v1.1.0

v1.1.0

Overview

This release introduced explicit support for netcoreapp2.1, and is mainly focused on perf-improvements over v1.0.0. Some key-points are:

• improved SIMD reduction #49
• loop unrolling "eats" more iterations #52
• extensively used intruction level parallelism (ISP) #57
• SIMD operations are done aligned to register size #60

Perf-improvement

Perf improved about +25...+200% (depending on data-size, whether parallelized compuation is done or not, and so on).
As example a graph for the combined compuation of average and variance is shown.

v1.1.0-preview-3

v1.1.0-preview-3

Overview

This release introduced explicit support for netcoreapp2.1, and is mainly focused on perf-improvements over v1.0.0. Some key-points are:

• improved SIMD reduction #49
• loop unrolling "eats" more iterations #52
• extensively used intruction level parallelism (ISP) #57
• SIMD operations are done aligned to register size #60

Perf-improvement

Perf improved about +25...+200% (depending on data-size, whether parallelized compuation is done or not, and so on).
As example a graph for the combined compuation of average and variance is shown.