diff options
| author | Paul E. Murphy <murp@ibm.com> | 2021-03-10 17:06:54 -0600 |
|---|---|---|
| committer | Lynn Boger <laboger@linux.vnet.ibm.com> | 2021-03-15 12:30:38 +0000 |
| commit | 4350e4961a6ea3d36a33271423735b37c96dd5bf (patch) | |
| tree | 6315d542cb50cdb176f4b73e098689fe9004d424 /src/bytes/buffer.go | |
| parent | 6ccb5c49cca52766f6d288d128db67be6392c579 (diff) | |
| download | go-4350e4961a6ea3d36a33271423735b37c96dd5bf.tar.xz | |
crypto/md5: improve ppc64x performance
This is mostly cleanup and simplification. This removes
many unneeded register moves, loads, and bit twiddlings
which were holdovers from porting this from the amd64
version.
The updated code loads each block once per iteration
instead of once per round. Similarly, the logical
operations now match the original md5 specification.
Likewise, add extra sizes to the benchtest to give more
data points on how the implementation scales with input
size.
All in all, this is roughly a 20% improvement on ppc64le
code running on POWER9 (POWER8 is similar, but around
16%):
name old time/op new time/op delta
Hash8Bytes 297ns ± 0% 255ns ± 0% -14.14%
Hash64 527ns ± 0% 444ns ± 0% -15.76%
Hash128 771ns ± 0% 645ns ± 0% -16.35%
Hash256 1.26µs ± 0% 1.05µs ± 0% -16.68%
Hash512 2.23µs ± 0% 1.85µs ± 0% -16.82%
Hash1K 4.16µs ± 0% 3.46µs ± 0% -16.83%
Hash8K 31.2µs ± 0% 26.0µs ± 0% -16.74%
Hash1M 3.58ms ± 0% 2.98ms ± 0% -16.74%
Hash8M 26.1ms ± 0% 21.7ms ± 0% -16.81%
Hash8BytesUnaligned 297ns ± 0% 255ns ± 0% -14.08%
Hash1KUnaligned 4.16µs ± 0% 3.46µs ± 0% -16.79%
Hash8KUnaligned 31.2µs ± 0% 26.0µs ± 0% -16.78%
name old speed new speed delta
Hash8Bytes 26.9MB/s ± 0% 31.4MB/s ± 0% +16.45%
Hash64 122MB/s ± 0% 144MB/s ± 0% +18.69%
Hash128 166MB/s ± 0% 199MB/s ± 0% +19.54%
Hash256 203MB/s ± 0% 244MB/s ± 0% +20.01%
Hash512 230MB/s ± 0% 276MB/s ± 0% +20.18%
Hash1K 246MB/s ± 0% 296MB/s ± 0% +20.26%
Hash8K 263MB/s ± 0% 315MB/s ± 0% +20.11%
Hash1M 293MB/s ± 0% 352MB/s ± 0% +20.10%
Hash8M 321MB/s ± 0% 386MB/s ± 0% +20.21%
Hash8BytesUnaligned 26.9MB/s ± 0% 31.4MB/s ± 0% +16.41%
Hash1KUnaligned 246MB/s ± 0% 296MB/s ± 0% +20.19%
Hash8KUnaligned 263MB/s ± 0% 315MB/s ± 0% +20.15%
Change-Id: I269bfa6878966bb4f6a64dc349100f5dc453ab7c
Reviewed-on: https://go-review.googlesource.com/c/go/+/300613
Run-TryBot: Paul Murphy <murp@ibm.com>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>
Diffstat (limited to 'src/bytes/buffer.go')
0 files changed, 0 insertions, 0 deletions