, Inc.

Just 2024.

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <fenv.h>
> +#include <math.h>
> +#include <stdbool.h>
> +#include <stdio.h>
> +
> +#include <math-tests.h>
> +
> +#ifndef FE_INVALID
> +# define FE_INVALID 0
> +#endif
> +
> +static bool any_supported = false;
> +
> +#define TEST_FUNC(NAME, FLOAT, SUFFIX)                                 \
> +static int                                                             \
> +NAME (void)                                                            \
> +{                                                                      \
> +  int result = 0;                                                      \
> +  if (!EXCEPTION_TESTS (FLOAT))                                                \
> +    return 0;                                                          \
> +  any_supported = true;                                                        \
> +  volatile FLOAT a, b __attribute__ ((unused));                                \
> +  a = 1.0;                                                             \
> +  /* trunc must not clear already-raised exceptions.  */               \
> +  feraiseexcept (FE_ALL_EXCEPT);                                       \
> +  b = trunc ## SUFFIX (a);                                             \
> +  if (fetestexcept (FE_ALL_EXCEPT) == FE_ALL_EXCEPT)                   \
> +    puts ("PASS: " #FLOAT);                                            \
> +  else                                                                 \
> +    {                                                                  \
> +      puts ("FAIL: " #FLOAT);                                          \
> +      result = 1;                                                      \
> +    }                                                                  \
> +  /* But it mustn't lose exceptions from sNaN arguments.  */           \
> +  if (SNAN_TESTS (FLOAT))                                              \
> +    {                                                                  \
> +      static volatile FLOAT snan = __builtin_nans ## SUFFIX ("");      \
> +      volatile FLOAT c __attribute__ ((unused));                       \
> +      feclearexcept (FE_ALL_EXCEPT);                                   \
> +      c = trunc ## SUFFIX (snan);                                      \
> +      if (fetestexcept (FE_INVALID) == FE_INVALID)                     \
> +       puts ("PASS: " #FLOAT " sNaN");                                 \
> +      else                                                             \
> +       {                                                               \
> +         puts ("FAIL: " #FLOAT " sNaN");                               \
> +         result = 1;                                                   \
> +       }                                                               \
> +    }                                                                  \
> +  return result;                                                       \
> +}
> +
> +TEST_FUNC (float_test, float, f)
> +TEST_FUNC (double_test, double, )
> +TEST_FUNC (ldouble_test, long double, l)
> +
> +static int
> +do_test (void)
> +{
> +  int result = float_test ();
> +  result |= double_test ();
> +  result |= ldouble_test ();
> +  if (!any_supported)
> +    return 77;
> +  return result;
> +}
> +
> +#include <support/test-driver.c>
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 07/12] x86: Do not raise inexact exception on ceill
  2023-12-28 17:20 ` [PATCH 07/12] x86: Do not raise inexact exception on ceill Adhemerval Zanella
@ 2024-04-01 14:12   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:12 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:23 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
>
> Checked on x86_64-linux-gnu and i686-linux-gnu.
> ---
>  sysdeps/i386/fpu/s_ceill.S   | 39 ------------------------------------
>  sysdeps/x86/fpu/s_ceill.c    | 38 +++++++++++++++++++++++++++++++++++
>  sysdeps/x86_64/fpu/s_ceill.S | 34 -------------------------------
>  3 files changed, 38 insertions(+), 73 deletions(-)
>  delete mode 100644 sysdeps/i386/fpu/s_ceill.S
>  create mode 100644 sysdeps/x86/fpu/s_ceill.c
>  delete mode 100644 sysdeps/x86_64/fpu/s_ceill.S
>
> diff --git a/sysdeps/i386/fpu/s_ceill.S b/sysdeps/i386/fpu/s_ceill.S
> deleted file mode 100644
> index a551fce7f9..0000000000
> --- a/sysdeps/i386/fpu/s_ceill.S
> +++ /dev/null
> @@ -1,39 +0,0 @@
> -/*
> - * Public domain.
> - */
> -
> -#include <libm-alias-ldouble.h>
> -#include <machine/asm.h>
> -
> -RCSID("$NetBSD: $")
> -
> -ENTRY(__ceill)
> -       fldt    4(%esp)
> -       subl    $32,%esp
> -       cfi_adjust_cfa_offset (32)
> -
> -       fnstenv 4(%esp)                 /* store fpu environment */
> -
> -       /* We use here %edx although only the low 1 bits are defined.
> -          But none of the operations should care and they are faster
> -          than the 16 bit operations.  */
> -       movl    $0x0800,%edx            /* round towards +oo */
> -       orl     4(%esp),%edx
> -       andl    $0xfbff,%edx
> -       movl    %edx,(%esp)
> -       fldcw   (%esp)                  /* load modified control word */
> -
> -       frndint                         /* round */
> -
> -       /* Preserve "invalid" exceptions from sNaN input.  */
> -       fnstsw
> -       andl    $0x1, %eax
> -       orl     %eax, 8(%esp)
> -
> -       fldenv  4(%esp)                 /* restore original environment */
> -
> -       addl    $32,%esp
> -       cfi_adjust_cfa_offset (-32)
> -       ret
> -END (__ceill)
> -libm_alias_ldouble (__ceil, ceil)
> diff --git a/sysdeps/x86/fpu/s_ceill.c b/sysdeps/x86/fpu/s_ceill.c
> new file mode 100644
> index 0000000000..96c901e383
> --- /dev/null
> +++ b/sysdeps/x86/fpu/s_ceill.c
> @@ -0,0 +1,38 @@
> +/* Return smallest integral value not less than argument.  x86 version.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024.

