Refactor part of math Makefile
diff mbox

Message ID 8d257095-c2da-0b8c-1740-00c8552d4ec8@linux.vnet.ibm.com
State New
Headers show

Commit Message

Paul E. Murphy June 14, 2016, 8:26 p.m. UTC
This is related to https://sourceware.org/ml/libc-alpha/2016-06/msg00100.html
based on feedback.
---8<---
In order to support more types, the Makefile needs a few bits
shuffled.

F is explictly used as a placeholder to substitute for the
appropriate type suffix.  This removes the need to demangle
_r suffixed objects.

The variable is compat-calls is added to house any objects which
are only built to provide compat symbols within libm.  That is,
no newly added type should ever attempt building these.

New types would be added as noted in the comments preceding
type-TYPE-{suffix,routines,yes} variables.  However, some manual
additions will still need to be done to add appropriate flags
when building the various variants of libm-test.c for a new type.

	* math/Makefile (libm-calls): Insert F placeholder into
	object names.  Move w_lgamma_compat* routines into
	(libm-compat-calls): New variable.
	(calls): Insert F placeholder.
	(generated): Redefine using new type-foreach macro call.
	(routines): Likewise.

	(types): New variable to hold a list of symbolic names
	for each type supported by the target machine.
	(libm-routines): Redefine using new types variable.
	(type-foreach): macro to replace fooF with the
	appropriately suffixed foo for each types.

	(type-ldouble-suffix): New variable.
	(type-ldouble-routines): Likewise.
	(type-ldouble-yes): Likewise.
	(type-double-suffix): Likewise.
	(type-double-routines): Likewise.
	(type-float-suffix): Likewise.
	(type-float-routines): Likewise.

	(dbl-only-routines): Remove.
	(long-c-yes): Likewise.
	(long-m-routines): Likewise.
	(long-m-support): Likewise.
	(long-m-yes): Likewise.
	(test-longdouble-yes): Likewise.

	(test-float.o): Simplify by increasing the scope of the
	recipe which adds libm-test.stmp to the vector types.
	(test-ifloat.o): Likewise.
	(test-float-finite.o): Likewise.
	(test-double.o): Likewise
	(test-idouble.o): Likewise.
	(test-double-finite.o): Likewise.
	(test-ldouble.o): Likewise
	(test-ildoubl.o): Likewise.
	(test-ldouble-finite.o): Likewise.

	(libm-tests): Redefine type tests using foreach and the
	list of types in types.
---
 math/Makefile | 127 +++++++++++++++++++++++++++++++++-------------------------
 1 file changed, 72 insertions(+), 55 deletions(-)

Comments

Joseph Myers June 14, 2016, 9:04 p.m. UTC | #1
On Tue, 14 Jun 2016, Paul E. Murphy wrote:

> +	     gamma_productF k_standardF lgamma_negF lgamma_productF		\
> +
> +libm-compat-calls = w_lgamma_compatf w_lgamma_compat w_lgamma_compatl

Given the conclusion that we should obsolete matherr / _LIB_VERSION as 
part of the preparation for adding a new type, k_standard should end up in 
the set of compat calls as well.  (The new cleaner wrappers used for new 
types, and for static linking / new ports for existing types, would handle 
errno setting directly without calling __kernel_standard.)

> +libm-tests = $(foreach t,$(types),test-$(t) 				\
> +				  test-$(t)-finite 			\
> +				  test-i$(subst ldouble,ldoubl,$(t)))	\

I don't think we should keep the irregularity of the test being called 
test-ildoubl rather than ildouble.  Rename it and you no longer need a 
special $(subst ldouble,ldoubl,$(t)).
Paul E. Murphy June 14, 2016, 9:26 p.m. UTC | #2
On 06/14/2016 04:04 PM, Joseph Myers wrote:
> On Tue, 14 Jun 2016, Paul E. Murphy wrote:
> 
>> +	     gamma_productF k_standardF lgamma_negF lgamma_productF		\
>> +
>> +libm-compat-calls = w_lgamma_compatf w_lgamma_compat w_lgamma_compatl
> 
> Given the conclusion that we should obsolete matherr / _LIB_VERSION as 
> part of the preparation for adding a new type, k_standard should end up in 
> the set of compat calls as well.  (The new cleaner wrappers used for new 
> types, and for static linking / new ports for existing types, would handle 
> errno setting directly without calling __kernel_standard.)

