about summary refs log tree commit diff
path: root/default.nix
diff options
context:
space:
mode:
authorVincent Ambo <tazjin@tvl.su>2024-08-12T22·21+0300
committertazjin <tazjin@tvl.su>2024-08-19T11·02+0000
commitadf9b4c54aadae7e1ba0bb9ab30efb5043d9843a (patch)
tree05f700b4b11e0c3d98b518f21adb7015e7d124d1 /default.nix
parentd6c57eb957abc9c9101779600e04b34209d5c436 (diff)
refactor(tvix/eval): remove use of imbl::Vector r/8520
This vector type has served us well for now, but it contains internal refcounts
which are incompatible with upcoming changes related to garbage collection.

The performance impact of this change within all benchmarks I ran was within the
margin of error:

[nix-shell:/tmp/perf]$ hyperfine "./before -E '(import <nixpkgs> {}).firefox.outPath' --log-level ERROR --no-warnings"
Benchmark 1: ./u64 -E '(import <nixpkgs> {}).firefox.outPath' --log-level ERROR --no-warnings
  Time (mean ± σ):      7.528 s ±  0.272 s    [User: 6.578 s, System: 0.631 s]
  Range (min … max):    7.160 s …  8.012 s    10 runs

nix-shell:/tmp/perf]$ hyperfine "./std-vec -E '(import <nixpkgs> {}).firefox.outPath' --log-level ERROR --no-warnings"
Benchmark 1: ./std-vec -E '(import <nixpkgs> {}).firefox.outPath' --log-level ERROR --no-warnings
  Time (mean ± σ):      7.515 s ±  0.178 s    [User: 6.508 s, System: 0.652 s]
  Range (min … max):    7.276 s …  7.861 s    10 runs

Change-Id: Ib95f871956e336a1e5771f6293583854b1efb276
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12197
Reviewed-by: aspen <root@gws.fyi>
Tested-by: BuildkiteCI
Diffstat (limited to 'default.nix')
0 files changed, 0 insertions, 0 deletions