Context
In-memory benchmarks (500MB, 1GB) for all codecs have been removed because the MoonBit native runtime truncates array/Bytes length to 28 bits, silently corrupting allocations >= 256MB.
See: moonbitlang/moonbit-docs#1155
What was removed
Go and MoonBit in-memory benchmarks at these sizes:
{flate,gzip,zlib}_compress_default_{500mb,1gb}{flate,gzip,zlib}_decompress_{500mb,1gb}
The benchmark generator (tools/gen_benchmarks.py) caps SIZES at 100MB.
When to re-enable
Once the upstream issue is resolved, update SIZES in tools/gen_benchmarks.py to include ("500mb", 524288000) and ("1gb", 1073741824), re-run the generator, and add corresponding Go benchmarks back to tools/bench_test.go.
Streaming benchmarks are unaffected
Streaming benchmarks process data in 4096-byte chunks and never hold > 256MB in a single Bytes, so they work at any size.
Context
In-memory benchmarks (500MB, 1GB) for all codecs have been removed because the MoonBit native runtime truncates array/Bytes length to 28 bits, silently corrupting allocations >= 256MB.
See: moonbitlang/moonbit-docs#1155
What was removed
Go and MoonBit in-memory benchmarks at these sizes:
{flate,gzip,zlib}_compress_default_{500mb,1gb}{flate,gzip,zlib}_decompress_{500mb,1gb}The benchmark generator (
tools/gen_benchmarks.py) capsSIZESat 100MB.When to re-enable
Once the upstream issue is resolved, update
SIZESintools/gen_benchmarks.pyto include("500mb", 524288000)and("1gb", 1073741824), re-run the generator, and add corresponding Go benchmarks back totools/bench_test.go.Streaming benchmarks are unaffected
Streaming benchmarks process data in 4096-byte chunks and never hold > 256MB in a single
Bytes, so they work at any size.