RFR: 8349814: [leyden] Avoid uncommon traps in preload code [v2]

[Vladimir Ivanov]

On Tue, 11 Feb 2025 16:39:39 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

>> Preload code is the C2 code we load very early, before it gets substituted by more advanced C2 code. We are seeing lots of uncommon traps from the preloaded code, though, which invalidates it too early. We can do precompile code without uncommon traps to alleviate this. 
>> 
>> Additional testing:
>>  - [x] Ad-hoc perf tests (see comment below)
>>  - [x] Linux x86_64 server fastdebug, `runtime/cds`
>
> Aleksey Shipilev has updated the pull request with a new target base due to a merge or a rebase. The incremental webrev excludes the unrelated changes brought in by the merge/rebase. The pull request contains three additional commits since the last revision:
> 
>  - Merge branch 'premain' into JDK-8349814-preload-avoid-uncommon-traps
>  - Comment
>  - Fix

src/hotspot/share/compiler/precompiler.cpp line 235:

> 233: 
> 234:     {
> 235:       // For preload code, avoid any uncommon traps.

Why don't you instrument `Compile::too_many_traps()` instead? (Check for compilation reason and unconditionally return true when compiling code for preloading.)  

Also, it would be useful to make new logic conditional (controlled by a diagnostic flag). It would make it easier to measure deoptimization-induced performance effects at runtime.

src/hotspot/share/runtime/deoptimization.cpp line 2130:

> 2128:                                 tm->name_and_sig_as_C_string(), trap_bci, nm->compiler_name());
> 2129: 
> 2130:       if (nm->preloaded() && (action != Action_none)) {

At some point, I added SCC-related info to `TraceDeoptimization` output [1], but it turns out I didn't upstream it. (I had a script which parsed the log and aggregated information about deopts in archived code.) 

Maybe it's better to upstream it instead of introducing such adhoc logging.

[1] https://github.com/iwanowww/leyden/blame/premain-precompile/src/hotspot/share/runtime/deoptimization.cpp#L2151

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

PR Review Comment: https://git.openjdk.org/leyden/pull/38#discussion_r1951247171
PR Review Comment: https://git.openjdk.org/leyden/pull/38#discussion_r1951576538