Re: [Qemu-devel] [PATCH v2 00/10] tb hash improvements

[Alex Bennée]

Emilio G. Cota <address@hidden> writes:

> See v1 for context:
>  https://lists.gnu.org/archive/html/qemu-devel/2016-04/msg00587.html
>
> All patches in v2 are checkpatch-clean, except 05 (checkpatch should
> be ignored for this one) and 06, which I took unmodified (later patches
> fix those warnings while doing other things, anyway).
>
> Note that patch 01 has already been picked up by Paolo; I left it
> here for completeness.
>
> Another patch that is related to this series is the transition of
> tb->flags to uint32_t; this has been sent as a separate patch since
> it touches all targets. It shouldn't conflict with this patchset.

I've just gotten through v1 so I'll leave this for others to look at and
pick up my review on v3. I have been benchmarking though and I like what
I see.

>
> Changes from v1:
> - Drop QEMU_CACHELINE, define QEMU_ALIGNED()
>   + Remove excessive caution about Windows: it supports it.
>   + define QHT_BUCKET_ALIGN as 64
> - Add some reviewed-by tags from Alex.
> - Drop POSIX spinlock wrapper; use the one pointed out by Paolo
>   + Add a couple of fixes over this spinlock implementation:
>     * define cpu_relax() for some architectures
>     * Optimize spin_lock() for uncontended cases
> - Add tb_hash_func5, a version of xxhash32 customized for tb-hash, so
>   that the input values do not have to be contiguous in memory.
>   + Drop xxhash.h, add only the customized version at exec/tb-hash-xx.h
> - qht
>   + use size_t for counts in qht
>     * Drop some 2**32 size checks in there; it's not really a problem
>       in practice.
>   + s/__func/func__locked/
>   + Test program: hash with tb_hash_func5
>
> Thanks,
>
>               Emilio


--
Alex Bennée