mirror_zfs/module/zcommon/zfs_fletcher_intel.c
Gvozden Neskovic fc897b24b2 Rework of fletcher_4 module
- Benchmark memory block is increased to 128kiB to reflect real block sizes more
accurately. Measurements include all three stages needed for checksum generation,
i.e. `init()/compute()/fini()`. The inner loop is repeated multiple times to offset
overhead of time function.

- Fastest implementation selects native and byteswap methods independently in
benchmark. To support this new function pointers `init_byteswap()/fini_byteswap()`
are introduced.

- Implementation mutex lock is replaced by atomic variable.

- To save time, benchmark is not executed in userspace. Instead, highest supported
implementation is used for fastest. Default userspace selector is still 'cycle'.

- `fletcher_4_native/byteswap()` methods use incremental methods to finish
calculation if data size is not multiple of vector stride (currently 64B).

- Added `fletcher_4_native_varsize()` special purpose method for use when buffer size
is not known in advance. The method does not enforce 4B alignment on buffer size, and
will ignore last (size % 4) bytes of the data buffer.

- Benchmark `kstat` is changed to match the one of vdev_raidz. It now shows
throughput for all supported implementations (in B/s), native and byteswap,
as well as the code [fastest] is running.

Example of `fletcher_4_bench` running on `Intel(R) Xeon(R) CPU E5-2660 v3 @ 2.60GHz`:
implementation   native         byteswap
scalar           4768120823     3426105750
sse2             7947841777     4318964249
ssse3            7951922722     6112191941
avx2             13269714358    11043200912
fastest          avx2           avx2

Example of `fletcher_4_bench` running on `Intel(R) Xeon Phi(TM) CPU 7210 @ 1.30GHz`:
implementation   native         byteswap
scalar           1291115967     1031555336
sse2             2539571138     1280970926
ssse3            2537778746     1080016762
avx2             4950749767     1078493449
avx512f          9581379998     4010029046
fastest          avx512f        avx512f

Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #4952
2016-08-16 14:11:55 -07:00

151 lines
4.7 KiB
C

/*
* Implement fast Fletcher4 with AVX2 instructions. (x86_64)
*
* Use the 256-bit AVX2 SIMD instructions and registers to compute
* Fletcher4 in four incremental 64-bit parallel accumulator streams,
* and then combine the streams to form the final four checksum words.
*
* Copyright (C) 2015 Intel Corporation.
*
* Authors:
* James Guilford <james.guilford@intel.com>
* Jinshan Xiong <jinshan.xiong@intel.com>
*
* This software is available to you under a choice of one of two
* licenses. You may choose to be licensed under the terms of the GNU
* General Public License (GPL) Version 2, available from the file
* COPYING in the main directory of this source tree, or the
* OpenIB.org BSD license below:
*
* Redistribution and use in source and binary forms, with or
* without modification, are permitted provided that the following
* conditions are met:
*
* - Redistributions of source code must retain the above
* copyright notice, this list of conditions and the following
* disclaimer.
*
* - Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following
* disclaimer in the documentation and/or other materials
* provided with the distribution.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#if defined(HAVE_AVX) && defined(HAVE_AVX2)
#include <linux/simd_x86.h>
#include <sys/spa_checksum.h>
#include <zfs_fletcher.h>
static void
fletcher_4_avx2_init(zio_cksum_t *zcp)
{
kfpu_begin();
/* clear avx2 registers */
asm volatile("vpxor %ymm0, %ymm0, %ymm0");
asm volatile("vpxor %ymm1, %ymm1, %ymm1");
asm volatile("vpxor %ymm2, %ymm2, %ymm2");
asm volatile("vpxor %ymm3, %ymm3, %ymm3");
}
static void
fletcher_4_avx2_fini(zio_cksum_t *zcp)
{
uint64_t __attribute__((aligned(32))) a[4];
uint64_t __attribute__((aligned(32))) b[4];
uint64_t __attribute__((aligned(32))) c[4];
uint64_t __attribute__((aligned(32))) d[4];
uint64_t A, B, C, D;
asm volatile("vmovdqu %%ymm0, %0":"=m" (a));
asm volatile("vmovdqu %%ymm1, %0":"=m" (b));
asm volatile("vmovdqu %%ymm2, %0":"=m" (c));
asm volatile("vmovdqu %%ymm3, %0":"=m" (d));
asm volatile("vzeroupper");
kfpu_end();
A = a[0] + a[1] + a[2] + a[3];
B = 0 - a[1] - 2*a[2] - 3*a[3]
+ 4*b[0] + 4*b[1] + 4*b[2] + 4*b[3];
C = a[2] + 3*a[3]
- 6*b[0] - 10*b[1] - 14*b[2] - 18*b[3]
+ 16*c[0] + 16*c[1] + 16*c[2] + 16*c[3];
D = 0 - a[3]
+ 4*b[0] + 10*b[1] + 20*b[2] + 34*b[3]
- 48*c[0] - 64*c[1] - 80*c[2] - 96*c[3]
+ 64*d[0] + 64*d[1] + 64*d[2] + 64*d[3];
ZIO_SET_CHECKSUM(zcp, A, B, C, D);
}
static void
fletcher_4_avx2_native(const void *buf, uint64_t size, zio_cksum_t *unused)
{
const uint64_t *ip = buf;
const uint64_t *ipend = (uint64_t *)((uint8_t *)ip + size);
for (; ip < ipend; ip += 2) {
asm volatile("vpmovzxdq %0, %%ymm4"::"m" (*ip));
asm volatile("vpaddq %ymm4, %ymm0, %ymm0");
asm volatile("vpaddq %ymm0, %ymm1, %ymm1");
asm volatile("vpaddq %ymm1, %ymm2, %ymm2");
asm volatile("vpaddq %ymm2, %ymm3, %ymm3");
}
}
static void
fletcher_4_avx2_byteswap(const void *buf, uint64_t size, zio_cksum_t *unused)
{
static const struct {
uint64_t v[4] __attribute__((aligned(32)));
} mask = {
.v = { 0xFFFFFFFF00010203, 0xFFFFFFFF08090A0B,
0xFFFFFFFF00010203, 0xFFFFFFFF08090A0B }
};
const uint64_t *ip = buf;
const uint64_t *ipend = (uint64_t *)((uint8_t *)ip + size);
asm volatile("vmovdqa %0, %%ymm5"::"m"(mask));
for (; ip < ipend; ip += 2) {
asm volatile("vpmovzxdq %0, %%ymm4"::"m" (*ip));
asm volatile("vpshufb %ymm5, %ymm4, %ymm4");
asm volatile("vpaddq %ymm4, %ymm0, %ymm0");
asm volatile("vpaddq %ymm0, %ymm1, %ymm1");
asm volatile("vpaddq %ymm1, %ymm2, %ymm2");
asm volatile("vpaddq %ymm2, %ymm3, %ymm3");
}
}
static boolean_t fletcher_4_avx2_valid(void)
{
return (zfs_avx_available() && zfs_avx2_available());
}
const fletcher_4_ops_t fletcher_4_avx2_ops = {
.init_native = fletcher_4_avx2_init,
.fini_native = fletcher_4_avx2_fini,
.compute_native = fletcher_4_avx2_native,
.init_byteswap = fletcher_4_avx2_init,
.fini_byteswap = fletcher_4_avx2_fini,
.compute_byteswap = fletcher_4_avx2_byteswap,
.valid = fletcher_4_avx2_valid,
.name = "avx2"
};
#endif /* defined(HAVE_AVX) && defined(HAVE_AVX2) */