RFR: 8302150: Speed up compiler/codegen/Test7100757.java

[Tobias Hartmann]

On Fri, 17 Feb 2023 15:46:25 GMT, Emanuel Peter <epeter at openjdk.org> wrote:

> This test took over `11 seconds`. It had a very high iteration count of `5_000_000`.
> 
> I lowered it to `50_000`, it now takes about `0.7 seconds`, so over `15x speedup`.
> 
> I ensured that the same compilations still happen, with and without `-Xbatch`, this is the reason why I could not set it all the way down to `10_000` (the speedup at this point was hardly noticable anyway):
> 
> 
>     554 1023       3       compiler.codegen.Test7100757::test (274 bytes)
>     561 1028 %     4       compiler.codegen.Test7100757::test @ 14 (274 bytes)
>     578 1033       4       compiler.codegen.Test7100757::test (274 bytes)
>     593 1023       3       compiler.codegen.Test7100757::test (274 bytes)   made not entrant
>     599 1034 %     3       compiler.codegen.Test7100757::main @ 32 (65 bytes)
>     600 1035       3       compiler.codegen.Test7100757::main (65 bytes)
>     615 1036 %     4       compiler.codegen.Test7100757::main @ 32 (65 bytes)
>     635 1034 %     3       compiler.codegen.Test7100757::main @ 32 (65 bytes)   made not entrant
>     676 1036 %     4       compiler.codegen.Test7100757::main @ 32 (65 bytes)   made not entrant
> 
> 
> I am not sure if the test only reproduced intermittently, but if it did it would now probably reproduce with lower frequency, but hopefully still eventually (maybe 100x less less often). Or maybe the original author just did not notice that the iteration count was a bit excessive.

Looks good.

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

Marked as reviewed by thartmann (Reviewer).

PR: https://git.openjdk.org/jdk/pull/12618