Date: Sun, 8 May 2011 10:10:43 -0400 From: Attilio Rao <attilio@freebsd.org> To: Bruce Evans <brde@optusnet.com.au> Cc: svn-src-projects@freebsd.org, src-committers@freebsd.org Subject: Re: svn commit: r221614 - projects/largeSMP/sys/powerpc/include Message-ID: <BANLkTi=mnZc6fy8TGpLMsMLDV8UpK5J-Gw@mail.gmail.com> In-Reply-To: <20110508202725.K1192@besplex.bde.org> References: <201105080039.p480doiZ021493@svn.freebsd.org> <20110508202725.K1192@besplex.bde.org>
next in thread | previous in thread | raw e-mail | index | archive | help
2011/5/8 Bruce Evans <brde@optusnet.com.au>:
> On Sun, 8 May 2011, Attilio Rao wrote:
>
>> Log:
>> =C2=A0All architectures define the size-bounded types (uint32_t, uint64_=
t,
>> etc.)
>> =C2=A0starting from base C types (int, long, etc).
>> =C2=A0That is also reflected when building atomic operations, as the
>> =C2=A0size-bounded types are built from the base C types.
>>
>> =C2=A0However, powerpc does the inverse thing, leading to a serie of nas=
ty
>
> I think you mean that it does the inverse thing for atomic ops only.
>
>> =C2=A0bugs.
>> =C2=A0Cleanup the atomic implementation by defining as base the base C t=
ype
>> =C2=A0version and depending on them, appropriately.
>
> I think that it is mostly the other arches that are backwards here,
> except for plain ints. =C2=A0MI code cannot use atomic ops for anything
> except plain ints, since no other type is guaranteed to be supported
> by the MD code. =C2=A0For example, sparc64 doesn't support any 8-bit or
> 16-bit types, and i386 doesn't support any 64-bit types (until recently;
> it now uses cmpxchg8b on some CPUs and a hack on other CPUs to support
> 64, bit types), and my version of i386 doesn't support any 8-bit or
> 16-bit types or long since these types are unusable in MI code and
> unused in MD code (long is misused in some MI code but I don't use
> such broken code).
>
> At least one type must be supported, and to keep things sane, int must
> be supported. =C2=A0The type could be int32_t instead, but then you would=
have
> a hard-coded assumption that int is int32_t instead of a hard-coded
> assumption that int can be an atomic type. =C2=A0The latter is a better
> assumption for MI code. =C2=A0But MD code can make the former assumption
> except of course on arches where it is false, but there are none of
> those yet. =C2=A0This gives the following structure in atomic.h:
>
> o basic definitions in terms of uint8/16/32/64_t. =C2=A0But don't define =
the
> =C2=A08/16/64 bit ones unless they are actually used in MD code, which is
> =C2=A0rare. =C2=A0MI code cannot use these.
> o assume that u_int is uint32_t and #define all the atomic_foo_int()
> =C2=A0interfaces as atomic_foo_32. =C2=A0Similarly for pointers, except u=
se
> =C2=A0atomic_foo_64 in some cases.
> o do something for long. =C2=A0MI code cannot use atomic ops on longs, bu=
t does.
> =C2=A0Correctly-sized longs are usually fundamentally non-atomic since th=
ey
> =C2=A0are twice as long as the register width and the register width is u=
sually
> =C2=A0the same as the bus width and atomicness for widths wider than the =
bus is
> =C2=A0unnatural and/or inefficient. =C2=A0But no supported arch has corre=
ctly-sized
> =C2=A0longs.
So the thing is that we define the uintXX_t type from int,long,etc and
then we should really expect to follow the same pattern for atomic
instructions.
>> Modified: projects/largeSMP/sys/powerpc/include/atomic.h
>>
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D
>> --- projects/largeSMP/sys/powerpc/include/atomic.h =C2=A0 =C2=A0 =C2=A0S=
at May =C2=A07
>> 23:34:14 2011 =C2=A0 =C2=A0 =C2=A0 =C2=A0(r221613)
>> +++ projects/largeSMP/sys/powerpc/include/atomic.h =C2=A0 =C2=A0 =C2=A0S=
un May =C2=A08
>> 00:39:49 2011 =C2=A0 =C2=A0 =C2=A0 =C2=A0(r221614)
>> @@ -48,13 +48,7 @@
>> =C2=A0* { *p +=3D v; }
>> =C2=A0*/
>>
>> -#define __ATOMIC_ADD_8(p, v, t) =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> - =C2=A0 =C2=A08-bit atomic_add not implemented
>> -
>> -#define __ATOMIC_ADD_16(p, v, t) =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> - =C2=A0 =C2=A016-bit atomic_add not implemented
>> -
>
> Already correctly left out except for bogusly #defining it and #undefinin=
g
> it. =C2=A0Now not bogusly #defined either.