Is there a reason not to continue calling the function __kernel_standard
and build a compat variant named __kernel_standard_compat to preserve the
legacy behavior (with similar additions to this variable).

Anyhow, that change should be done on its own.  Its a trivial change to
this patch.

> 
>> +libm-tests = $(foreach t,$(types),test-$(t) 				\
>> +				  test-$(t)-finite 			\
>> +				  test-i$(subst ldouble,ldoubl,$(t)))	\
> 
> I don't think we should keep the irregularity of the test being called 
> test-ildoubl rather than ildouble.  Rename it and you no longer need a 
> special $(subst ldouble,ldoubl,$(t)).

I'm happy to oblige there.  Do you happen to know the reason for the odd
naming?
Joseph Myers June 14, 2016, 9:37 p.m. UTC | #3
On Tue, 14 Jun 2016, Paul E. Murphy wrote:

> > Given the conclusion that we should obsolete matherr / _LIB_VERSION as 
> > part of the preparation for adding a new type, k_standard should end up in 
> > the set of compat calls as well.  (The new cleaner wrappers used for new 
> > types, and for static linking / new ports for existing types, would handle 
> > errno setting directly without calling __kernel_standard.)
> 
> Is there a reason not to continue calling the function __kernel_standard
> and build a compat variant named __kernel_standard_compat to preserve the
> legacy behavior (with similar additions to this variable).

__kernel_standard is a very strange interface involving an enumerated list 
of functions and recomputing error returns.  It's not something to 
replicate for new types.  The natural approach is:

* Call the function being wrapped (unconditionally), which will set 
exceptions and produce an appropriate return.

* Check for NaN / infinity / zero (as appropriate) as results, and set 
errno directly instead of calling __kernel_standard, and return the 
original error value instead of recomputing it.

No __kernel_standard.  No checks on the writable global (not thread-local) 
variable _LIB_VERSION.

Note: it's likely that various existing function implementations do not in 
fact get their error returns right, and this is invisible in testing 
because those original error returns are overridden by __kernel_standard 
unless you use -lieee.  So when adding cleaner wrappers to be used for new 
types and ports and static linking, it's important to test them to find 
and fix such cases (which are user-visible bugs, though fairly obscure 
ones).

> >> +libm-tests = $(foreach t,$(types),test-$(t) 				\
> >> +				  test-$(t)-finite 			\
> >> +				  test-i$(subst ldouble,ldoubl,$(t)))	\
> > 
> > I don't think we should keep the irregularity of the test being called 
> > test-ildoubl rather than ildouble.  Rename it and you no longer need a 
> > special $(subst ldouble,ldoubl,$(t)).
> 
> I'm happy to oblige there.  Do you happen to know the reason for the odd
> naming?

My guess would be long-ago 14-character limits on filenames 
(test-ildouble.c is 15 characters).  We no longer need to care about that 
and have filenames up to 34 characters in glibc (e.g. 
pthread_mutexattr_getprioceiling.c), and git uses filenames longer than 
that.

