HCoop / hcoop / debian / mlton.git / blame
| Commit | Line | Data |
|---|---|---|
| 7f918cf1 CE |
1 | DeepFlatten |
| 2 | =========== | |
| 3 | ||
| 4 | <:DeepFlatten:> is an optimization pass for the <:SSA2:> | |
| 5 | <:IntermediateLanguage:>, invoked from <:SSA2Simplify:>. | |
| 6 | ||
| 7 | == Description == | |
| 8 | ||
| 9 | This pass flattens into mutable fields of objects and into vectors. | |
| 10 | ||
| 11 | For example, an `(int * int) ref` is represented by a 2 word | |
| 12 | object, and an `(int * int) array` contains pairs of `int`-s, | |
| 13 | rather than pointers to pairs of `int`-s. | |
| 14 | ||
| 15 | == Implementation == | |
| 16 | ||
| 17 | * <!ViewGitFile(mlton,master,mlton/ssa/deep-flatten.fun)> | |
| 18 | ||
| 19 | == Details and Notes == | |
| 20 | ||
| 21 | There are some performance issues with the deep flatten pass, where it | |
| 22 | consumes an excessive amount of memory. | |
| 23 | ||
| 24 | * http://www.mlton.org/pipermail/mlton/2005-April/026990.html | |
| 25 | * http://www.mlton.org/pipermail/mlton-user/2010-June/001626.html | |
| 26 | * http://www.mlton.org/pipermail/mlton/2010-December/030876.html | |
| 27 | ||
| 28 | A number of applications require compilation with | |
| 29 | `-disable-pass deepFlatten` to avoid exceeding available memory. It is | |
| 30 | often asked whether the deep flatten pass usually has a significant | |
| 31 | impact on performance. The standard benchmark suite was run with and | |
| 32 | without the deep flatten pass enabled when the pass was first | |
| 33 | introduced: | |
| 34 | ||
| 35 | * http://www.mlton.org/pipermail/mlton/2004-August/025760.html | |
| 36 | ||
| 37 | The conclusion is that it does not have a significant impact. | |
| 38 | However, these are micro benchmarks; other applications may derive | |
| 39 | greater benefit from the pass. |