I don't see the point to have something like that, also because is
never used in the kernel.
>> -#define __ATOMIC_ADD_32(p, v, t) =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> +#define __atomic_add_int(p, v, t) =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> =C2=A0 =C2=A0__asm __volatile( =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0=
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0"1: =C2=A0 =C2=A0 lwarx =C2=A0 %0, 0, %2\n" =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0\
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0" =C2=A0 =C2=A0 =C2=A0 add =C2=A0 =C2=A0 %0, =
%3, %0\n" =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> @@ -63,10 +57,10 @@
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0: "=3D&r" (t), "=3Dm" (*p) =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0: "r" (p), "r" (v), "m" (*p) =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0\
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0: "cc", "memory") =C2=A0 =C2=A0 =C2=A0 =C2=A0=
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> - =C2=A0 =C2=A0/* __ATOMIC_ADD_32 */
>> + =C2=A0 =C2=A0/* __atomic_add_int */
>
> No type problems at this level.
Yes, but I also want uniformity in the implementation, thus try to
give a common pattern.
>> #ifdef __powerpc64__
>> -#define __ATOMIC_ADD_64(p, v, t) =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> +#define __atomic_add_long(p, v, t) =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> =C2=A0 =C2=A0__asm __volatile( =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0=
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0"1: =C2=A0 =C2=A0 ldarx =C2=A0 %0, 0, %2\n" =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0\
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0" =C2=A0 =C2=A0 =C2=A0 add =C2=A0 =C2=A0 %0, =
%3, %0\n" =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> @@ -75,69 +69,78 @@
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0: "=3D&r" (t), "=3Dm" (*p) =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0: "r" (p), "r" (v), "m" (*p) =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0\
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0: "cc", "memory") =C2=A0 =C2=A0 =C2=A0 =C2=A0=
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> - =C2=A0 =C2=A0/* __ATOMIC_ADD_64 */
>> + =C2=A0 =C2=A0/* __atomic_add_long */
>> #else
>> -#define =C2=A0 =C2=A0 =C2=A0 =C2=A0__ATOMIC_ADD_64(p, v, t) =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> - =C2=A0 =C2=A064-bit atomic_add not implemented
>
> Now correctly left out when it is not supported.
See below, for non powerpc64 case it must not be used.
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0__atomic_add_long(p, v, t) =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0long atomic_add not implemented
>> #endif
>
> atomic_add_long() should never be supported, but it is (mis)used in
> standard kernel files so not having it will break more than just LINT.
> But the above is not atomic_add_long()...hacks are used below to make
> that sort of work.
In which sense "should never be supported"?
If you are on a 64 bit arch and you want and atomic operation for long
you are supposed to use 64-bit version directly? I don't get the
point.
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_acq_64 =C2=A0 =C2=A0 =C2=
=A0 atomic_add_acq_long
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_rel_64 =C2=A0 =C2=A0 =C2=
=A0 atomic_add_rel_long
>> +
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_ptr(p, v) =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_add_long((volatile u_long *)(p), (u_long)(=
v))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_acq_ptr(p, v) =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_add_acq_long((volatile u_long *)(p), (u_lo=
ng)(v))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_rel_ptr(p, v) =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_add_rel_long((volatile u_long *)(p), (u_lo=
ng)(v))
>
> This uses bogus casts to break warnings. =C2=A0amd64 is still missing thi=
s bug.
> Someone added this bug to i386 after reviewers objected to it. =C2=A0Even=
i386
> only has this bug for accesses to pointers. =C2=A0For char/short/long, it=
uses
> type-safe code with separate functions/#defines like the ones for int.
I'll probabilly may have get from the i386 version, I'll verify, thanks.
> The casts for pointers break jhb's intentions that:
> - atomic accesses to pointers should be rare
> - when they are needed, the caller must cast
>
> I don't know why we don't have separate functions/#defines for pointers,
> but like not having the bloat for it.
