On Mon, Nov 27, 2017 at 03:40:58PM -0700, Martin Sebor wrote:
> On 11/27/2017 12:21 PM, Michael Meissner wrote:
> >The new -Wattribute-alias option now issues warnings for old-style ifunc
> >declarations that coerce the pointer to the function to void *. The
> >float128-ifunc.c module in libgcc/config/rs6000 now gets a lot of warnings of
> >the form:
> >
> >../float128-ifunc.c:109:1: warning: ‘ifunc’ resolver for ‘__negkf2’ should
> >return ‘TFtype (*)(TFtype) {aka _Float128 (*)(_Float128)}’
> >[-Wattribute-alias]
> >
> >This patch fixes these warnings. I have done a full bootstrap build and test
> >suite run. I have verified that the ifunc handler works correctly, using
> >software emulation on a power8 and the hardware instructions on power9. Can
> >I
> >check this into the trunk?
>
> Just as a side note, a convenient way to deal with this is to use
> typeof to deduce the return type of the resolver from the type of
> the function it returns. I would expect something like
> the following untested change to do it and make the typedefs
> unnecessary:
>
> -static void *
> +static __typeof__ (__addkf3_sw) *
> __addkf3_resolve (void)
> {
> - return (void *) SW_OR_HW (__addkf3_sw, __addkf3_hw);
> + return SW_OR_HW (__addkf3_sw, __addkf3_hw);
> }
>
> Martin
Thanks, that does simplify things. I redid the patch for both PR libgcc/83112
(avoid the warnings in float128-ifunc) and PR libgcc/83103 (optimize complex
float128 multiply/divide when running on ISA 3.0 hardware).
This patch combines both PRs. I have bootstrapped it and done a regression
test with no regressions. Can I install this patch in the trunk?
2017-11-27 Michael Meissner <meiss...@linux.vnet.ibm.com>
PR libgcc/83112
* config/rs6000/float128-ifunc.c (__addkf3_resolve): Use the
correct type for all ifunc resolvers to silence -Wattribute-alias
warnings. Eliminate the forward declaration of the resolver
functions which is no longer needed.
(__subkf3_resolve): Likewise.
(__mulkf3_resolve): Likewise.
(__divkf3_resolve): Likewise.
(__negkf2_resolve): Likewise.
(__eqkf2_resolve): Likewise.
(__nekf2_resolve): Likewise.
(__gekf2_resolve): Likewise.
(__gtkf2_resolve): Likewise.
(__lekf2_resolve): Likewise.
(__ltkf2_resolve): Likewise.
(__unordkf2_resolve): Likewise.
(__extendsfkf2_resolve): Likewise.
(__extenddfkf2_resolve): Likewise.
(__trunckfsf2_resolve): Likewise.
(__trunckfdf2_resolve): Likewise.
(__fixkfsi_resolve): Likewise.
(__fixkfdi_resolve): Likewise.
(__fixunskfsi_resolve): Likewise.
(__fixunskfdi_resolve): Likewise.
(__floatsikf_resolve): Likewise.
(__floatdikf_resolve): Likewise.
(__floatunsikf_resolve): Likewise.
(__floatundikf_resolve): Likewise.
(__extendkftf2_resolve): Likewise.
(__trunctfkf2_resolve): Likewise.
PR libgcc/83103
* config/rs6000/quad-float128.h (TF): Don't define if long double
is IEEE 128-bit floating point.
(TCtype): Define as either TCmode or KCmode, depending on whether
long double is IEEE 128-bit floating point.
(__mulkc3_sw): Add declarations for software/hardware versions of
complex multiply/divide.
(__divkc3_sw): Likewise.
(__mulkc3_hw): Likewise.
(__divkc3_hw): Likewise.
* config/rs6000/_mulkc3.c (_mulkc3): If we are building ifunc
handlers to switch between using software emulation and hardware
float128 instructions, build the complex multiply/divide functions
for both software and hardware support.
* config/rs6000/_divkc3.c (_divkc3): Likewise.
* config/rs6000/float128-ifunc.c (__mulkc3_resolve): Likewise.
(__divkc3_resolve): Likewise.
(__mulkc3): Likewise.
(__divkc3): Likewise.
* config/rs6000/t-float128-hw (fp128_hardfp_src): Likewise.
