Replies: 3 comments 2 replies
-
Hi @Excelnabb ,
Thank you!
To be honest, we haven't tried with the EE of GraamVM. Ideally, it should work. However, TornadoVM depends on the Graal IR (intermediate compiler representation). Since we haven't used the EE, we can't tell the IR scheduling and structure of the Graal IR EE is that different that TornadoVM IR cannot understand. For example, auto-vectorization. TornadoVM also does vectorization for OpenCL and SPIR-V using the Graal IR in a specific format. What we know is that it should work well with Graal CE. But feel free to give it a try.
TornadoVM for "Windows" is experimental. Indeed, what we currently do is compile TornadoVM through MSys64. I personally use Windows 11 and it seems to work well for PTX and OpenCL. Alternatively, you can use WSL2 (e.g., Ubuntu). In my case, it works really well. the installation is exactly the same as any other Linux System, and you can use it also for running on Intel Integrated GPUs, Intel ARC GPUs (discrete) and FPGAs (emulation mode) if you have Intel CPU/GPU. To answer your question: No, you do not have to your Java code in the same location as your Java home. You can use For reference, this is my #!/bin/bash
export JAVA_HOME="C:\Users\jjfum\Documents\bin\jvms\graalvm-ce-java17-windows-amd64-22.2.0\graalvm-ce-java17-22.2.0"
export CUDA_PATH="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.6"
export PTX_LDFLAGS=-L\"$CUDA_PATH/lib/x64\"
# LEAVE THE REST OF FILE 'AS IS'
# DON'T ALTER!
export PATH=$PWD/bin/bin:$PATH ## This directory will be automatically generated during Tornado compilation
export TORNADO_SDK=$PWD/bin/sdk ## This directory will be automatically generated during Tornado compilation
CMAKE_FILE=$(where cmake | head -n 1)
export CMAKE_ROOT=${CMAKE_FILE%\\*\\*}
It seems there is an error with some of the maven dependencies. Can you try again? Also, note that TornadoVM assumes a correct installation of CUDA, or any supported vendor implementation for OpenCL (e.g., Intel, AMD, etc). For example, if you have an NVIDIA GPU, quite likely the driver is already installed. You only need to install the CUDA SDK.
Hope this helps. Feel free to ask and open new issues. We are keen to get feedback from the community. |
Beta Was this translation helpful? Give feedback.
-
Hi |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
Hi
First of all, thanks for developing such a program!
I'm trying to follow this guide to run it on Windows 10:
https://github.com/beehive-lab/TornadoVM/blob/master/tornado-assembly/src/docs/20_INSTALL_WINDOWS_WITH_GRAALVM.md
However I stumble on some problems:
It says that I need to install and download GraalVM Community Edition 22.2.0 (https://github.com/graalvm/graalvm-ce-builds/releases/tag/vm-22.2.0).
For the program I want to run with the GPU I'm already running with the enterprise version of the code, which is 20% faster graalvm-ee-java17-windows-amd64-22.3.0 will this work? Or do I need to use the community edition 22.2.0?
When it says # UPDATE PATH TO ACTUAL LOCATION OF THE JDK OR GRAAL (REMEMBER OF UNIX_STYLE SLASHES AND SPACES!!!)
export JAVA_HOME="C:/graalvm-ce-java11-22.2.0"
Does this mean the location of the Java code for the particular program I want to run with GPU? In my case: C:\StrategyQuantX\j64
(folder where I extracted the graalvm-ee-java17-windows-amd64-22.3.0 )
Lastly, when I try to compile the program, I get these errors, not sure how to fix them: Error on building Tornado.txt
Thank you!
Beta Was this translation helpful? Give feedback.
All reactions