Nice history, I didn't know that explicitly, but I'll check that later
along with the cast thing.
>> +#else
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_long(p, v) =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \
>> + =C2=A0 =C2=A0 =C2=A0 atomic_add_int((volatile u_int *)(p), (u_int)(v))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_acq_long(p, v) =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 \
>> + =C2=A0 =C2=A0 =C2=A0 atomic_add_acq_int((volatile u_int *)(p), (u_int)=
(v))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_rel_long(p, v) =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 \
>> + =C2=A0 =C2=A0 =C2=A0 atomic_add_rel_int((volatile u_int *)(p), (u_int)=
(v))
>> +
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_ptr(p, v) =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_add_int((volatile u_int *)(p), (u_int)(v))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_acq_ptr(p, v) =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_add_acq_int((volatile u_int *)(p), (u_int)=
(v))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_add_rel_ptr(p, v) =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_add_rel_int((volatile u_int *)(p), (u_int)=
(v))
>> +#endif
>
> Now you need the bogus casts for the longs, since although atomic
> accesses to longs should be rarer than ones to pointers (never on MI code=
),
> it is not intended that the caller must bogusly cast for them, especially
> since the bogus casts are MD.
>
> The bogus casts for the pointers should have no effect except to break
> warnings in new code, since at least old MI code must already have the
> casts in callers else it wouldn't compile on amd64.
>
>> [... often I snip bits without noting this as here]
>> -#if 0
>> -_ATOMIC_CLEAR(8, 8, uint8_t)
>> -_ATOMIC_CLEAR(8, char, u_char)
>> -_ATOMIC_CLEAR(16, 16, uint16_t)
>> -_ATOMIC_CLEAR(16, short, u_short)
>> -#endif
>> -_ATOMIC_CLEAR(32, 32, uint32_t)
>> -_ATOMIC_CLEAR(32, int, u_int)
>> -#ifdef __powerpc64__
>> -_ATOMIC_CLEAR(64, 64, uint64_t)
>> -_ATOMIC_CLEAR(64, long, u_long)
>> -_ATOMIC_CLEAR(64, ptr, uintptr_t)
>> -#else
>> -_ATOMIC_CLEAR(32, long, u_long)
>> -_ATOMIC_CLEAR(32, ptr, uintptr_t)
>> -#endif
>
> Seems a better way, and still used on amd64 and i386, to generate
> separate type-safe code for each type supported. =C2=A0Here the only erro=
r
> relative to amd64 and i386 seems to have been to generate from long
> to a 32/64 suffix and #define from that back to a long suffix, instead
> of to generate from long to a long suffix and #define from that to a
> 32/64 suffix.
I'm a bit confused on what you are stating here exactly... are in
favor of this change or not?
>> ...
>> @@ -622,39 +638,59 @@ atomic_cmpset_rel_long(volatile u_long *
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0return (atomic_cmpset_long(p, cmpval, newval)=
);
>> }
>>
>> -#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_acq_int =C2=A0 atomic_=
cmpset_acq_32
>> -#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_rel_int =C2=A0 atomic_=
cmpset_rel_32
>> -
>> -#ifdef __powerpc64__
>> -#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_acq_ptr(dst, old, new)=
=C2=A0 =C2=A0\
>> - =C2=A0 =C2=A0atomic_cmpset_acq_long((volatile u_long *)(dst), (u_long)=
(old),
>> (u_long)(new))
>> -#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_rel_ptr(dst, old, new)=
=C2=A0 =C2=A0\
>> - =C2=A0 =C2=A0atomic_cmpset_rel_long((volatile u_long *)(dst), (u_long)=
(old),
>> (u_long)(new))
>> -#else
>> -#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_acq_ptr(dst, old, new)=
=C2=A0 =C2=A0\
>> - =C2=A0 =C2=A0atomic_cmpset_acq_32((volatile u_int *)(dst), (u_int)(old=
),
>> (u_int)(new))
>> -#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_rel_ptr(dst, old, new)=
=C2=A0 =C2=A0\
>> - =C2=A0 =C2=A0atomic_cmpset_rel_32((volatile u_int *)(dst), (u_int)(old=
),
>> (u_int)(new))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_64 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0atomic_cmpset_long
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_acq_64 =C2=A0 =C2=A0at=
omic_cmpset_acq_long
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_rel_64 =C2=A0 =C2=A0at=
omic_cmpset_rel_long
>> +
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_ptr(dst, old, new)
>> =C2=A0 =C2=A0 =C2=A0\
>
> No bogus casts for these now.