Patch
diff mbox

diff --git a/math/Makefile b/math/Makefile
index 6cd3cf1..9d18ade 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -42,50 +42,77 @@  libm-support = s_lib_version s_matherr s_signgam			\
 	       ftestexcept fegetround fesetround fegetenv feholdexcpt	\
 	       fesetenv feupdateenv t_exp fedisblxcpt feenablxcpt	\
 	       fegetexcept
-libm-calls = e_acos e_acosh e_asin e_atan2 e_atanh e_cosh e_exp e_fmod	\
-	     e_hypot e_j0 e_j1 e_jn e_lgamma_r e_log e_log10 e_pow	\
-	     e_rem_pio2 e_remainder e_scalb e_sinh e_sqrt e_gamma_r	\
-	     e_ilogb							\
-	     k_cos k_rem_pio2 k_sin k_tan s_asinh s_atan s_cbrt		\
-	     s_ceil s_cos s_erf s_expm1 s_fabs				\
-	     s_floor s_log1p w_log1p s_logb				\
-	     s_nextafter s_nexttoward s_rint s_scalbln w_scalbln	\
-	     s_significand s_sin s_tan s_tanh w_acos w_acosh w_asin	\
-	     w_atan2 w_atanh w_cosh w_exp w_exp2 w_exp10 w_fmod		\
-	     w_tgamma w_hypot w_j0 w_j1 w_jn w_lgamma w_lgamma_r	\
-	     w_log w_log10 w_pow w_remainder w_scalb w_sinh w_sqrt	\
-	     w_ilogb							\
-	     s_fpclassify s_fmax s_fmin s_fdim s_nan s_trunc		\
-	     s_remquo e_log2 e_exp2 s_round s_nearbyint s_sincos	\
-	     conj cimag creal cabs carg s_cexp s_csinh s_ccosh s_clog	\
-	     s_catan s_casin s_ccos s_csin s_ctan s_ctanh s_cacos	\
-	     s_casinh s_cacosh s_catanh s_csqrt s_cpow s_cproj s_clog10 \
-	     s_fma s_lrint s_llrint s_lround s_llround e_exp10 w_log2	\
-	     s_issignaling $(calls:s_%=m_%) x2y2m1 k_casinh	\
-	     gamma_product k_standard lgamma_neg lgamma_product		\
-	     w_lgamma_compat
-
-dbl-only-routines := branred doasin dosincos halfulp mpa mpatan2	\
-		     mpatan mpexp mplog mpsqrt mptan sincos32 slowexp	\
-		     slowpow sincostab
-libm-routines = $(strip $(libm-support) $(libm-calls) \
-			$(patsubst %_rf,%f_r,$(libm-calls:=f))	\
-			$(long-m-$(long-double-fcts))) \
-		$(dbl-only-routines)
-long-m-routines = $(patsubst %_rl,%l_r,$(libm-calls:=l))
-long-m-support = t_sincosl k_sincosl
-long-m-yes = $(long-m-routines) $(long-m-support)
+
+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_rem_pio2F e_remainderF e_scalbF e_sinhF e_sqrtF e_gammaF_r	\
+	     e_ilogbF								\
+	     k_cosF k_rem_pio2F k_sinF k_tanF s_asinhF s_atanF s_cbrtF		\
+	     s_ceilF s_cosF s_erfF s_expm1F s_fabsF				\
+	     s_floorF s_log1pF w_log1pF s_logbF					\
+	     s_nextafterF s_nexttowardF s_rintF s_scalblnF w_scalblnF		\
+	     s_significandF s_sinF s_tanF s_tanhF w_acosF w_acoshF w_asinF	\
+	     w_atan2F w_atanhF w_coshF w_expF w_exp2F w_exp10F w_fmodF		\
+	     w_tgammaF w_hypotF w_j0F w_j1F w_jnF w_lgammaF w_lgammaF_r		\
+	     w_logF w_log10F w_powF w_remainderF w_scalbF w_sinhF w_sqrtF	\
+	     w_ilogbF								\
+	     s_fpclassifyF s_fmaxF s_fminF s_fdimF s_nanF s_truncF		\
+	     s_remquoF e_log2F e_exp2F s_roundF s_nearbyintF s_sincosF		\
+	     conjF cimagF crealF cabsF cargF s_cexpF s_csinhF s_ccoshF s_clogF	\
+	     s_catanF s_casinF s_ccosF s_csinF s_ctanF s_ctanhF s_cacosF	\
+	     s_casinhF s_cacoshF s_catanhF s_csqrtF s_cpowF s_cprojF s_clog10F 	\
+	     s_fmaF s_lrintF s_llrintF s_lroundF s_llroundF e_exp10F w_log2F	\
+	     s_issignalingF $(calls:s_%=m_%) x2y2m1F k_casinhF			\
+	     gamma_productF k_standardF lgamma_negF lgamma_productF		\
+
+libm-compat-calls = w_lgamma_compatf w_lgamma_compat w_lgamma_compatl
+
+
+# Type specific routine support.
+#
+# The following three variables control what is included for each type:
+#
+# type-floatN-suffix = The suffix of the type
+# type-floatN-routines = Type specific support objects
+# type-floatN-yes = If the type is supported, evaluates to floatN
+#
+# 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 = $(type-ldouble-$(long-double-fcts)) double float
+
+# long double support
+type-ldouble-suffix := l
+type-ldouble-routines := t_sincosl k_sincosl
+type-ldouble-yes := ldouble
+
+# double support
+type-double-suffix :=
+type-double-routines := branred doasin dosincos halfulp mpa mpatan2	\
+		       mpatan mpexp mplog mpsqrt mptan sincos32 slowexp	\
+		       slowpow sincostab
+
+# float support
+type-float-suffix := f
+type-float-routines :=
+
+
+# Apply suffix to each type in arg 1
+type-foreach = $(foreach t,$(types),$(subst F,$(type-$(t)-suffix),$(1)))
+
+libm-routines = $(strip $(libm-support) $(libm-compat-calls)		\
+			$(call type-foreach, $(libm-calls))		\
+			$(foreach t, $(types), $(type-$(t)-routines))) 	\
 
