Closed Bug 1552744 Opened 6 years ago Closed 4 years ago

Cranelift: Performance parity with Baldr on x86_64 for zlib

Categories

(Core :: JavaScript: WebAssembly, task, P3)

Product:
Core
Component:
JavaScript: WebAssembly
Platform:
x86_64
All
Type:
task

Tracking

()

Status:
RESOLVED INACTIVE

People

(Reporter: lth, Unassigned)

References

Details

The zlib embenchen benchmark is part of the set that determines whether we've reached parity.

As of this morning, cranelift code on zlib is very, very slow - barely more than 50% the speed of baldr code, and actually slower than rabaldr code.

It is known that some if this problem is from excessive spilling and redundant memory-to-memory copies but we've no in-depth analysis as of yet.

Priority: -- → P3
Status: NEW → RESOLVED
Closed: 4 years ago
Resolution: --- → INACTIVE
You need to log in before you can comment on or make changes to this bug.