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#define NO_MATH_REDIRECT
> +#include <math.h>
> +#include <fenv_private.h>
> +#include <libm-alias-ldouble.h>
> +
> +long double
> +__ceill (long double x)
> +{
> +  fenv_t fenv;
> +  long double r;
> +
> +  libc_feholdexcept_setround_387 (&fenv, FE_UPWARD);
> +  asm volatile ("frndint" : "=t" (r) : "0" (x));
> +  /* Preserve "invalid" exceptions from sNaN input.  */
> +  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
> +  libc_fesetenv_387 (&fenv);
> +
> +  return r;
> +}
> +libm_alias_ldouble (__ceil, ceil)
> diff --git a/sysdeps/x86_64/fpu/s_ceill.S b/sysdeps/x86_64/fpu/s_ceill.S
> deleted file mode 100644
> index 16dbecd56d..0000000000
> --- a/sysdeps/x86_64/fpu/s_ceill.S
> +++ /dev/null
> @@ -1,34 +0,0 @@
> -/*
> - * Public domain.
> - */
> -
> -#include <libm-alias-ldouble.h>
> -#include <machine/asm.h>
> -
> -
> -ENTRY(__ceill)
> -       fldt    8(%rsp)
> -
> -       fnstenv -28(%rsp)               /* store fpu environment */
> -
> -       /* We use here %edx although only the low 1 bits are defined.
> -          But none of the operations should care and they are faster
> -          than the 16 bit operations.  */
> -       movl    $0x0800,%edx            /* round towards +oo */
> -       orl     -28(%rsp),%edx
> -       andl    $0xfbff,%edx
> -       movl    %edx,-32(%rsp)
> -       fldcw   -32(%rsp)               /* load modified control word */
> -
> -       frndint                         /* round */
> -
> -       /* Preserve "invalid" exceptions from sNaN input.  */
> -       fnstsw
> -       andl    $0x1, %eax
> -       orl     %eax, -24(%rsp)
> -
> -       fldenv  -28(%rsp)               /* restore original environment */
> -
> -       ret
> -END (__ceill)
> -libm_alias_ldouble (__ceil, ceil)
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 08/12] x86: Do not raise inexact exception on floorl
  2023-12-28 17:20 ` [PATCH 08/12] x86: Do not raise inexact exception on floorl Adhemerval Zanella
@ 2024-04-01 14:13   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:13 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:23 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
>
> Checked on x86_64-linux-gnu and i686-linux-gnu.
> ---
>  sysdeps/i386/fpu/s_floorl.S   | 39 -----------------------------------
>  sysdeps/x86/fpu/s_floorl.c    | 38 ++++++++++++++++++++++++++++++++++
>  sysdeps/x86_64/fpu/s_floorl.S | 33 -----------------------------
>  3 files changed, 38 insertions(+), 72 deletions(-)
>  delete mode 100644 sysdeps/i386/fpu/s_floorl.S
>  create mode 100644 sysdeps/x86/fpu/s_floorl.c
>  delete mode 100644 sysdeps/x86_64/fpu/s_floorl.S
>
> diff --git a/sysdeps/i386/fpu/s_floorl.S b/sysdeps/i386/fpu/s_floorl.S
> deleted file mode 100644
> index 3ec28b477b..0000000000
> --- a/sysdeps/i386/fpu/s_floorl.S
> +++ /dev/null
> @@ -1,39 +0,0 @@
> -/*
> - * Public domain.
> - */
> -
> -#include <libm-alias-ldouble.h>
> -#include <machine/asm.h>
> -
> -RCSID("$NetBSD: $")
> -
> -ENTRY(__floorl)
> -       fldt    4(%esp)
> -       subl    $32,%esp
> -       cfi_adjust_cfa_offset (32)
> -
> -       fnstenv 4(%esp)                 /* store fpu environment */
> -
> -       /* We use here %edx although only the low 1 bits are defined.
> -          But none of the operations should care and they are faster
> -          than the 16 bit operations.  */
> -       movl    $0x400,%edx             /* round towards -oo */
> -       orl     4(%esp),%edx
> -       andl    $0xf7ff,%edx
> -       movl    %edx,(%esp)
> -       fldcw   (%esp)                  /* load modified control word */
> -
> -       frndint                         /* round */
> -
> -       /* Preserve "invalid" exceptions from sNaN input.  */
> -       fnstsw
> -       andl    $0x1, %eax
> -       orl     %eax, 8(%esp)
> -
> -       fldenv  4(%esp)                 /* restore original environment */
> -
> -       addl    $32,%esp
> -       cfi_adjust_cfa_offset (-32)
> -       ret
> -END (__floorl)
> -libm_alias_ldouble (__floor, floor)
> diff --git a/sysdeps/x86/fpu/s_floorl.c b/sysdeps/x86/fpu/s_floorl.c
> new file mode 100644
> index 0000000000..edaf5e954f
> --- /dev/null
> +++ b/sysdeps/x86/fpu/s_floorl.c
> @@ -0,0 +1,38 @@
> +/* Return largest integral value not less than argument.  x86 version.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024.

