RFR: 8254317: C2: Resource consumption of ConvI2LNode::Ideal() grows exponentially

[Roberto Castañeda Lozano]

On Mon, 19 Oct 2020 09:01:48 GMT, Roberto Castañeda Lozano <rcastanedalo at openjdk.org> wrote:

>> Tested on tier1-3 on windows-x64, linux-x64, linux-aarch64, and macosx-x64 in both release and debug mode.
>
> The microbenchmark gives the following results before and after the proposed fix (linux-x86_64-server-release build run
> on Intel Core i7-9850H CPU @ 2.60GHz with 32 GB main memory): | Version | Mode | Cnt | Score | Error | Units |
> |---|---|---|---|---|---|
> | before the fix | ss | 10 | 15402.110 | 171.255 | ms/op |
> | after the fix | ss | 10 | 0.976 | 0.069 | ms/op |
> 
> "Single-shot" (ss) benchmarking mode is used here because the focus is on measuring C2 execution time.

In the optimization ConvI2L(AddI(x, y)) -> AddL(ConvI2L(x), ConvI2L(y)) within
ConvI2LNode::Ideal(), handle the special case x = y by feeding both inputs of
AddL from a single ConvI2L node rather than creating two semantically equivalent
ConvI2L nodes. This avoids an exponential number of calls to
ConvI2LNode::Ideal() when dealing with long chains of AddI nodes. Disable the
optimization for the pattern ConvI2L(SubI(x, x)), which is simplified to zero
during parsing anyway. Add a set of regression tests for the transformation that
cover different shapes of AddI subgraphs. Also add a microbenchmark that
exercises the special case, for performance regression testing.

-------------

PR: https://git.openjdk.java.net/jdk/pull/727