From 849aae3b1440cc2516592ce97c09541db06242fc Mon Sep 17 00:00:00 2001 From: Milan K Date: Wed, 4 Nov 2020 16:22:39 +0000 Subject: [PATCH] Update README.md --- README.md | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/README.md b/README.md index 3c83c14..da1c6d0 100644 --- a/README.md +++ b/README.md @@ -56,10 +56,9 @@ And similarly for the other number types. Then on an Intel(R) Core(R) i5 (Ice La | rounding mode | Float32 | BFloat16 | Float64 | Float16 | | --------------------- | ---------- | ---------- | --------- | --------- | -| default | 460.421 μs | 588.813 μs | 1.151ms | 16.446 ms | -| + stochastic rounding | 2.458 ms | 3.398 ms | n/a | 17.318 ms | +| default | 460 μs | 556 μs | 1.151ms | 16.446 ms | +| + stochastic rounding | 2.585 ms | 3.820 ms | n/a | 20.714 ms | -Stochastic rounding imposes an about x5 performance decrease for Float32/BFloat16, but is negligible for Float16. For Float32sr about 50% of the time is -spend on the random number generation, a bit less than 50% on the addition in -Float64 and the rest is the addition of the random number on the result and -round to nearest. +Stochastic rounding imposes an about x5-7 performance decrease for Float32/BFloat16, but is almost negligible for Float16. +For Float32sr about 50% of the time is spend on the random number generation, a bit less than 50% on the addition in +Float64 and the rest is the addition of the random number on the result and round to nearest.