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#define NO_MATH_REDIRECT
> +#include <math.h>
> +#include <fenv_private.h>
> +#include <libm-alias-ldouble.h>
> +
> +long double
> +__floorl (long double x)
> +{
> +  fenv_t fenv;
> +  long double r;
> +
> +  libc_feholdexcept_setround_387 (&fenv, FE_DOWNWARD);
> +  asm volatile ("frndint" : "=t" (r) : "0" (x));
> +  /* Preserve "invalid" exceptions from sNaN input.  */
> +  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
> +  libc_fesetenv_387 (&fenv);
> +
> +  return r;
> +}
> +libm_alias_ldouble (__floor, floor)
> diff --git a/sysdeps/x86_64/fpu/s_floorl.S b/sysdeps/x86_64/fpu/s_floorl.S
> deleted file mode 100644
> index b74d1a4d6b..0000000000
> --- a/sysdeps/x86_64/fpu/s_floorl.S
> +++ /dev/null
> @@ -1,33 +0,0 @@
> -/*
> - * Public domain.
> - */
> -
> -#include <libm-alias-ldouble.h>
> -#include <machine/asm.h>
> -
> -ENTRY(__floorl)
> -       fldt    8(%rsp)
> -
> -       fnstenv -28(%rsp)               /* store fpu environment */
> -
> -       /* We use here %edx although only the low 1 bits are defined.
> -          But none of the operations should care and they are faster
> -          than the 16 bit operations.  */
> -       movl    $0x400,%edx             /* round towards -oo */
> -       orl     -28(%rsp),%edx
> -       andl    $0xf7ff,%edx
> -       movl    %edx,-32(%rsp)
> -       fldcw   -32(%rsp)               /* load modified control word */
> -
> -       frndint                         /* round */
> -
> -       /* Preserve "invalid" exceptions from sNaN input.  */
> -       fnstsw
> -       andl    $0x1, %eax
> -       orl     %eax, -24(%rsp)
> -
> -       fldenv  -28(%rsp)               /* restore original environment */
> -
> -       ret
> -END (__floorl)
> -libm_alias_ldouble (__floor, floor)
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 09/12] x86: Do not raise inexact exception on truncl
  2023-12-28 17:20 ` [PATCH 09/12] x86: Do not raise inexact exception on truncl Adhemerval Zanella
@ 2024-04-01 14:14   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:14 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:23 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
>
> Checked on x86_64-linux-gnu and i686-linux-gnu.
> ---
>  .../fpu/s_truncl.S => x86/fpu/s_truncl.c}     | 40 +++++++++----------
>  sysdeps/x86_64/fpu/s_truncl.S                 | 36 -----------------
>  2 files changed, 19 insertions(+), 57 deletions(-)
>  rename sysdeps/{i386/fpu/s_truncl.S => x86/fpu/s_truncl.c} (61%)
>  delete mode 100644 sysdeps/x86_64/fpu/s_truncl.S
>
> diff --git a/sysdeps/i386/fpu/s_truncl.S b/sysdeps/x86/fpu/s_truncl.c
> similarity index 61%
> rename from sysdeps/i386/fpu/s_truncl.S
> rename to sysdeps/x86/fpu/s_truncl.c
> index cfd4e75252..249b307004 100644
> --- a/sysdeps/i386/fpu/s_truncl.S
> +++ b/sysdeps/x86/fpu/s_truncl.c
> @@ -1,5 +1,5 @@
> -/* Truncate long double value.
> -   Copyright (C) 1997-2023 Free Software Foundation, Inc.
> +/* Round to integer, toward zero.  x86 version.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024.