(fp128_hw_src): Likewise.
(fp128_hw_static_obj): Likewise.
(fp128_hw_shared_obj): Likewise.
(_mulkc3-hw.c): Likewise.
(_divkc3-hw.c): Likewise.
* config/rs6000/t-float128 (clean-float128): Delete _mulkc3-hw.c
and _divkc3-hw.c.
--
Michael Meissner, IBM
IBM, M/S 2506R, 550 King Street, Littleton, MA 01460-6245, USA
email: meiss...@linux.vnet.ibm.com, phone: +1 (978) 899-4797
Index: libgcc/config/rs6000/quad-float128.h
===================================================================
--- libgcc/config/rs6000/quad-float128.h (revision 255177)
+++ libgcc/config/rs6000/quad-float128.h (working copy)
@@ -30,13 +30,20 @@
/* quad.h defines the TFtype type by:
typedef float TFtype __attribute__ ((mode (TF)));
- This define forces it to use KFmode (aka, ieee 128-bit floating point). */
+ This define forces it to use KFmode (aka, ieee 128-bit floating point).
+ However, when the compiler's default is changed so that long double is IEEE
+ 128-bit floating point, we need to go back to using TFmode and TCmode. */
+#ifndef __LONG_DOUBLE_IEEE128__
#define TF KF
/* We also need TCtype to represent complex ieee 128-bit float for
__mulkc3 and __divkc3. */
typedef __complex float TCtype __attribute__ ((mode (KC)));
+#else
+typedef __complex float TCtype __attribute__ ((mode (TC)));
+#endif
+
/* Force the use of the VSX instruction set. */
#if defined(_ARCH_PPC) && (!defined(__VSX__) || !defined(__FLOAT128__))
#pragma GCC target ("vsx,float128")
@@ -88,6 +95,8 @@ extern TFtype __floatunsikf_sw (USItype_
extern TFtype __floatundikf_sw (UDItype_ppc);
extern IBM128_TYPE __extendkftf2_sw (TFtype);
extern TFtype __trunctfkf2_sw (IBM128_TYPE);
+extern TCtype __mulkc3_sw (TFtype, TFtype, TFtype, TFtype);
+extern TCtype __divkc3_sw (TFtype, TFtype, TFtype, TFtype);
#ifdef _ARCH_PPC64
/* We do not provide ifunc resolvers for __fixkfti, __fixunskfti, __floattikf,
@@ -128,6 +137,8 @@ extern TFtype __floatunsikf_hw (USItype_
extern TFtype __floatundikf_hw (UDItype_ppc);
extern IBM128_TYPE __extendkftf2_hw (TFtype);
extern TFtype __trunctfkf2_hw (IBM128_TYPE);
+extern TCtype __mulkc3_hw (TFtype, TFtype, TFtype, TFtype);
+extern TCtype __divkc3_hw (TFtype, TFtype, TFtype, TFtype);
/* Ifunc function declarations, to automatically switch between software
emulation and hardware support. */
Index: libgcc/config/rs6000/_mulkc3.c
===================================================================
--- libgcc/config/rs6000/_mulkc3.c (revision 255177)
+++ libgcc/config/rs6000/_mulkc3.c (working copy)
@@ -31,6 +31,10 @@ typedef __complex float KCtype __attribu
#define isnan __builtin_isnan
#define isinf __builtin_isinf
+#if defined(FLOAT128_HW_INSNS) && !defined(__mulkc3)
+#define __mulkc3 __mulkc3_sw
+#endif
+
KCtype
__mulkc3 (KFtype a, KFtype b, KFtype c, KFtype d)
{
Index: libgcc/config/rs6000/_divkc3.c
===================================================================
--- libgcc/config/rs6000/_divkc3.c (revision 255177)
+++ libgcc/config/rs6000/_divkc3.c (working copy)
@@ -33,6 +33,10 @@ typedef __complex float KCtype __attribu
