-
Currently, enabling "Tiered JIT" will only show you the first tier of the tiered JIT.
It'd be handy to be able to choose the tier you want to inspect.
ghost updated
4 years ago
-
I'm seeing a big difference in computation time between `jax.jit()` and `jax.jit().lower().compile()`. Jitting and then executing is faster than precompiling. I expect that performance should be the …
-
### Run Information
Name | Value
-- | --
Architecture | x64
OS | ubuntu 22.04
Queue | ViperUbuntu
Baseline | [24e7d1b0959234a36fb09a5c64cfd2da20f10618](https://github.com/dotnet/runtime/co…
-
I am trying to reproduce the MNIST classification task with an MPS ansatz and encountered very slow compilation times when using jit. And even without jitting, the MPS contraction seems to be slower f…
-
```
Execution should not block on compiling a function with LLVM or
reoptimizing it with new data. We should send these work units to separate
worker threads, allowing the main threads to carry on uni…
-
```
We currently use llvm::ExecutionEngine::getPointerToFunction(func) to
translate LLVM IR into machine code. getPointerToFunction caches the
returned machine code for each input function, and will n…
-
```
The patch available at
http://codereview.appspot.com/206091
supports building the LLVM JIT as a dynamic extension module, _llvmjit. The
basic idea is that all calls into LLVM get indirected by a…
-
```
Problem reported at http://www.openintents.org/en/node/682
"When using OI Safe with roms that have dalvik vm JIT enabled, the database
decrypts without posting any errors, but the records are co…
-
```
Problem reported at http://www.openintents.org/en/node/682
"When using OI Safe with roms that have dalvik vm JIT enabled, the database
decrypts without posting any errors, but the records are co…
-
### Description
Hello team,
I have found unexpected behavior when using `shard_map`, `custom_vjp`, and `pallas` (`triton` works the same) simultaneously during auto grad. Please take a look at the…