>     This file is part of the GNU C Library.
>
>     The GNU C Library is free software; you can redistribute it and/or
> @@ -16,25 +16,23 @@
>     License along with the GNU C Library; if not, see
>     <https://www.gnu.org/licenses/>.  */
>
> +#define NO_MATH_REDIRECT
> +#include <math.h>
> +#include <fenv_private.h>
>  #include <libm-alias-ldouble.h>
> -#include <machine/asm.h>
>
> -ENTRY(__truncl)
> -       fldt    4(%esp)
> -       subl    $32, %esp
> -       cfi_adjust_cfa_offset (32)
> -       fnstenv 4(%esp)
> -       movl    $0xc00, %edx
> -       orl     4(%esp), %edx
> -       movl    %edx, (%esp)
> -       fldcw   (%esp)
> -       frndint
> -       fnstsw
> -       andl    $0x1, %eax
> -       orl     %eax, 8(%esp)
> -       fldenv  4(%esp)
> -       addl    $32, %esp
> -       cfi_adjust_cfa_offset (-32)
> -       ret
> -END(__truncl)
> +long double
> +__truncl (long double x)
> +{
> +  fenv_t fenv;
> +  long double r;
> +
> +  libc_feholdexcept_setround_387 (&fenv, FE_TOWARDZERO);
> +  asm volatile ("frndint" : "=t" (r) : "0" (x));
> +  /* Preserve "invalid" exceptions from sNaN input.  */
> +  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
> +  libc_fesetenv_387 (&fenv);
> +
> +  return r;
> +}
>  libm_alias_ldouble (__trunc, trunc)
> diff --git a/sysdeps/x86_64/fpu/s_truncl.S b/sysdeps/x86_64/fpu/s_truncl.S
> deleted file mode 100644
> index 8d319e68b0..0000000000
> --- a/sysdeps/x86_64/fpu/s_truncl.S
> +++ /dev/null
> @@ -1,36 +0,0 @@
> -/* Truncate long double value.
> -   Copyright (C) 1997-2023 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
> -   <https://www.gnu.org/licenses/>.  */
> -
> -#include <libm-alias-ldouble.h>
> -#include <machine/asm.h>
> -
> -ENTRY(__truncl)
> -       fldt    8(%rsp)
> -       fnstenv -28(%rsp)
> -       movl    $0xc00, %edx
> -       orl     -28(%rsp), %edx
> -       movl    %edx, -32(%rsp)
> -       fldcw   -32(%rsp)
> -       frndint
> -       fnstsw
> -       andl    $0x1, %eax
> -       orl     %eax, -24(%rsp)
> -       fldenv  -28(%rsp)
> -       ret
> -END(__truncl)
> -libm_alias_ldouble (__trunc, trunc)
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 10/12] x86: Do not raise inexact exception on floor/floorf
  2023-12-28 17:20 ` [PATCH 10/12] x86: Do not raise inexact exception on floor/floorf Adhemerval Zanella
@ 2024-04-01 14:15   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:15 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:24 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
>
> Checked on x86_64-linux-gnu and i686-linux-gnu.
> ---
>  sysdeps/i386/fpu/s_floor.S  | 34 ---------------------------------
>  sysdeps/i386/fpu/s_floor.c  | 38 +++++++++++++++++++++++++++++++++++++
>  sysdeps/i386/fpu/s_floorf.S | 34 ---------------------------------
>  sysdeps/i386/fpu/s_floorf.c | 38 +++++++++++++++++++++++++++++++++++++
>  4 files changed, 76 insertions(+), 68 deletions(-)
>  delete mode 100644 sysdeps/i386/fpu/s_floor.S
>  create mode 100644 sysdeps/i386/fpu/s_floor.c
>  delete mode 100644 sysdeps/i386/fpu/s_floorf.S
>  create mode 100644 sysdeps/i386/fpu/s_floorf.c
>
> diff --git a/sysdeps/i386/fpu/s_floor.S b/sysdeps/i386/fpu/s_floor.S
> deleted file mode 100644
> index 7143fdcc9a..0000000000
> --- a/sysdeps/i386/fpu/s_floor.S
> +++ /dev/null
> @@ -1,34 +0,0 @@
> -/*
> - * Public domain.
> - */
> -
> -#include <machine/asm.h>
> -#include <libm-alias-double.h>
> -
> -RCSID("$NetBSD: s_floor.S,v 1.4 1995/05/09 00:01:59 jtc Exp $")
> -
> -ENTRY(__floor)
> -       fldl    4(%esp)
> -       subl    $32,%esp
> -       cfi_adjust_cfa_offset (32)
> -
> -       fnstenv 4(%esp)                 /* store fpu environment */
> -
> -       /* We use here %edx although only the low 1 bits are defined.
> -          But none of the operations should care and they are faster
> -          than the 16 bit operations.  */
> -       movl    $0x400,%edx             /* round towards -oo */
> -       orl     4(%esp),%edx
> -       andl    $0xf7ff,%edx
> -       movl    %edx,(%esp)
> -       fldcw   (%esp)                  /* load modified control word */
> -
> -       frndint                         /* round */
> -
> -       fldenv  4(%esp)                 /* restore original environment */
> -
> -       addl    $32,%esp
> -       cfi_adjust_cfa_offset (-32)
> -       ret
> -END (__floor)
> -libm_alias_double (__floor, floor)
> diff --git a/sysdeps/i386/fpu/s_floor.c b/sysdeps/i386/fpu/s_floor.c
> new file mode 100644
> index 0000000000..90ba61da92
> --- /dev/null
> +++ b/sysdeps/i386/fpu/s_floor.c
> @@ -0,0 +1,38 @@
> +/* Return largest integral value not less than argument.  i386 version.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024.