#define isinf __builtin_isinf
#define isfinite __builtin_isfinite
+#if defined(FLOAT128_HW_INSNS) && !defined(__divkc3)
+#define __divkc3 __divkc3_sw
+#endif
+
KCtype
__divkc3 (KFtype a, KFtype b, KFtype c, KFtype d)
{
Index: libgcc/config/rs6000/float128-ifunc.c
===================================================================
--- libgcc/config/rs6000/float128-ifunc.c (revision 255177)
+++ libgcc/config/rs6000/float128-ifunc.c (working copy)
@@ -54,190 +54,175 @@
128-bit integer types and 128-bit IEEE floating point, or vice versa. So
use the emulator functions for these conversions. */
-static void *__addkf3_resolve (void);
-static void *__subkf3_resolve (void);
-static void *__mulkf3_resolve (void);
-static void *__divkf3_resolve (void);
-static void *__negkf2_resolve (void);
-static void *__eqkf2_resolve (void);
-static void *__nekf2_resolve (void);
-static void *__gekf2_resolve (void);
-static void *__gtkf2_resolve (void);
-static void *__lekf2_resolve (void);
-static void *__ltkf2_resolve (void);
-static void *__unordkf2_resolve (void);
-static void *__extendsfkf2_resolve (void);
-static void *__extenddfkf2_resolve (void);
-static void *__trunckfsf2_resolve (void);
-static void *__trunckfdf2_resolve (void);
-static void *__fixkfsi_resolve (void);
-static void *__fixkfdi_resolve (void);
-static void *__fixunskfsi_resolve (void);
-static void *__fixunskfdi_resolve (void);
-static void *__floatsikf_resolve (void);
-static void *__floatdikf_resolve (void);
-static void *__floatunsikf_resolve (void);
-static void *__floatundikf_resolve (void);
-static void *__extendkftf2_resolve (void);
-static void *__trunctfkf2_resolve (void);
-
-static void *
+static __typeof__ (__addkf3_sw) *
__addkf3_resolve (void)
{
- return (void *) SW_OR_HW (__addkf3_sw, __addkf3_hw);
+ return SW_OR_HW (__addkf3_sw, __addkf3_hw);
}
-static void *
+static __typeof__ (__subkf3_sw) *
__subkf3_resolve (void)
{
- return (void *) SW_OR_HW (__subkf3_sw, __subkf3_hw);
+ return SW_OR_HW (__subkf3_sw, __subkf3_hw);
}
-static void *
+static __typeof__ (__mulkf3_sw) *
__mulkf3_resolve (void)
{
- return (void *) SW_OR_HW (__mulkf3_sw, __mulkf3_hw);
+ return SW_OR_HW (__mulkf3_sw, __mulkf3_hw);
}
-static void *
+static __typeof__ (__divkf3_sw) *
__divkf3_resolve (void)
{
- return (void *) SW_OR_HW (__divkf3_sw, __divkf3_hw);
+ return SW_OR_HW (__divkf3_sw, __divkf3_hw);
}
-static void *
+static __typeof__ (__negkf2_sw) *
__negkf2_resolve (void)
{
- return (void *) SW_OR_HW (__negkf2_sw, __negkf2_hw);
+ return SW_OR_HW (__negkf2_sw, __negkf2_hw);
}
-static void *
+static __typeof__ (__floatsikf_sw) *
__floatsikf_resolve (void)
{
- return (void *) SW_OR_HW (__floatsikf_sw, __floatsikf_hw);
+ return SW_OR_HW (__floatsikf_sw, __floatsikf_hw);
}
-static void *
+static __typeof__ (__floatdikf_sw) *
__floatdikf_resolve (void)
{
- return (void *) SW_OR_HW (__floatdikf_sw, __floatdikf_hw);
+ return SW_OR_HW (__floatdikf_sw, __floatdikf_hw);
}
-static void *
+static __typeof__ (__floatunsikf_sw) *
__floatunsikf_resolve (void)
{