 # These functions are in libc instead of libm because __printf_fp
 # calls them, so any program using printf will need them linked in,
 # 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_isinf s_isnan s_finite s_copysign s_modf s_scalbn s_frexp s_ldexp \
-	s_signbit
-generated += $(foreach s,.c .S l.c l.S f.c f.S,$(calls:s_%=m_%$s))
-routines = $(calls) $(calls:=f) $(long-c-$(long-double-fcts))
-long-c-yes = $(calls:=l)
+calls = s_isinfF s_isnanF s_finiteF s_copysignF s_modfF s_scalbnF s_frexpF \
+	s_ldexpF s_signbitF
+generated += $(foreach s,.c .S,$(call type-foreach, $(calls:s_%=m_%)))
+routines = $(call type-foreach, $(calls))
 
 ifeq ($(build-mathvec),yes)
 # We need to install libm.so as linker script
@@ -120,9 +147,6 @@  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
-# We do the `long double' tests only if this data type is available and
-# distinct from `double'.
-test-longdouble-yes = test-ldouble test-ildoubl test-ldouble-finite
 
 ifneq (,$(CXX))
 tests += test-math-isinff
@@ -130,9 +154,12 @@  endif
 
 ifneq (no,$(PERL))
 libm-vec-tests = $(addprefix test-,$(libmvec-tests))
-libm-tests = test-float test-double $(test-longdouble-$(long-double-fcts)) \
-	test-ifloat test-idouble test-float-finite test-double-finite \
-	$(libm-vec-tests)
+
+libm-tests = $(foreach t,$(types),test-$(t) 				\
+				  test-$(t)-finite 			\
+				  test-i$(subst ldouble,ldoubl,$(t)))	\
+	     $(libm-vec-tests)
+
 libm-tests.o = $(addsuffix .o,$(libm-tests))
 
 tests += $(libm-tests)
@@ -155,16 +182,6 @@  $(objpfx)libm-test.stmp: $(ulps-file) libm-test.inc gen-libm-test.pl \
 	$(PERL) gen-libm-test.pl -u $< -o "$(objpfx)"
 	$(SHELL) gen-libm-have-vector-test.sh > $(objpfx)libm-have-vector-test.h
 	@echo > $@
-
-$(objpfx)test-float.o: $(objpfx)libm-test.stmp
-$(objpfx)test-ifloat.o: $(objpfx)libm-test.stmp
-$(objpfx)test-float-finite.o: $(objpfx)libm-test.stmp
-$(objpfx)test-double.o: $(objpfx)libm-test.stmp
-$(objpfx)test-idouble.o: $(objpfx)libm-test.stmp
-$(objpfx)test-double-finite.o: $(objpfx)libm-test.stmp
-$(objpfx)test-ldouble.o: $(objpfx)libm-test.stmp
-$(objpfx)test-ildoubl.o: $(objpfx)libm-test.stmp
-$(objpfx)test-ldouble-finite.o: $(objpfx)libm-test.stmp
 endif
 
 libm-test-fast-math-cflags = -fno-builtin -D__FAST_MATH__ -DTEST_FAST_MATH
@@ -247,7 +264,7 @@  include ../Rules
 
 ifneq (no,$(PERL))
 # This must come after the inclusion of sysdeps Makefiles via Rules.
-$(addprefix $(objpfx), $(addsuffix .o, $(libm-vec-tests))): $(objpfx)libm-test.stmp
+$(addprefix $(objpfx), $(libm-tests.o)): $(objpfx)libm-test.stmp
 
 # Run the math programs to automatically generate ULPs files.
 .PHONY: regen-ulps