Reduce compilation times #328
Comments
fthomas
added a commit
that referenced
this issue
Sep 25, 2017
This specializes the macro implementations for `autoRefineV` and
`autoRefineT`. Both methods were defined in terms of the generic
`RefineMacro.impl` which works for any `RefType[F]`. This generic
implementation has the drawback that it needs to `eval` the implicit
`RefType` parameter at compile-time. Since `autoRefine{V,T}` already
know which `RefType` instance they will use, it is unnecessary to `eval`
them in the macro.
Here are the benchmark results for master and this change:
```
master:
[info] Result "eu.timepit.refined.benchmark.RefineMacroBenchmark.autoRefineV_PosInt":
[info] 0.549 ±(99.9%) 0.004 s/op [Average]
[info] (min, avg, max) = (0.520, 0.549, 0.593), stdev = 0.015
[info] CI (99.9%): [0.546, 0.553] (assumes normal distribution)
[info] # Run complete. Total time: 00:17:45
[info] Benchmark Mode Cnt Score Error Units
[info] RefineMacroBenchmark.autoRefineV_PosInt avgt 200 0.549 ± 0.004 s/op
impl_Refined:
[info] Result "eu.timepit.refined.benchmark.RefineMacroBenchmark.autoRefineV_PosInt":
[info] 0.403 ±(99.9%) 0.004 s/op [Average]
[info] (min, avg, max) = (0.377, 0.403, 0.471), stdev = 0.017
[info] CI (99.9%): [0.399, 0.407] (assumes normal distribution)
[info] # Run complete. Total time: 00:09:07
[info] Benchmark Mode Cnt Score Error Units
[info] RefineMacroBenchmark.autoRefineV_PosInt avgt 200 0.403 ± 0.004 s/op
```
That is a decrease in compilation time of ~26%. This is first PR towards
reduced compilation times: #328
fthomas
added a commit
that referenced
this issue
Sep 25, 2017
This specializes the macro implementations for `autoRefineV` and
`autoRefineT`. Both methods were defined in terms of the generic
`RefineMacro.impl` which works for any `RefType[F]`. This generic
implementation has the drawback that it needs to `eval` the implicit
`RefType` parameter at compile-time. Since `autoRefine{V,T}` already
know which `RefType` instance they will use, it is unnecessary to `eval`
them in the macro.
Here are the benchmark results for master and this change:
```
master:
[info] Result "eu.timepit.refined.benchmark.RefineMacroBenchmark.autoRefineV_PosInt":
[info] 0.549 ±(99.9%) 0.004 s/op [Average]
[info] (min, avg, max) = (0.520, 0.549, 0.593), stdev = 0.015
[info] CI (99.9%): [0.546, 0.553] (assumes normal distribution)
[info] # Run complete. Total time: 00:17:45
[info] Benchmark Mode Cnt Score Error Units
[info] RefineMacroBenchmark.autoRefineV_PosInt avgt 200 0.549 ± 0.004 s/op
impl_Refined:
[info] Result "eu.timepit.refined.benchmark.RefineMacroBenchmark.autoRefineV_PosInt":
[info] 0.403 ±(99.9%) 0.004 s/op [Average]
[info] (min, avg, max) = (0.377, 0.403, 0.471), stdev = 0.017
[info] CI (99.9%): [0.399, 0.407] (assumes normal distribution)
[info] # Run complete. Total time: 00:09:07
[info] Benchmark Mode Cnt Score Error Units
[info] RefineMacroBenchmark.autoRefineV_PosInt avgt 200 0.403 ± 0.004 s/op
```
That is a decrease in compilation time of ~26%. This is first PR towards
reduced compilation times: #328
|
The first step towards this goal has been merged: ~26% reduced compilation times with #332. |
|
Reduced compilation time when refining with full refined types: #388 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
This tweet https://twitter.com/missingfaktor/status/909064863917510656 shows that increased compilation time with refined is a major issue. We should investigate ways to get better in this regard.
The text was updated successfully, but these errors were encountered: