Crate honeycomb_benches 

honeycomb-benches

This crate contains all benchmarks of the project. It also contains simple binaries used to profile and further optimize the implementation.

Binary

The package provides a single binary, hc-bench, which exposes several benchmarks as subcommands. For details on options and arguments, run:

cargo run --bin hc-bench -- --help

Benchmarks are described in the documentation of their respective modules.

Features

Optional features can be enabled to affect implementation:

• bind-threads – enabled by default – uses hwlocality to bind threads to physical cores,
• jemalloc – uses tikv-jemallocator to replace the default allocator,
• profiling – enable perf fifo interactions to allow per-section profiling,
• _single_precision – compile cargo benches (not the binary) to use f32 instead of f64.

Available benchmarks

Criterion-based

• builder - grid building routines at fixed size
• builder-grid-size - grid building routines over a range of grid sizes
• fetch_icells - CMap2::iter_<CELL> methods
• grisubal - grisubal kernel with a fixed size grid
• grisubal-grid-size - grisubal kernel over a range of grid granularity
• triangulate-quads - triangulate all cells of a mixed-mesh

Iai-callgrind-based

• prof-dim2-basic - CMap2 basic operations benchmarks
• prof-dim2-build - CMap2 constructor & building functions benchmarks
• prof-dim2-sewing-unsewing - CMap2 (un)sewing & (un)linking methods benchmarks

Modules

cut_edges

cut-edges benchmark

grid_gen

Grid generation benchmark

grisubal

grisubal benchmark

remesh

2D remshing pipeline benchmark

shift

shift benchmark

Macros

prof_init

Attempt to open a fifo at the path /tmp/hc_perf_control.

prof_start

Write to the /tmp/hc_perf_control to enable perf sampling if ${$var} is defined.

prof_stop

Write to the /tmp/hc_perf_control to disable perf sampling if ${$var} is defined.