From 8cd351ddc25bc5e6c549b99dc5fd3e9607f5a580 Mon Sep 17 00:00:00 2001 From: David Phillips Date: Tue, 23 May 2017 14:51:31 +1200 Subject: Update readme with new benchmarks --- README.md | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index cb18582..266cbaf 100644 --- a/README.md +++ b/README.md @@ -29,11 +29,18 @@ that keys are sane. Preliminary benching shows sand leek to be faster than some of the other similar tools out there when pushing work across cores. +I have also written a "slightly parallel" base32 algorithm which uses +SSSE3 if support is given from the compiler and target platform. +Preliminary benchmarks seem to indicate that this gives a performance +benefit of between roughly 3% and 30%. Although, these higher +performance gains seem to be only when running at a reduced worker +thread count. + | CPU(s) | Max throughput | -t | |---------------------------------------------|---------------:|---:| -| 2× Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz | 95 MH/s | 32 | -| Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz | 38 MH/s | 8 | -| AMD A4-1200 APU with Radeon(TM) HD Graphics | 2.3 MH/s | 2 | +| 2× Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz | 103.3 MH/s | 32 | +| Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz | 39.2 MH/s | 8 | +| AMD A4-1200 APU with Radeon(TM) HD Graphics | 2.6 MH/s | 2 | | ARMv6-compatible processor rev 7 (v6l) | 0.22 MH/s | 1 | ## Inspiration -- cgit v1.1