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#define NO_MATH_REDIRECT
> +#include <math.h>
> +#include <fenv_private.h>
> +#include <libm-alias-double.h>
> +
> +double
> +__floor (double x)
> +{
> +  fenv_t fenv;
> +  double r;
> +
> +  libc_feholdexcept_setround_387 (&fenv, FE_DOWNWARD);
> +  asm volatile ("frndint" : "=t" (r) : "0" (x));
> +  /* Preserve "invalid" exceptions from sNaN input.  */
> +  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
> +  libc_fesetenv_387 (&fenv);
> +
> +  return r;
> +}
> +libm_alias_double (__floor, floor)
> diff --git a/sysdeps/i386/fpu/s_floorf.S b/sysdeps/i386/fpu/s_floorf.S
> deleted file mode 100644
> index 8fad9c0698..0000000000
> --- a/sysdeps/i386/fpu/s_floorf.S
> +++ /dev/null
> @@ -1,34 +0,0 @@
> -/*
> - * Public domain.
> - */
> -
> -#include <machine/asm.h>
> -#include <libm-alias-float.h>
> -
> -RCSID("$NetBSD: s_floorf.S,v 1.3 1995/05/09 00:04:32 jtc Exp $")
> -
> -ENTRY(__floorf)
> -       flds    4(%esp)
> -       subl    $32,%esp
> -       cfi_adjust_cfa_offset (32)
> -
> -       fnstenv 4(%esp)                 /* store fpu environment */
> -
> -       /* We use here %edx although only the low 1 bits are defined.
> -          But none of the operations should care and they are faster
> -          than the 16 bit operations.  */
> -       movl    $0x400,%edx             /* round towards -oo */
> -       orl     4(%esp),%edx
> -       andl    $0xf7ff,%edx
> -       movl    %edx,(%esp)
> -       fldcw   (%esp)                  /* load modified control word */
> -
> -       frndint                         /* round */
> -
> -       fldenv  4(%esp)                 /* restore original environment */
> -
> -       addl    $32,%esp
> -       cfi_adjust_cfa_offset (-32)
> -       ret
> -END (__floorf)
> -libm_alias_float (__floor, floor)
> diff --git a/sysdeps/i386/fpu/s_floorf.c b/sysdeps/i386/fpu/s_floorf.c
> new file mode 100644
> index 0000000000..0e7e8223a8
> --- /dev/null
> +++ b/sysdeps/i386/fpu/s_floorf.c
> @@ -0,0 +1,38 @@
> +/* Return largest integral value not less than argument.  i386 version.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024.

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#define NO_MATH_REDIRECT
> +#include <math.h>
> +#include <fenv_private.h>
> +#include <libm-alias-float.h>
> +
> +float
> +__floorf (float x)
> +{
> +  fenv_t fenv;
> +  float r;
> +
> +  libc_feholdexcept_setround_387 (&fenv, FE_DOWNWARD);
> +  asm volatile ("frndint" : "=t" (r) : "0" (x));
> +  /* Preserve "invalid" exceptions from sNaN input.  */
> +  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
> +  libc_fesetenv_387 (&fenv);
> +
> +  return r;
> +}
> +libm_alias_float (__floor, floor)
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 11/12] i386: Do not raise inexact exception on ceil/ceilf
  2023-12-28 17:20 ` [PATCH 11/12] i386: Do not raise inexact exception on ceil/ceilf Adhemerval Zanella
@ 2024-04-01 14:16   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:16 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:25 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
>
> Checked on x86_64-linux-gnu and i686-linux-gnu.
> ---
>  sysdeps/i386/fpu/s_ceil.S  | 34 ----------------------------------
>  sysdeps/i386/fpu/s_ceil.c  | 38 ++++++++++++++++++++++++++++++++++++++
>  sysdeps/i386/fpu/s_ceilf.S | 34 ----------------------------------
>  sysdeps/i386/fpu/s_ceilf.c | 38 ++++++++++++++++++++++++++++++++++++++
>  4 files changed, 76 insertions(+), 68 deletions(-)
>  delete mode 100644 sysdeps/i386/fpu/s_ceil.S
>  create mode 100644 sysdeps/i386/fpu/s_ceil.c
>  delete mode 100644 sysdeps/i386/fpu/s_ceilf.S
>  create mode 100644 sysdeps/i386/fpu/s_ceilf.c
>
> diff --git a/sysdeps/i386/fpu/s_ceil.S b/sysdeps/i386/fpu/s_ceil.S
> deleted file mode 100644
> index 99984f9b8d..0000000000
> --- a/sysdeps/i386/fpu/s_ceil.S
> +++ /dev/null
> @@ -1,34 +0,0 @@
> -/*
> - * Public domain.
> - */
> -
> -#include <machine/asm.h>
> -#include <libm-alias-double.h>
> -
> -RCSID("$NetBSD: s_ceil.S,v 1.4 1995/05/08 23:52:13 jtc Exp $")
> -
> -ENTRY(__ceil)
> -       fldl    4(%esp)
> -       subl    $32,%esp
> -       cfi_adjust_cfa_offset (32)
> -
> -       fnstenv 4(%esp)                 /* store fpu environment */
> -
> -       /* We use here %edx although only the low 1 bits are defined.
> -          But none of the operations should care and they are faster
> -          than the 16 bit operations.  */
> -       movl    $0x0800,%edx            /* round towards +oo */
> -       orl     4(%esp),%edx
> -       andl    $0xfbff,%edx
> -       movl    %edx,(%esp)
> -       fldcw   (%esp)                  /* load modified control word */
> -
> -       frndint                         /* round */
> -
> -       fldenv  4(%esp)                 /* restore original environment */
> -
> -       addl    $32,%esp
> -       cfi_adjust_cfa_offset (-32)
> -       ret
> -END (__ceil)
> -libm_alias_double (__ceil, ceil)
> diff --git a/sysdeps/i386/fpu/s_ceil.c b/sysdeps/i386/fpu/s_ceil.c
> new file mode 100644
> index 0000000000..ec948ef83f
> --- /dev/null
> +++ b/sysdeps/i386/fpu/s_ceil.c
> @@ -0,0 +1,38 @@
> +/* Return smallest integral value not less than argument.  i386 version.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024.

