RFR: 8326615: C1/C2 don't handle allocation failure properly during initialization (RuntimeStub::new_runtime_stub fatal crash) [v5]

[Damon Fenacci]

On Tue, 23 Jul 2024 16:45:26 GMT, Vladimir Kozlov <kvn at openjdk.org> wrote:

> You should run build/linux-x64/images/jdk/bin/java instead which is used in production.

Oh, I see! Thanks @vnkozlov!
Here are the updated values:

### `-version`
|                | Int   | C1 only | C2 only | C1/C2 tiered |
|----------------|-------|---------|---------|--------------|
| linux-x64      | 515Kb | 1125Kb  | 604Kb   | 1200Kb       |
| macosx-aarch64 | 479Kb | 1080Kb  | 542Kb   | 1135Kb       |
| linux-aarch64  | 478Kb | 1085Kb  | 542Kb   | 1166Kb       |

### `HelloWorld.java`
|                | Int   | C1 only | C2 only | C1/C2 tiered |
|----------------|-------|---------|---------|--------------|
| linux-x64      | 720Kb | 6956Kb  | 1289Kb  | 5771Kb       |
| macosx-aarch64 | 609Kb | 2760Kb  | 1138Kb  | 3768Kb       |
| linux-aarch64  | 608Kb | 6833Kb  | 1155Kb  | 5827Kb       |

I was still puzzled by C1 values being higher than C2 ones but noticed that most of the difference is made by the _C1 temporary CodeBuffer_(s) (563Kb each on `linux-x64`).

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

PR Comment: https://git.openjdk.org/jdk/pull/19280#issuecomment-2247953556