Commit 49548b09e0fa9fddb64e3cd47266193e36b73144 [MIPS: math-emu: Cleanup
definition of structs describe sp/dp floats.] introduced a regression for
pre-4.7 GCC versions that apparently have issues with static initialisers
setting anonymous struct/union members, even though such members have been
supported by GCC in other uses for a little bit longer.
This has been figured out by Al Viro and documented in commit
05064084e82d057f8d74590c51581650e060fbb8 [fix __swap_writepage() compile
failure on old gcc versions] (thanks Al!).
The regression manifests itself by a load of compilation errors like:
arch/mips/math-emu/ieee754.c:45: error: unknown field 'sign' specified in initializer
arch/mips/math-emu/ieee754.c:45: warning: missing braces around initializer
arch/mips/math-emu/ieee754.c:45: warning: (near initialization for '__ieee754dp_spcvals[0].<anonymous>')
arch/mips/math-emu/ieee754.c:45: error: unknown field 'bexp' specified in initializer
arch/mips/math-emu/ieee754.c:45: warning: excess elements in union initializer
arch/mips/math-emu/ieee754.c:45: warning: (near initialization for '__ieee754dp_spcvals[0]')
arch/mips/math-emu/ieee754.c:45: error: unknown field 'mant' specified in initializer
arch/mips/math-emu/ieee754.c:45: warning: excess elements in union initializer
arch/mips/math-emu/ieee754.c:45: warning: (near initialization for '__ieee754dp_spcvals[0]')
-- and so on, and so on.
This change gives a name back to the anonymous union members and reverts
parts of the problematic commit while retaining what I believe were the
good parts of the cleanup. I concluded a workaround such as one made by
Al in his commit is infeasible in this context, there's simply too much
static initialisation across math-emu.
It is worth noting here that we still in principle support GCC versions
back to 3.2:
$ grep 'Gnu C' Documentation/Changes
o Gnu C 3.2 # gcc --version
$
and I think it makes sense to keep this support where feasible as I think
from the perspective of the Linux kernel it is better if developers' time
is spent on the kernel itself rather than on sorting out GCC versions.
These anonymous struct/union members are a semi-standard language
extension BTW, citing from a recent version of the GCC manual:
--------------------------------------------------------------------------
6 Extensions to the C Language Family
*************************************
GNU C provides several language features not found in ISO standard C.
(The `-pedantic' option directs GCC to print a warning message if any
of these features is used.) To test for the availability of these
features in conditional compilation, check for a predefined macro
`__GNUC__', which is always defined under GCC.
6.59 Unnamed struct/union fields within structs/unions
======================================================
As permitted by ISO C11 and for compatibility with other compilers, GCC
allows you to define a structure or union that contains, as fields,
structures and unions without names.
--------------------------------------------------------------------------
-- note the term "permitted" rather than "required".
We do make use of a few GCC language extensions, most notably inline
assembly, however in this case we merely save a couple of characters here
and there and this is IMO not worth breaking people's development
environments.
Signed-off-by: Maciej W. Rozycki <macro@xxxxxxxxxxxxxx>
---
linux-mips-anon-ieee754.patch
Index: linux-20140623-4maxp64/arch/mips/math-emu/dp_sqrt.c
===================================================================
--- linux-20140623-4maxp64.orig/arch/mips/math-emu/dp_sqrt.c
+++ linux-20140623-4maxp64/arch/mips/math-emu/dp_sqrt.c