> +   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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#define NO_MATH_REDIRECT
> +#include <math.h>
> +#include <fenv_private.h>
> +#include <libm-alias-double.h>
> +
> +double
> +__ceil (double x)
> +{
> +  fenv_t fenv;
> +  double r;
> +
> +  libc_feholdexcept_setround_387 (&fenv, FE_UPWARD);
> +  asm volatile ("frndint" : "=t" (r) : "0" (x));
> +  /* Preserve "invalid" exceptions from sNaN input.  */
> +  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
> +  libc_fesetenv_387 (&fenv);
> +
> +  return r;
> +}
> +libm_alias_double (__ceil, ceil)
> diff --git a/sysdeps/i386/fpu/s_ceilf.S b/sysdeps/i386/fpu/s_ceilf.S
> deleted file mode 100644
> index 03e8e22609..0000000000
> --- a/sysdeps/i386/fpu/s_ceilf.S
> +++ /dev/null
> @@ -1,34 +0,0 @@
> -/*
> - * Public domain.
> - */
> -
> -#include <machine/asm.h>
> -#include <libm-alias-float.h>
> -
> -RCSID("$NetBSD: s_ceilf.S,v 1.3 1995/05/08 23:52:44 jtc Exp $")
> -
> -ENTRY(__ceilf)
> -       flds    4(%esp)
> -       subl    $32,%esp
> -       cfi_adjust_cfa_offset (32)
> -
> -       fnstenv 4(%esp)                 /* store fpu environment */
> -
> -       /* We use here %edx although only the low 1 bits are defined.
> -          But none of the operations should care and they are faster
> -          than the 16 bit operations.  */
> -       movl    $0x0800,%edx            /* round towards +oo */
> -       orl     4(%esp),%edx
> -       andl    $0xfbff,%edx
> -       movl    %edx,(%esp)
> -       fldcw   (%esp)                  /* load modified control word */
> -
> -       frndint                         /* round */
> -
> -       fldenv  4(%esp)                 /* restore original environment */
> -
> -       addl    $32,%esp
> -       cfi_adjust_cfa_offset (-32)
> -       ret
> -END (__ceilf)
> -libm_alias_float (__ceil, ceil)
> diff --git a/sysdeps/i386/fpu/s_ceilf.c b/sysdeps/i386/fpu/s_ceilf.c
> new file mode 100644
> index 0000000000..36159610b2
> --- /dev/null
> +++ b/sysdeps/i386/fpu/s_ceilf.c
> @@ -0,0 +1,38 @@
> +/* Return largest integral value not less than argument.  i386 version.
> +   Copyright (C) 2023 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
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#define NO_MATH_REDIRECT
> +#include <math.h>
> +#include <fenv_private.h>
> +#include <libm-alias-float.h>
> +
> +float
> +__ceilf (float x)
> +{
> +  fenv_t fenv;
> +  float r;
> +
> +  libc_feholdexcept_setround_387 (&fenv, FE_UPWARD);
> +  asm volatile ("frndint" : "=t" (r) : "0" (x));
> +  /* Preserve "invalid" exceptions from sNaN input.  */
> +  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
> +  libc_fesetenv_387 (&fenv);
> +
> +  return r;
> +}
> +libm_alias_float (__ceil, ceil)
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 12/12] i386: Do not raise inexact exception on trunc/truncf
  2023-12-28 17:20 ` [PATCH 12/12] i386: Do not raise inexact exception on trunc/truncf Adhemerval Zanella
@ 2024-04-01 14:17   ` H.J. Lu
  0 siblings, 0 replies; 27+ messages in thread
From: H.J. Lu @ 2024-04-01 14:17 UTC (permalink / raw)
  To: Adhemerval Zanella; +Cc: libc-alpha