>
>> + =C2=A0 =C2=A0 =C2=A0 atomic_cmpset_long((volatile u_long *)(dst), (u_l=
ong)(old), =C2=A0 =C2=A0 \
>> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (u_long)(new))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_acq_ptr(dst, old, new)
>> =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_cmpset_acq_long((volatile u_long *)(dst), =
(u_long)(old), \
>> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (u_long)(new))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_rel_ptr(dst, old, new)
>> =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_cmpset_rel_long((volatile u_long *)(dst), =
(u_long)(old), \
>> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (u_long)(new))
>
> These bogus casts seem more bogus than before -- they seem to do nothing
> except break type safety, since you already have functions with the right
> suffixes and types (except the suffix says `long' but only u_longs are
> supported).
>
>> +#else
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_long(dst, old, new)
>> =C2=A0 =C2=A0 =C2=A0 \
>> + =C2=A0 =C2=A0 =C2=A0 atomic_cmpset_int((volatile u_int *)(dst), (u_int=
)(old), =C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (u_int)(new))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_acq_long(dst, old, new=
)
>> =C2=A0 =C2=A0 =C2=A0 \
>> + =C2=A0 =C2=A0 =C2=A0 atomic_cmpset_acq_int((volatile u_int *)(dst), (u=
_int)(old), =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (u_int)(new))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_rel_long(dst, old, new=
)
>> =C2=A0 =C2=A0 =C2=A0 \
>> + =C2=A0 =C2=A0 =C2=A0 atomic_cmpset_rel_int((volatile u_int *)(dst), (u=
_int)(old), =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (u_int)(new))
>
> Usual bogus casts to type-pun from long to int in the 32-bit case. =C2=A0=
Even
> here, the ones on the non-pointer args have no useful effect.
>
>> +
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_ptr(dst, old, new)
>> =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_cmpset_int((volatile u_int *)(dst), (u_int=
)(old), =C2=A0 =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (u_int)(new))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_acq_ptr(dst, old, new)
>> =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_cmpset_acq_int((volatile u_int *)(dst), (u=
_int)(old), =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (u_int)(new))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_cmpset_rel_ptr(dst, old, new)
>> =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 atomic_cmpset_rel_int((volatile u_int *)(dst), (u=
_int)(old), =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (u_int)(new))
>> #endif
>
> Usual bogus casts for the pointer case.
>
>> #ifdef __powerpc64__
>> -static __inline uint64_t
>> -atomic_fetchadd_64(volatile uint64_t *p, uint64_t v)
>> +static __inline u_long
>> +atomic_fetchadd_long(volatile u_long *p, u_long v)
>> {
>> - =C2=A0 =C2=A0 =C2=A0 uint64_t value;
>> + =C2=A0 =C2=A0 =C2=A0 u_long value;
>>
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0do {
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0value =3D *p;
>> @@ -662,10 +698,10 @@ atomic_fetchadd_64(volatile uint64_t *p,
>> =C2=A0 =C2=A0 =C2=A0 =C2=A0return (value);
>> }
>>
>> -#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_fetchadd_long =C2=A0 =C2=A0at=
omic_fetchadd_64
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_fetchadd_64 =C2=A0 =C2=A0 =C2=
=A0atomic_fetchadd_long
>> #else
>> -#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_fetchadd_long(p, v) =C2=A0 =
=C2=A0 =C2=A0\
>> - =C2=A0 =C2=A0(u_long)atomic_fetchadd_32((volatile u_int *)(p), (u_int)=
(v))
>> +#define =C2=A0 =C2=A0 =C2=A0 =C2=A0atomic_fetchadd_long(p, v)
>> =C2=A0 =C2=A0 =C2=A0\
>> + =C2=A0 =C2=A0 =C2=A0 (u_long)atomic_fetchadd_int((volatile u_int *)(p)=
, (u_int)(v))
>
> i386 uses a function to type-pun from fetchadd_long to fetchadd_int().
> This has minor technical advantages. =C2=A0Why be different?
I'm not entirely sure what you mean with 'function to type-pun' but
will look at that as well, thanks.
Attilio
--=20
Peace can only be achieved by understanding - A. Einstein
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?BANLkTi=mnZc6fy8TGpLMsMLDV8UpK5J-Gw>