Re: ZFS committed to the FreeBSD base.

On Thu, Apr 12, 2007 at 11:06:03AM -0500, Rick C. Petty wrote:
On Thu, Apr 12, 2007 at 10:54:17AM +0200, Dag-Erling Sm?rgrav wrote:

Our native atomic operations are all defined as either macros or
static inline functions in machine/atomic.h, so we can easily make
this choice at compile time based on a config option.

Is there any way we could make the choice at boot time, by checking for
presence of the CX8 feature? Either as something like:

extern int feature_cx8; /* or MIB variable */
#define CMPXCHG8(a) (feature_cx8 ? { _asm "..." } : emulate_cmpxch8(a))

For something this low level my opinion is it's better to stay with
compile time options. After all, in the above example, cmpxchg8 is a
single machine instruction. How much overhead does it add to retrieve a
variable from memory and check it, then jump to the correct place?
Enough that it outweighs the benefit of using that instruction in the
first place?

For entire functions that have been optimized (bzero comes to mind) you
can always either use a function pointer or overwrite the code in memory
with the optimized version. The function call overhead presumably isn't
that much compared to the work that the function is doing.

Otherwise something like ZFS which utilizes this feature a lot could
check the MIB variable and set different fn ptr in its device structure,
or something along those lines. Of course, that would require compiling
the same code twice essentially, but it had the advantage that it would
work on non-CX8 systems and that it would be fast on systems with CX8.

I agree this makes sense for some things, but atomic operations are
supposed to be as fast as possible -- preferably single machine
instructions I can't think of anything short of JIT compiling the kernel
that wouldn't be a high price to pay.

Craig
_______________________________________________
freebsd-current@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@xxxxxxxxxxx"

Relevant Pages

  • Re: performance of hardware dynamic scheduling
    ... I suspect enough IPL can be obtained at compile time. ... The biggest problem with static scheduling of instructions ... instruction scheduling can hide the ... Today someone told me that the access latency of L1 cache in general ...
    (comp.arch)
  • Re: As the Crow flies.
    ... I can do this in a single assembly language instruction: ... At compile time, it takes as long as the ... Both REP STOSB and REP MOVS require that you _know_ the count in ... important in high memory is burned into ROM so no problem. ...
    (talk.origins)
  • Re: As the Crow flies.
    ... I can do this in a single assembly language instruction: ... something of a surprise to the guys who wrote the ISO-C89/99 ... At compile time, it takes as long as the ... Both REP STOSB and REP MOVS require that you _know_ the count in ...
    (talk.origins)
  • Re: ZFS committed to the FreeBSD base.
    ... static inline functions in machine/atomic.h, ... this choice at compile time based on a config option. ... presence of the CX8 feature? ... work on non-CX8 systems and that it would be fast on systems with CX8. ...
    (freebsd-current)