On Thu, Dec 28, 2023 at 9:23 AM Adhemerval Zanella
<adhemerval.zanella@linaro.org> wrote:
>
> It is not allowed anymore on ISO C23.
>
> Checked on x86_64-linux-gnu and i686-linux-gnu.
> ---
>  sysdeps/i386/fpu/{s_trunc.S => s_trunc.c}   | 37 +++++++++++----------
>  sysdeps/i386/fpu/{s_truncf.S => s_truncf.c} | 37 +++++++++++----------
>  2 files changed, 38 insertions(+), 36 deletions(-)
>  rename sysdeps/i386/fpu/{s_trunc.S => s_trunc.c} (61%)
>  rename sysdeps/i386/fpu/{s_truncf.S => s_truncf.c} (61%)
>
> diff --git a/sysdeps/i386/fpu/s_trunc.S b/sysdeps/i386/fpu/s_trunc.c
> similarity index 61%
> rename from sysdeps/i386/fpu/s_trunc.S
> rename to sysdeps/i386/fpu/s_trunc.c
> index 145288bc8e..b071ad9750 100644
> --- a/sysdeps/i386/fpu/s_trunc.S
> +++ b/sysdeps/i386/fpu/s_trunc.c
> @@ -1,5 +1,5 @@
> -/* Truncate double value.
> -   Copyright (C) 1997-2023 Free Software Foundation, Inc.
> +/* Round to integer, toward zero.  i386 version.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024.

>     This file is part of the GNU C Library.
>
>     The GNU C Library is free software; you can redistribute it and/or
> @@ -16,22 +16,23 @@
>     License along with the GNU C Library; if not, see
>     <https://www.gnu.org/licenses/>.  */
>
> -#include <machine/asm.h>
> +#define NO_MATH_REDIRECT
> +#include <math.h>
> +#include <fenv_private.h>
>  #include <libm-alias-double.h>
>
> -ENTRY(__trunc)
> -       fldl    4(%esp)
> -       subl    $32, %esp
> -       cfi_adjust_cfa_offset (32)
> -       fnstenv 4(%esp)
> -       movl    $0xc00, %edx
> -       orl     4(%esp), %edx
> -       movl    %edx, (%esp)
> -       fldcw   (%esp)
> -       frndint
> -       fldenv  4(%esp)
> -       addl    $32, %esp
> -       cfi_adjust_cfa_offset (-32)
> -       ret
> -END(__trunc)
> +double
> +__trunc (double x)
> +{
> +  fenv_t fenv;
> +  double r;
> +
> +  libc_feholdexcept_setround_387 (&fenv, FE_TOWARDZERO);
> +  asm volatile ("frndint" : "=t" (r) : "0" (x));
> +  /* Preserve "invalid" exceptions from sNaN input.  */
> +  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
> +  libc_fesetenv_387 (&fenv);
> +
> +  return r;
> +}
>  libm_alias_double (__trunc, trunc)
> diff --git a/sysdeps/i386/fpu/s_truncf.S b/sysdeps/i386/fpu/s_truncf.c
> similarity index 61%
> rename from sysdeps/i386/fpu/s_truncf.S
> rename to sysdeps/i386/fpu/s_truncf.c
> index 7bdf0e5111..86d46ff660 100644
> --- a/sysdeps/i386/fpu/s_truncf.S
> +++ b/sysdeps/i386/fpu/s_truncf.c
> @@ -1,5 +1,5 @@
> -/* Truncate float value.
> -   Copyright (C) 1997-2023 Free Software Foundation, Inc.
> +/* Round to integer, toward zero.  i386 version.
> +   Copyright (C) 2023 Free Software Foundation, Inc.

2024.

>     This file is part of the GNU C Library.
>
>     The GNU C Library is free software; you can redistribute it and/or
> @@ -16,22 +16,23 @@
>     License along with the GNU C Library; if not, see
>     <https://www.gnu.org/licenses/>.  */
>
> -#include <machine/asm.h>
> +#define NO_MATH_REDIRECT
> +#include <math.h>
> +#include <fenv_private.h>
>  #include <libm-alias-float.h>
>
> -ENTRY(__truncf)
> -       flds    4(%esp)
> -       subl    $32, %esp
> -       cfi_adjust_cfa_offset (32)
> -       fnstenv 4(%esp)
> -       movl    $0xc00, %edx
> -       orl     4(%esp), %edx
> -       movl    %edx, (%esp)
> -       fldcw   (%esp)
> -       frndint
> -       fldenv  4(%esp)
> -       addl    $32, %esp
> -       cfi_adjust_cfa_offset (-32)
> -       ret
> -END(__truncf)
> +float
> +__truncf (float x)
> +{
> +  fenv_t fenv;
> +  float r;
> +
> +  libc_feholdexcept_setround_387 (&fenv, FE_TOWARDZERO);
> +  asm volatile ("frndint" : "=t" (r) : "0" (x));
> +  /* Preserve "invalid" exceptions from sNaN input.  */
> +  fenv.__status_word |= libc_fetestexcept_387 (FE_INVALID);
> +  libc_fesetenv_387 (&fenv);
> +
> +  return r;
> +}
>  libm_alias_float (__trunc, trunc)
> --
> 2.34.1
>