- return (void *) SW_OR_HW (__floatunsikf_sw, __floatunsikf_hw);
+ return SW_OR_HW (__floatunsikf_sw, __floatunsikf_hw);
}
-static void *
+static __typeof__ (__floatundikf_sw) *
__floatundikf_resolve (void)
{
- return (void *) SW_OR_HW (__floatundikf_sw, __floatundikf_hw);
+ return SW_OR_HW (__floatundikf_sw, __floatundikf_hw);
}
-static void *
+static __typeof__ (__fixkfsi_sw) *
__fixkfsi_resolve (void)
{
- return (void *) SW_OR_HW (__fixkfsi_sw, __fixkfsi_hw);
+ return SW_OR_HW (__fixkfsi_sw, __fixkfsi_hw);
}
-static void *
+static __typeof__ (__fixkfdi_sw) *
__fixkfdi_resolve (void)
{
- return (void *) SW_OR_HW (__fixkfdi_sw, __fixkfdi_hw);
+ return SW_OR_HW (__fixkfdi_sw, __fixkfdi_hw);
}
-static void *
+static __typeof__ (__fixunskfsi_sw) *
__fixunskfsi_resolve (void)
{
- return (void *) SW_OR_HW (__fixunskfsi_sw, __fixunskfsi_hw);
+ return SW_OR_HW (__fixunskfsi_sw, __fixunskfsi_hw);
}
-static void *
+static __typeof__ (__fixunskfdi_sw) *
__fixunskfdi_resolve (void)
{
- return (void *) SW_OR_HW (__fixunskfdi_sw, __fixunskfdi_hw);
+ return SW_OR_HW (__fixunskfdi_sw, __fixunskfdi_hw);
}
-static void *
+static __typeof__ (__extendsfkf2_sw) *
__extendsfkf2_resolve (void)
{
- return (void *) SW_OR_HW (__extendsfkf2_sw, __extendsfkf2_hw);
+ return SW_OR_HW (__extendsfkf2_sw, __extendsfkf2_hw);
}
-static void *
+static __typeof__ (__extenddfkf2_sw) *
__extenddfkf2_resolve (void)
{
- return (void *) SW_OR_HW (__extenddfkf2_sw, __extenddfkf2_hw);
+ return SW_OR_HW (__extenddfkf2_sw, __extenddfkf2_hw);
}
-static void *
+static __typeof__ (__trunckfsf2_sw) *
__trunckfsf2_resolve (void)
{
- return (void *) SW_OR_HW (__trunckfsf2_sw, __trunckfsf2_hw);
+ return SW_OR_HW (__trunckfsf2_sw, __trunckfsf2_hw);
}
-static void *
+static __typeof__ (__trunckfdf2_sw) *
__trunckfdf2_resolve (void)
{
return (void *) SW_OR_HW (__trunckfdf2_sw, __trunckfdf2_hw);
}
-static void *
+static __typeof__ (__extendkftf2_sw) *
__extendkftf2_resolve (void)
{
- return (void *) SW_OR_HW (__extendkftf2_sw, __extendkftf2_hw);
+ return SW_OR_HW (__extendkftf2_sw, __extendkftf2_hw);
}
-static void *
+static __typeof__ (__trunctfkf2_sw) *
__trunctfkf2_resolve (void)
{
- return (void *) SW_OR_HW (__trunctfkf2_sw, __trunctfkf2_hw);
+ return SW_OR_HW (__trunctfkf2_sw, __trunctfkf2_hw);
}
-static void *
+static __typeof__ (__mulkc3_sw) *
+__mulkc3_resolve (void)
+{
+ return SW_OR_HW (__mulkc3_sw, __mulkc3_hw);
+}
+
+static __typeof__ (__divkc3_sw) *
+__divkc3_resolve (void)
+{
+ return SW_OR_HW (__divkc3_sw, __divkc3_hw);
+}
+
+static __typeof__ (__eqkf2_sw) *
__eqkf2_resolve (void)
{
- return (void *) SW_OR_HW (__eqkf2_sw, __eqkf2_hw);
+ return SW_OR_HW (__eqkf2_sw, __eqkf2_hw);
}
-static void *
+static __typeof__ (__gekf2_sw) *
__gekf2_resolve (void)
{
- return (void *) SW_OR_HW (__gekf2_sw, __gekf2_hw);
+ return SW_OR_HW (__gekf2_sw, __gekf2_hw);
}
-static void *
+static __typeof__ (__lekf2_sw) *
__lekf2_resolve (void)
{
- return (void *) SW_OR_HW (__lekf2_sw, __lekf2_hw);
+ return SW_OR_HW (__lekf2_sw, __lekf2_hw);
}
-static void *
