Timofey Titovets <nefelim4ag@xxxxxxxxx> writes:
> xxhash much faster then jhash,
> ex. for x86_64 host:
> PAGE_SIZE: 4096, loop count: 1048576
> jhash2: 0xacbc7a5b time: 1907 ms, th: 2251.9 MiB/s
> xxhash32: 0x570da981 time: 739 ms, th: 5809.4 MiB/s
> xxhash64: 0xa1fa032ab85bbb62 time: 371 ms, th: 11556.6 MiB/s
>
> xxhash64 on x86_32 work with ~ same speed as jhash2.
> xxhash32 on x86_32 work with ~ same speed as for x86_64
Which CPU is that?
>
> So replace jhash with xxhash,
> and use fastest version for current target ARCH.
Can you do some macro-benchmarking too? Something that uses
KSM and show how the performance changes.
You could manually increase the scan rate to make it easier
to see.
> @@ -51,6 +52,12 @@
> #define DO_NUMA(x) do { } while (0)
> #endif
>
> +#if BITS_PER_LONG == 64
> +typedef u64 xxhash;
> +#else
> +typedef u32 xxhash;
> +#endif
This should be in xxhash.h ?
xxhash_t would seem to be a better name.
> - u32 checksum;
> + xxhash checksum;
> void *addr = kmap_atomic(page);
> - checksum = jhash2(addr, PAGE_SIZE / 4, 17);
> +#if BITS_PER_LONG == 64
> + checksum = xxh64(addr, PAGE_SIZE, 0);
> +#else
> + checksum = xxh32(addr, PAGE_SIZE, 0);
> +#endif
This should also be generic in xxhash.h
-Andi
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
