Use of Best Platform Specific Instruction in JIT Compilation and VM Code Execution

eclipse-openj9 / openj9

Eclipse OpenJ9: A Java Virtual Machine for OpenJDK that's optimized for small footprint, fast start-up, and high throughput. Builds on Eclipse OMR (https://github.com/eclipse/omr) and combines with the Extensions for OpenJDK for OpenJ9 repo.

Other

3.27k stars 720 forks source link

Use of Best Platform Specific Instruction in JIT Compilation and VM Code Execution #10373

Open sirinath opened 4 years ago

sirinath commented 4 years ago

When JIT compiling and running of the VM is it possible to ensure, perhaps through additional tests, that the bytecode gets compiled and VM implementation, is the best possible set of instruction. This can include ensuring:

vectorisation
SIMD instructions
GPU acceleration
Other platform-specific instruction in the compiled VM code and JIT-compiled code corresponds to a set of independently written platform-specific optimal assembly instructions.

gacholio commented 4 years ago

The VM is entirely in C and is compiled only once for the minimum supported hardware level, so there's nothing to be done in the VM.

gacholio commented 4 years ago

The JIT already selects the best instructions available, so there's no work for the JIT either.

I suggest closing this issue.

sirinath commented 4 years ago

My issue is Graal VM EE and all other JVMs outperforms OpenJ9 so there must be some room for improvements in terms of:

VM overhead
The efficiency of the JITed code

sirinath commented 4 years ago

Also, it might be an idea to implement a C/C++ compiler and interpreter (like Cling) using OMR so you have control over the VM compilation also to better optimise this code and efficiency.