Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Refactor one of the core runtime classes - API new methods for adding Compilation Flags #539

Merged
merged 10 commits into from
Aug 26, 2024

Conversation

jjfumero
Copy link
Member

Description

This PR refactors the old AbstractMetaData class and it adds a new method in the TornadoExecutionPlan to select the compilation flags. This addition is due to the refactoring the of the aforementioned class.

API Changes:

 try (TornadoExecutionPlan executionPlan = new TornadoExecutionPlan(taskGraph.snapshot())) {
            executionPlan.withCompilerFlags(TornadoVMBackendType.OPENCL, "-cl-opt-disable") //
                    .withCompilerFlags(TornadoVMBackendType.PTX, " ") //
                    .withCompilerFlags(TornadoVMBackendType.SPIRV, "-ze-opt-level 1") //
                    .execute();
        }

We can specify the flags we want per backend. This is by design, because the TornadoVM runtime can migrate execution from one device to another, and even from different backends.

Problem description

This PR removes old fields and old functionality no longer needed.

Backend/s tested

Mark the backends affected by this PR.

  • OpenCL
  • PTX
  • SPIRV

OS tested

Mark the OS where this PR is tested.

  • Linux
  • OSx
  • Windows

Did you check on FPGAs?

If it is applicable, check your changes on FPGAs.

  • Yes
  • No

How to test the new patch?

make tests
tornado-test -V uk.ac.manchester.tornado.unittests.compiler.TestCompilerFlagsAPI

if (meta instanceof TaskMetaData) {
TaskMetaData metaData = (TaskMetaData) task.meta();
TaskContextInterface meta = task.meta();
if (meta instanceof TaskDataContext) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this can be simplified to avoid type casting

/**
* Default PTX Compiler Flags.
*/
public static final String DEFAULT_PTX_COMPILER_FLAGS = getProperty("tornado.ptx.compiler.flags", "");
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

are we sure that PTX does not require any flags that we can use by default?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I did not detect we pass any flags to the PTX driver compiler.

@@ -571,38 +571,6 @@ public boolean useDefaultThreadScheduler() {
return defaultScheduler;
}

public void createImmutableExecutionContext(TornadoExecutionContext executionContext) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why we removed this?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This method is never invoked.

@mikepapadim
Copy link
Member

LGTM, just a few comments

Copy link
Collaborator

@stratika stratika left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, it looks good to me. Shall we test if this option works also for the FPGAs? Have you tested it? is it supposed to overlap with the compiler flags for HLS in the configuration file?

jjfumero and others added 4 commits August 26, 2024 09:56
…ExecutionPlan.java

Co-authored-by: Thanos Stratikopoulos <34061419+stratika@users.noreply.github.com>
…ExecutionPlan.java

Co-authored-by: Thanos Stratikopoulos <34061419+stratika@users.noreply.github.com>
…ExecutionPlan.java

Co-authored-by: Thanos Stratikopoulos <34061419+stratika@users.noreply.github.com>
…ExecutionPlan.java

Co-authored-by: Thanos Stratikopoulos <34061419+stratika@users.noreply.github.com>
@jjfumero
Copy link
Member Author

Thanks, it looks good to me. Shall we test if this option works also for the FPGAs? Have you tested it? is it supposed to overlap with the compiler flags for HLS in the configuration file?

I did not test on FPGAs. It does not overlap with the FPGA compilation option, but let's do some testing.

@stratika
Copy link
Collaborator

Thanks, it looks good to me. Shall we test if this option works also for the FPGAs? Have you tested it? is it supposed to overlap with the compiler flags for HLS in the configuration file?

I did not test on FPGAs. It does not overlap with the FPGA compilation option, but let's do some testing.

I have just tried it in emulation mode, and it works.

@jjfumero
Copy link
Member Author

Great, I'll merge this PR.

