Running Examples and Benchmarks
Running TornadoVM Programs
TornadoVM includes a tool for launching applications from the command-line:
$ tornado --help
usage: tornado [-h] [--version] [-version] [--debug] [--threadInfo] [--igv] [--igvLowTier] [--printKernel] [--printBytecodes] [--enableProfiler ENABLE_PROFILER] [--dumpProfiler DUMP_PROFILER] [--printJavaFlags] [--devices] [--ea]
[--module-path MODULE_PATH] [--classpath CLASSPATH] [--jvm JVM_OPTIONS] [-m MODULE_APPLICATION] [-jar JAR_FILE] [--params APPLICATION_PARAMETERS]
[application]
Tool for running TornadoVM Applications. This tool sets all Java options for enabling TornadoVM.
positional arguments:
application
optional arguments:
-h, --help show this help message and exit
--version Print version of TornadoVM
-version Print JVM Version
--debug Enable debug mode
--threadInfo Print thread deploy information per task on the accelerator
--igv Debug Compilation Graphs using Ideal Graph Visualizer (IGV)
--igvLowTier Debug Low Tier Compilation Graphs using Ideal Graph Visualizer (IGV)
--printKernel, -pk Print generated kernel (OpenCL, PTX or SPIR-V)
--printBytecodes, -pc
Print the generated TornadoVM bytecodes
--enableProfiler ENABLE_PROFILER
Enable the profiler {silent|console}
--dumpProfiler DUMP_PROFILER
Dump the profiler to a file
--printJavaFlags Print all the Java flags to enable the execution with TornadoVM
--devices Print information about the accelerators available
--ea, -ea Enable assertions
--module-path MODULE_PATH
Module path option for the JVM
--classpath CLASSPATH, -cp CLASSPATH, --cp CLASSPATH
Set class-path
--jvm JVM_OPTIONS, -J JVM_OPTIONS
Pass Java options to the JVM. Use without spaces: e.g., --jvm="-Xms10g" or -J"-Xms10g"
-m MODULE_APPLICATION
Application using Java modules
-jar JAR_FILE Main Java application in a JAR File
--params APPLICATION_PARAMETERS
Command-line parameters for the host-application. Example: --params="param1 param2..."
Some examples:
$ tornado -m tornado.examples/uk.ac.manchester.tornado.examples.compute.MatrixMultiplication1D
Use the following command to identify the ids of the devices that are being discovered by TornadoVM:
$ tornado --devices
Every device discovered by TornadoVM is associated with a pair of id numbers that correspond to the type of driver and the specific device, as follows:
Tornado device=<driverNumber>:<deviceNumber>
Example output:
Number of Tornado drivers: 2
Total number of PTX devices : 1
Tornado device=0:0
PTX -- GeForce GTX 1650
Global Memory Size: 3.8 GB
Local Memory Size: 48.0 KB
Workgroup Dimensions: 3
Max WorkGroup Configuration: [1024, 1024, 64]
Device OpenCL C version: N/A
Total number of OpenCL devices : 4
Tornado device=1:0
NVIDIA CUDA -- GeForce GTX 1650
Global Memory Size: 3.8 GB
Local Memory Size: 48.0 KB
Workgroup Dimensions: 3
Max WorkGroup Configuration: [1024, 1024, 64]
Device OpenCL C version: OpenCL C 1.2
Tornado device=1:1
Intel(R) OpenCL HD Graphics -- Intel(R) Gen9 HD Graphics NEO
Global Memory Size: 24.8 GB
Local Memory Size: 64.0 KB
Workgroup Dimensions: 3
Max WorkGroup Configuration: [256, 256, 256]
Device OpenCL C version: OpenCL C 2.0
Tornado device=1:2
Intel(R) OpenCL -- Intel(R) Core(TM) i7-7700HQ CPU @ 2.80GHz
Global Memory Size: 31.0 GB
Local Memory Size: 32.0 KB
Workgroup Dimensions: 3
Max WorkGroup Configuration: [8192, 8192, 8192]
Device OpenCL C version: OpenCL C 1.2
Tornado device=1:3
AMD Accelerated Parallel Processing -- Intel(R) Core(TM) i7-7700HQ CPU @ 2.80GHz
Global Memory Size: 31.0 GB
Local Memory Size: 32.0 KB
Workgroup Dimensions: 3
Max WorkGroup Configuration: [1024, 1024, 1024]
Device OpenCL C version: OpenCL C 1.2
The output might vary depending on which backends you have included in the build process. To run TornadoVM, you should see at least one device.
To run on a specific device use the following option:
-D<g>.<t>.device=<driverNumber>:<deviceNumber>
Where g is the TaskGraph name and t is the task name.
For example running on driver:device 1:1 (Intel HD Graphics in our example) will look like this:
$ tornado --jvm="-Ds0.t0.device=1:1" -m tornado.examples/uk.ac.manchester.tornado.examples.compute.MatrixMultiplication1D
The command above will run the MatrixMultiplication1D example on the integrated GPU (Intel HD Graphics).
Benchmarking
Running all benchmarks with default values
$ tornado-benchmarks.py
Running TornadoVM Benchmarks
[INFO] This process takes between 30-60 minutes
List of benchmarks:
*saxpy
*addImage
*stencil
*convolvearray
*convolveimage
*blackscholes
*montecarlo
*blurFilter
*renderTrack
*euler
*nbody
*sgemm
*dgemm
*mandelbrot
*dft
[INFO] TornadoVM options: -Xms24G -Xmx24G -server
...
Running a specific benchmark
$ tornado -m tornado.benchmarks/uk.ac.manchester.tornado.benchmarks.BenchmarkRunner --params="sgemm"
Unittests
To run all unittests in Tornado:
$ make tests
To run an individual unittest:
$ tornado-test uk.ac.manchester.tornado.unittests.TestHello
Also, it can be executed in verbose mode:
$ tornado-test --verbose uk.ac.manchester.tornado.unittests.TestHello
To test just a method of a unittest class:
$ tornado-test --verbose uk.ac.manchester.tornado.unittests.TestHello#testHello
To see the OpenCL/PTX generated kernel for a unittest:
$ tornado-test --verbose -pk uk.ac.manchester.tornado.unittests.TestHello#testHello
To execute in debug mode:
$ tornado-test --verbose --debug uk.ac.manchester.tornado.unittests.TestHello#testHello
task info: s0.t0
platform : NVIDIA CUDA
device : GeForce GTX 1050 CL_DEVICE_TYPE_GPU (available)
dims : 1
global work offset: [0]
global work size : [8]
local work size : [8]