RFR: 8349457: [leyden] Workaround Leyden problems with -XX:-TieredCompilation

[Vladimir Ivanov]

On Wed, 5 Feb 2025 13:59:03 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

>> src/hotspot/share/compiler/precompiler.cpp line 243:
>> 
>>> 241:     }
>>> 242: 
>>> 243:     if (TieredCompilation) {
>> 
>> How does this work as far as consistency between training/assembly run and production run is concerned? Do we need to enforce the same setting across these runs or can we e.g. train and then (assuming we fork) assemble with -TieredCompilation but run in production with +TieredCompilation -- or vice versa?
>
> I tested different combinations of `{+|-}TieredCompilation` with training and production runs, and it works well. I believe the normal mechanisms make it right. I.e. there is no guarantee that lower-tiers code would always be present with `+TieredCompilation`. So, no C1 code? Fine, nothing loaded.
> 
> The loaders are also aware of the compilers that are present. So, tier{2,3} code in archive, but C1 is disabled? We are not loading it then.
> 
> From performance angle, there seem to be a slight performance edge in training with C2 only, and then running with C1+C2: no penalty on loading lots of C1 compiled code, and quicker transition from intepreter to "real" C1 in only a few places where it is needed! Compiler dynamics is weird. This PR gains us access to these kind of studies :)

As I mentioned during the meeting, it looks weird to me to see `if (TieredCompilation)` checks in Precompiler code which is executed only during assembly phase. As a data point, regular CDS archive is used to be generated in interpreter-only mode (to make the process reproducible). 

Instead of supporting arbitrary JVM configurations during assembly phase, I'd prefer to see offending JVM parameters being overridden/dropped.

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

PR Review Comment: https://git.openjdk.org/leyden/pull/32#discussion_r1945512817