OCaml Forge
SCM

Forum: optimizations-in-hlvm

Posted by: Jon Harrop
Date: 2009-03-27 10:35
Summary:Optimizations in HLVM
Project:HLVM

HLVM now includes a simple term rewriter designed to support local optimization passes as well as a loop unrolling optimization pass.

The GC has also been optimized to record only addresses internally rather than (triple-word) references.

Unrolling is very beneficial in HLVM because it amortizes the cost of updating the shadow stack. Consequently, HLVM is now significantly faster than ocamlopt on all of the numerical benchmarks included in the HLVM test suite.

Latest News

Variant types and pattern matching in HLVM

Jon Harrop - 2010-04-28 09:37 -

HLVM on the ray tracer language comparison

Jon Harrop - 2010-01-21 17:01 -

High-performance parallelism with HLVM

Jon Harrop - 2010-01-03 18:30 -

Parallel programming in HLVM

Jon Harrop - 2009-11-30 09:08 -

New HLVM GC and LLVM 2.6 support

Jon Harrop - 2009-10-28 23:32 -

Discussion Forums: optimizations-in-hlvm

Monitor Forum | Start New Thread Start New Thread