@jjfumero jjfumero merged commit 68abbcd into beehive-lab:develop Aug 26, 2024
2 checks passed
@jjfumero jjfumero deleted the feat/plan/api branch August 26, 2024 09:23
jjfumero added a commit to jjfumero/TornadoVM that referenced this pull request Aug 30, 2024
Improvements
~~~~~~~~~~~~~~~~~~

- beehive-lab#468: Cleanup Abstract Metadata Class.
- beehive-lab#473: Add maven plugin to build TornadoVM source for the releases.
- beehive-lab#474: Refactor `<X>TornadoDevice` to place common methods in the `TornadoXPUInterface`.
- beehive-lab#482: Help messages improve when an out-of-memory exception is raised.
- beehive-lab#484: Double-type for the trigonometric functions added in the `TornadoMath` class.
- beehive-lab#487: Prebuilt API simplified.
- beehive-lab#494: Add test to trigger unsupported features related to direct use of Memory Segments.
- beehive-lab#509: Add a quick pass configuration to skip the heavy tests during active development.
- beehive-lab#532: Improve thread scheduler to support RISC-V Accelerators from Codeplay.
- beehive-lab#533: Support for scalar values to be passed via lambda expressions as tasks.
- beehive-lab#538: `README` file updated.
- beehive-lab#539: Refactor core classes and add new API methods to pass compilation flags to the low-level driver compilers (OpenCL, PTX and Level Zero).
- beehive-lab#542: Tagged LevelZero JNI and Beehive Toolkit dependencies added in the build and installer.

Compatibility
~~~~~~~~~~~~~~~~~~

- beehive-lab#465: Support for JDK 22 and GraalVM 24.0.2.
- beehive-lab#486: Temurin for Windows added in the list of supported JDKs.
- beehive-lab#525: Revert usage of String Templates in preparation for JDK 23.
- beehive-lab#527: SPIR-V version parameter added. TornadoVM may run previous SPIR-V versions (e.g., ComputeAorta from Codeplay).
- beehive-lab#513: LevelZero JNI Library updated to v0.1.4.

Bug Fixes
~~~~~~~~~~~~~~~~~~

- beehive-lab#470: README documentation fixed.
- beehive-lab#478: Fix the test names that are present in the white list.
- beehive-lab#488: FP64 Kind for radian operations and the PTX backend fixed.
- beehive-lab#493: Tests Whitelist for PTX backend fixed.
- beehive-lab#502: Fix barrier type in the documentation regarding programmability of reductions.
- beehive-lab#514: Installer script fixed.
- beehive-lab#540: Fix  issue with clean-up execution IDs function.
- beehive-lab#541: Fix Data Accessors for the prebuilt API.
- beehive-lab#543: Fix checkstyle condition and FP16 error message improved.
jjfumero added a commit to jjfumero/TornadoVM that referenced this pull request Aug 30, 2024
Improvements
~~~~~~~~~~~~~~~~~~

- beehive-lab#468: Cleanup Abstract Metadata Class.
- beehive-lab#473: Add maven plugin to build TornadoVM source for the releases.
- beehive-lab#474: Refactor `<X>TornadoDevice` to place common methods in the `TornadoXPUInterface`.
- beehive-lab#482: Help messages improved when an out-of-memory exception is raised.
- beehive-lab#484: Double-type for the trigonometric functions added in the `TornadoMath` class.
- beehive-lab#487: Prebuilt API simplified.
- beehive-lab#494: Add test to trigger unsupported features related to direct use of Memory Segments.
- beehive-lab#509: Add a quick pass configuration to skip the heavy tests during active development.
- beehive-lab#532: Improve thread scheduler to support RISC-V Accelerators from Codeplay.
- beehive-lab#533: Support for scalar values to be passed via lambda expressions as tasks.
- beehive-lab#538: `README` file updated.
- beehive-lab#539: Refactor core classes and add new API methods to pass compilation flags to the low-level driver compilers (OpenCL, PTX and Level Zero).
- beehive-lab#542: Tagged LevelZero JNI and Beehive Toolkit dependencies added in the build and installer.

Compatibility
~~~~~~~~~~~~~~~~~~

- beehive-lab#465: Support for JDK 22 and GraalVM 24.0.2.
- beehive-lab#486: Temurin for Windows added in the list of supported JDKs.
- beehive-lab#525: Revert usage of String Templates in preparation for JDK 23.
- beehive-lab#527: SPIR-V version parameter added. TornadoVM may run previous SPIR-V versions (e.g., ComputeAorta from Codeplay).
- beehive-lab#513: LevelZero JNI Library updated to v0.1.4.

Bug Fixes
~~~~~~~~~~~~~~~~~~

- beehive-lab#470: README documentation fixed.
- beehive-lab#478: Fix the test names that are present in the white list.
- beehive-lab#488: FP64 Kind for radian operations and the PTX backend fixed.
- beehive-lab#493: Tests Whitelist for PTX backend fixed.
- beehive-lab#502: Fix barrier type in the documentation regarding programmability of reductions.
- beehive-lab#514: Installer script fixed.
- beehive-lab#540: Fix  issue with clean-up execution IDs function.
- beehive-lab#541: Fix Data Accessors for the prebuilt API.
- beehive-lab#543: Fix checkstyle condition and FP16 error message improved.

minor change
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
Development

Successfully merging this pull request may close these issues.

3 participants