+static __typeof__ (__unordkf2_sw) *
__unordkf2_resolve (void)
{
- return (void *) SW_OR_HW (__unordkf2_sw, __unordkf2_hw);
+ return SW_OR_HW (__unordkf2_sw, __unordkf2_hw);
}
/* Resolve __nekf2, __gtkf2, __ltkf2 like __eqkf2, __gekf2, and __lekf2, since
the functions return the same values. */
-static void *
+static __typeof__ (__eqkf2_sw) *
__nekf2_resolve (void)
{
- return (void *) SW_OR_HW (__eqkf2_sw, __eqkf2_hw);
+ return SW_OR_HW (__eqkf2_sw, __eqkf2_hw);
}
-static void *
+static __typeof__ (__eqkf2_sw) *
__gtkf2_resolve (void)
{
- return (void *) SW_OR_HW (__gekf2_sw, __gekf2_hw);
+ return SW_OR_HW (__gekf2_sw, __gekf2_hw);
}
-static void *
+static __typeof__ (__eqkf2_sw) *
__ltkf2_resolve (void)
{
- return (void *) SW_OR_HW (__lekf2_sw, __lekf2_hw);
+ return SW_OR_HW (__lekf2_sw, __lekf2_hw);
}
@@ -320,3 +305,9 @@ IBM128_TYPE __extendkftf2 (TFtype)
TFtype __trunctfkf2 (IBM128_TYPE)
__attribute__ ((__ifunc__ ("__trunctfkf2_resolve")));
+
+TCtype __mulkc3 (TFtype, TFtype, TFtype, TFtype)
+ __attribute__ ((__ifunc__ ("__mulkc3_resolve")));
+
+TCtype __divkc3 (TFtype, TFtype, TFtype, TFtype)
+ __attribute__ ((__ifunc__ ("__divkc3_resolve")));
Index: libgcc/config/rs6000/t-float128-hw
===================================================================
--- libgcc/config/rs6000/t-float128-hw (revision 255177)
+++ libgcc/config/rs6000/t-float128-hw (working copy)
@@ -5,10 +5,12 @@
FLOAT128_HW_INSNS = -DFLOAT128_HW_INSNS
# New functions for hardware support
-fp128_hw_funcs = float128-hw
-fp128_hw_src = $(srcdir)/config/rs6000/float128-hw.c
-fp128_hw_static_obj = float128-hw$(objext)
-fp128_hw_shared_obj = float128-hw_s$(objext)
+fp128_hardfp_src = _mulkc3-hw.c _divkc3-hw.c
+fp128_hw_funcs = float128-hw _mulkc3-hw _divkc3-hw
+fp128_hw_src = $(srcdir)/config/rs6000/float128-hw.c _mulkc3-hw.c \
+ _divkc3-hw.c
+fp128_hw_static_obj = $(addsuffix $(objext),$(fp128_hw_funcs))
+fp128_hw_shared_obj = $(addsuffix _s$(objext),$(fp128_hw_funcs))
fp128_hw_obj = $(fp128_hw_static_obj) $(fp128_hw_shared_obj)
fp128_ifunc_funcs = float128-ifunc
@@ -33,3 +35,13 @@ $(fp128_hw_obj) : $(srcdir)/config/rs6
$(fp128_ifunc_obj) : INTERNAL_CFLAGS += $(FP128_CFLAGS_SW)
$(fp128_ifunc_obj) : $(srcdir)/config/rs6000/t-float128-hw
+
+_mulkc3-hw.c: $(srcdir)/config/rs6000/_mulkc3.c
+ rm -rf _mulkc3.c
+ (echo "#define __mulkc3 __mulkc3_hw"; \
+ cat $(srcdir)/config/rs6000/_mulkc3.c) > _mulkc3-hw.c
+
+_divkc3-hw.c: $(srcdir)/config/rs6000/_divkc3.c
+ rm -rf _divkc3.c
+ (echo "#define __divkc3 __divkc3_hw"; \
+ cat $(srcdir)/config/rs6000/_divkc3.c) > _divkc3-hw.c
Index: libgcc/config/rs6000/t-float128
===================================================================
--- libgcc/config/rs6000/t-float128 (revision 255177)
+++ libgcc/config/rs6000/t-float128 (working copy)
@@ -86,7 +86,7 @@ test:
for x in $(fp128_obj); do echo " $$x"; done;
clean-float128:
- rm -rf $(fp128_softfp_src)
+ rm -rf $(fp128_softfp_src) $(fp128_hardfp_src)
@$(MULTICLEAN) multi-clean DO=clean-float128
# For now, only put it in the static library