@@ -110,13 +110,13 @@ union ieee754dp ieee754dp_sqrt(union iee
/* triple to almost 56 sig. bits: y ~= sqrt(x) to within 1 ulp */
/* t=y*y; z=t; pt[n0]+=0x00100000; t+=z; z=(x-z)*y; */
z = t = ieee754dp_mul(y, y);
- t.bexp += 0x001;
+ t.parts.bexp += 0x001;
t = ieee754dp_add(t, z);
z = ieee754dp_mul(ieee754dp_sub(x, z), y);
/* t=z/(t+x) ; pt[n0]+=0x00100000; y+=t; */
t = ieee754dp_div(z, ieee754dp_add(t, x));
- t.bexp += 0x001;
+ t.parts.bexp += 0x001;
y = ieee754dp_add(y, t);
/* twiddle last bit to force y correctly rounded */
@@ -155,7 +155,7 @@ union ieee754dp ieee754dp_sqrt(union iee
}
/* py[n0]=py[n0]+scalx; ...scale back y */
- y.bexp += scalx;
+ y.parts.bexp += scalx;
/* restore rounding mode, possibly set inexact */
ieee754_csr = oldcsr;
Index: linux-20140623-4maxp64/arch/mips/math-emu/ieee754.c
===================================================================
--- linux-20140623-4maxp64.orig/arch/mips/math-emu/ieee754.c
+++ linux-20140623-4maxp64/arch/mips/math-emu/ieee754.c
@@ -36,9 +36,11 @@
#define DPCNST(s, b, m) \
{ \
- .sign = (s), \
- .bexp = (b) + DP_EBIAS, \
- .mant = (m) \
+ .parts = { \
+ .sign = (s), \
+ .bexp = (b) + DP_EBIAS, \
+ .mant = (m) \
+ } \
}
const union ieee754dp __ieee754dp_spcvals[] = {
@@ -63,9 +65,11 @@ const union ieee754dp __ieee754dp_spcval
#define SPCNST(s, b, m) \
{ \
- .sign = (s), \
- .bexp = (b) + SP_EBIAS, \
- .mant = (m) \
+ .parts = { \
+ .sign = (s), \
+ .bexp = (b) + SP_EBIAS, \
+ .mant = (m) \
+ } \
}
const union ieee754sp __ieee754sp_spcvals[] = {
Index: linux-20140623-4maxp64/arch/mips/math-emu/ieee754.h
===================================================================
--- linux-20140623-4maxp64.orig/arch/mips/math-emu/ieee754.h
+++ linux-20140623-4maxp64/arch/mips/math-emu/ieee754.h
@@ -37,7 +37,7 @@ union ieee754dp {
__BITFIELD_FIELD(unsigned int bexp:11,
__BITFIELD_FIELD(u64 mant:52,
;)))
- };
+ } parts;
u64 bits;
};
@@ -47,7 +47,7 @@ union ieee754sp {
__BITFIELD_FIELD(unsigned bexp:8,
__BITFIELD_FIELD(unsigned mant:23,
;)))
- };
+ } parts;
u32 bits;
};
Index: linux-20140623-4maxp64/arch/mips/math-emu/ieee754dp.h
===================================================================
--- linux-20140623-4maxp64.orig/arch/mips/math-emu/ieee754dp.h
+++ linux-20140623-4maxp64/arch/mips/math-emu/ieee754dp.h
@@ -36,9 +36,9 @@
#define DP_HIDDEN_BIT DP_MBIT(DP_FBITS)
#define DP_SIGN_BIT DP_MBIT(63)
-#define DPSIGN(dp) (dp.sign)
-#define DPBEXP(dp) (dp.bexp)
-#define DPMANT(dp) (dp.mant)
+#define DPSIGN(dp) (dp.parts.sign)
+#define DPBEXP(dp) (dp.parts.bexp)
+#define DPMANT(dp) (dp.parts.mant)
static inline int ieee754dp_finite(union ieee754dp x)
{
@@ -70,9 +70,9 @@ static inline union ieee754dp builddp(in
&& (bx) <= DP_EMAX + 1 + DP_EBIAS);
assert(((m) >> DP_FBITS) == 0);
- r.sign = s;
- r.bexp = bx;
- r.mant = m;
+ r.parts.sign = s;
+ r.parts.bexp = bx;
+ r.parts.mant = m;
return r;
}
Index: linux-20140623-4maxp64/arch/mips/math-emu/ieee754sp.h
===================================================================
--- linux-20140623-4maxp64.orig/arch/mips/math-emu/ieee754sp.h
+++ linux-20140623-4maxp64/arch/mips/math-emu/ieee754sp.h
@@ -36,9 +36,9 @@
#define SP_HIDDEN_BIT SP_MBIT(SP_FBITS)
#define SP_SIGN_BIT SP_MBIT(31)
-#define SPSIGN(sp) (sp.sign)
-#define SPBEXP(sp) (sp.bexp)
-#define SPMANT(sp) (sp.mant)
+#define SPSIGN(sp) (sp.parts.sign)
+#define SPBEXP(sp) (sp.parts.bexp)
+#define SPMANT(sp) (sp.parts.mant)
static inline int ieee754sp_finite(union ieee754sp x)
{
@@ -75,9 +75,9 @@ static inline union ieee754sp buildsp(in
&& (bx) <= SP_EMAX + 1 + SP_EBIAS);
assert(((m) >> SP_FBITS) == 0);
- r.sign = s;
- r.bexp = bx;
- r.mant = m;
+ r.parts.sign = s;
+ r.parts.bexp = bx;
+ r.parts.mant = m;
return r;
}
