llvm-project.git/llvm/test/Transforms/FunctionSpecialization/function-specialization-stats.ll, branch users/nico/python-2

[FuncSpec][NFC] Command line option renaming.

2023-03-15T18:03:44+00:00

Standardize all options with 'funcspec' prefix and shorter abreviations.

Differential Revision: https://reviews.llvm.org/D145378

[FuncSpec] Minor refactoring in statistics and debug messages.

2023-03-15T09:10:06+00:00

* Remove redundant variable `NbFunctionsSpecialized` as it is no longer
  used by the cost model.
* Rename statistic `NumFuncSpecialized` to `NumSpecsCreated` as a better
  description (the old name confusingly implied number of functions we have
  created clones for).
* Same for variable `SpecializedFuncs`. Renamed to `Specializations`.
* Move debug message in the destructor (avoids repetition when MaxIters > 1).

Differential Revision: https://reviews.llvm.org/D145375

[IPSCCP] Create a Pass parameter to control specialization of functions.

2022-12-23T16:54:45+00:00

Required for D140210 in order to disable FuncSpec at {Os, Oz}
optimization levels.

Differential Revision: https://reviews.llvm.org/D140564

[FuncSpec] Make the Function Specializer part of the IPSCCP pass.

2022-12-10T14:39:49+00:00

Reland 877a9f9abec61f06e39f1cd872e37b828139c2d1 since D138654 (parent)
has been fixed with 9ebaf4fef4aac89d4eff08e48185d61bc893f14e and with
8f1e11c5a7d70f96943a72649daa69f152d73e90.

Differential Revision: https://reviews.llvm.org/D126455

Revert "[FuncSpec] Make the Function Specializer part of the IPSCCP pass."

2022-12-08T12:41:43+00:00

This reverts commit 877a9f9abec61f06e39f1cd872e37b828139c2d1.

It depends on the parent revision 42c2dc401742266da3e0251b6c1ca491f4779963
which needs to be reverted as it broke some buildbots, so reverting both.

[FuncSpec] Make the Function Specializer part of the IPSCCP pass.

2022-12-08T12:14:27+00:00

The aim of this patch is to minimize the compilation time overhead of
running Function Specialization. It is about 40% slower to run as a
standalone pass (IPSCCP + FuncSpec vs IPSCCP with FuncSpec) according
to my measurements. I compiled the llvm testsuite with NewPM-O3 + LTO
and measured single threaded [user + system] time of IPSCCP and FuncSpec
by passing the '-time-passes' option to lld. Then I compared the two
configurations in terms of Instruction Count of the total compilation
(not of the individual passes) as in https://llvm-compile-time-tracker.com.
Geomean for non-LTO builds is -0.25% and LTO is -0.5% approximately.

You can find more info below:

https://discourse.llvm.org/t/rfc-should-we-enable-function-specialization/61518

Differential Revision: https://reviews.llvm.org/D126455

[NFC] Port all FunctionSpecialization tests to `-passes=` syntax

2022-12-07T23:38:43+00:00

FunctionSpecialization: Convert tests to opaque pointers

2022-11-28T14:35:48+00:00

[FuncSpec] Don't specialize function which are easy to inline

2021-08-23T11:20:21+00:00

It would waste time to specialize a function which would inline finally.
This patch did two things:

- Don't specialize functions which are always-inline.
- Don't spescialize functions whose lines of code are less than threshold
(100 by default).

For spec2017int, this patch could reduce the number of specialized
functions by 33%. Then the compile time didn't increase for every
benchmark.

Reviewed By: SjoerdMeijer, xbolva00, snehasish

Differential Revision: https://reviews.llvm.org/D107897

[FuncSpec] Remove other passes in a test RUN line. NFC.

2021-06-16T09:36:22+00:00