-- 
H.J.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH 02/12] powerpc: Add missing arch flags on rounding ifunc variants
  2024-04-01 13:59   ` H.J. Lu
@ 2024-04-01 16:50     ` Peter Bergner
  0 siblings, 0 replies; 27+ messages in thread
From: Peter Bergner @ 2024-04-01 16:50 UTC (permalink / raw)
  To: H.J. Lu, Adhemerval Zanella; +Cc: libc-alpha

On 4/1/24 8:59 AM, H.J. Lu wrote:
> On Thu, Dec 28, 2023 at 9:21 AM Adhemerval Zanella
> <adhemerval.zanella@linaro.org> wrote:
>>
>> The ifunc variants now uses the powerpc implementation which in turn
>> uses the compiler builtin.  Without the proper -mcpu switch the builtin
>> does not generate the expected optimization.
>>
>> Checked on powerpc-linux-gnu.
>> ---
>>  sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile | 6 ++++++
>>  1 file changed, 6 insertions(+)
>>
>> diff --git a/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile b/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile
>> index 64317506c6..2f5c0eded6 100644
>> --- a/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile
>> +++ b/sysdeps/powerpc/powerpc32/power4/fpu/multiarch/Makefile
>> @@ -34,6 +34,12 @@ CFLAGS-s_modff-power5+.c = -mcpu=power5+
>>  CFLAGS-s_logbl-power7.c = -mcpu=power7
>>  CFLAGS-s_logb-power7.c = -mcpu=power7
>>  CFLAGS-s_logbf-power7.c = -mcpu=power7
>> +CFLAGS-s_round-power5+.c += -mcpu=power5+
>> +CFLAGS-s_roundf-power5+.c += -mcpu=power5+
>> +CFLAGS-s_floor-power5+.c += -mcpu=power5+
>> +CFLAGS-s_floorf-power5+.c += -mcpu=power5+
>> +CFLAGS-s_trunc-power5+.c += -mcpu=power5+
>> +CFLAGS-s_truncf-power5+.c += -mcpu=power5+
>>
>>  # These files quiet sNaNs in a way that is optimized away without
>>  # -fsignaling-nans.
>> --
>> 2.34.1
>>
> 
> LGTM.
> 
> Reviewed-by: H.J. Lu <hjl.tools@gmail.com>

This predated me taking over the Power port, but...

LGTM too.

Reviewed-by: Peter Bergner <bergner@linux.ibm.com>


Peter




^ permalink raw reply	[flat|nested] 27+ messages in thread

end of thread, other threads:[~2024-04-01 16:50 UTC | newest]

Thread overview: 27+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-12-28 17:20 [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella
2023-12-28 17:20 ` [PATCH 01/12] math: Reformat Makefile Adhemerval Zanella
2024-04-01 13:57   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 02/12] powerpc: Add missing arch flags on rounding ifunc variants Adhemerval Zanella
2024-04-01 13:59   ` H.J. Lu
2024-04-01 16:50     ` Peter Bergner
2023-12-28 17:20 ` [PATCH 03/12] math: Add test to check if ceil raise inexact floating-point exception Adhemerval Zanella
2024-04-01 14:01   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 04/12] math: Add test to check if floor " Adhemerval Zanella
2024-04-01 14:02   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 05/12] math: Add test to check if trunc " Adhemerval Zanella
2024-04-01 14:03   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 06/12] math: Add test to check if round " Adhemerval Zanella
2024-04-01 14:02   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 07/12] x86: Do not raise inexact exception on ceill Adhemerval Zanella
2024-04-01 14:12   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 08/12] x86: Do not raise inexact exception on floorl Adhemerval Zanella
2024-04-01 14:13   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 09/12] x86: Do not raise inexact exception on truncl Adhemerval Zanella
2024-04-01 14:14   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 10/12] x86: Do not raise inexact exception on floor/floorf Adhemerval Zanella
2024-04-01 14:15   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 11/12] i386: Do not raise inexact exception on ceil/ceilf Adhemerval Zanella
2024-04-01 14:16   ` H.J. Lu
2023-12-28 17:20 ` [PATCH 12/12] i386: Do not raise inexact exception on trunc/truncf Adhemerval Zanella
2024-04-01 14:17   ` H.J. Lu
2024-02-16 17:47 ` [PATCH 00/12] Improve rounding to interger function for C23 Adhemerval Zanella Netto

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).