wip: GPU acceleration

Author: WrldEngine

.gitignore

@@ -101,6 +101,9 @@ ipython_config.py
 #   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
 poetry.lock
 
+# UV
+uv.lock
+
 # Cargo
 Cargo.lock
 
@@ -166,4 +169,8 @@ cython_debug/
 
 # Rust
 /target
-.env
+.env
+
+test.py
+
+*.class

CONTRIBUTING.md

@@ -3,5 +3,9 @@
 Thank you
 
 ## Rollout
-1. `pip install -r dev-requirements.txt` - Installing python dependencies
-2. `cargo build` - Installing rust dependencies
+1. `pip install -r dev-requirements.txt` or alternatively `uv sync` - Installing python dependencies
+2. `cargo build` - Installing rust dependencies
+
+## Run benchmarks
+- `uv run pytest python/benches/compare_benchmark_test.py` (via uv)
+- `pytest python/benches/compare_benchmark_test.py`

Cargo.toml

@@ -19,10 +19,10 @@ crate-type = ["cdylib", "rlib"]
 numpy = "0.25.0"
 pyo3 = { version = "0.25.1", features = ["extension-module"] }
 rayon = "1.10.0"
+anyhow = "1.0"
 
-[dependencies.opencl3]
-version = "0.11"
-features = ["CL_VERSION_2_1", "CL_VERSION_2_2", "CL_VERSION_3_0"]
+[dependencies.rem_math_gpu]
+path = "rem_math_gpu"
 
 [dev-dependencies]
 criterion = "0.3"
@@ -34,3 +34,8 @@ harness = false
 [profile.release]
 lto = true
 codegen-units = 1
+
+[workspace]
+members = [
+  "rem_math_gpu", "rem_math_simd",
+]

README.md

@@ -155,9 +155,17 @@ Found 15 outliers among 100 measurements (15.00%)
   8 (8.00%) high severe
 ```
 
+## Integrations (java will be done later)
+<img height="50" src="https://raw.githubusercontent.com/marwin1991/profile-technology-icons/refs/heads/main/icons/python.png">
+<img height="50" src="https://raw.githubusercontent.com/marwin1991/profile-technology-icons/refs/heads/main/icons/java.png">
+
+## GPU Integrations
+![nVIDIA](https://img.shields.io/badge/cuda-000000.svg?style=for-the-badge&logo=nVIDIA&logoColor=green)
+![nVIDIA](https://img.shields.io/badge/opencl-000000.svg?style=for-the-badge&logo=opencl&logoColor=green)
+
 ## Roadmap
 
-- Add GPU-accelerated operations for improved performance.
+- Add GPU-accelerated operations for improved performance. (__in progress__)
 - Implement own custom type objects for best performance from ecosystem.
 - Expand mathematical functionality with additional features and algorithms.
 

dev-requirements.txt

@@ -1,10 +1,61 @@
-colorama==0.4.6 ; python_version >= "3.11" and sys_platform == "win32"
-iniconfig==2.1.0 ; python_version >= "3.11"
-maturin==1.9.0 ; python_version >= "3.11"
-numpy==2.3.1 ; python_version >= "3.11"
-packaging==25.0 ; python_version >= "3.11"
-pluggy==1.6.0 ; python_version >= "3.11"
-py-cpuinfo==9.0.0 ; python_version >= "3.11"
-pygments==2.19.2 ; python_version >= "3.11"
-pytest-benchmark==5.1.0 ; python_version >= "3.11"
-pytest==8.4.1 ; python_version >= "3.11"
+# This file was autogenerated by uv via the following command:
+#    uv export --group dev --no-hashes --format requirements-txt
+-e .
+click==8.3.1
+    # via mkdocs
+colorama==0.4.6 ; sys_platform == 'win32'
+    # via
+    #   click
+    #   mkdocs
+    #   pytest
+ghp-import==2.1.0
+    # via mkdocs
+iniconfig==2.3.0
+    # via pytest
+jinja2==3.1.6
+    # via mkdocs
+markdown==3.10
+    # via mkdocs
+markupsafe==3.0.3
+    # via
+    #   jinja2
+    #   mkdocs
+maturin==1.10.2
+mergedeep==1.3.4
+    # via
+    #   mkdocs
+    #   mkdocs-get-deps
+mkdocs==1.6.1
+mkdocs-get-deps==0.2.0
+    # via mkdocs
+numpy==2.4.0
+packaging==25.0
+    # via
+    #   mkdocs
+    #   pytest
+pathspec==0.12.1
+    # via mkdocs
+platformdirs==4.5.1
+    # via mkdocs-get-deps
+pluggy==1.6.0
+    # via pytest
+py-cpuinfo==9.0.0
+    # via pytest-benchmark
+pygments==2.19.2
+    # via pytest
+pytest==9.0.2
+    # via pytest-benchmark
+pytest-benchmark==5.2.3
+python-dateutil==2.9.0.post0
+    # via ghp-import
+pyyaml==6.0.3
+    # via
+    #   mkdocs
+    #   mkdocs-get-deps
+    #   pyyaml-env-tag
+pyyaml-env-tag==1.1
+    # via mkdocs
+six==1.17.0
+    # via python-dateutil
+watchdog==6.0.0
+    # via mkdocs

java/Cargo.toml

@@ -0,0 +1,8 @@
+[package]
+name = "rem_math_java"
+version = "0.1.0"
+edition = "2024"
+
+[dependencies]
+anyhow = "1.0"
+jni = "0.21.1"

java/RemMath.h

@@ -0,0 +1,23 @@
+/*
+ * Experimental module.
+ *
+ * This code is under active development and provided for internal testing only.
+ * APIs, behavior, and implementation details are unstable and may change
+ * without notice. Do not rely on this functionality in production or
+ * draw conclusions from current results.
+*/
+
+class RemMath {
+    private static native int[] sum_two_ints32(int[] a, int[] b, String method);
+    static {
+        System.loadLibrary("rem_math");
+    }
+
+    public static void main(String[] args) {
+        int[] a = {1, 2, 3};
+        int[] b = {1, 2, 3};
+
+        int[] output = RemMath.sum_two_ints32(a, b, "gpu");
+        System.out.println(output);
+    }
+}

java/RemMath.java

@@ -0,0 +1,20 @@
+/*
+ RemMath Java SDK integration.
+ Note, this is template function, it will be replaced soon
+*/
+
+use jni::JNIEnv;
+use jni::objects::{JClass, JString};
+use jni::sys::jstring;
+
+#[no_mangle]
+pub extern "system" fn Java_HelloWorld_hello<'local>(
+    mut env: JNIEnv<'local>,
+    class: JClass<'local>,
+    input: JString<'local>
+) -> jstring {
+    let input: String = env.get_string(&input).expect("Couldn't get java string!").into();
+    let output = env.new_string(format!("Hello, {}!", input)).expect("Couldn't create java string!");
+
+    output.into_raw()
+}

java/src/lib.rs

@@ -10,16 +10,18 @@ requires-python = ">=3.11"
 dependencies = [
 ]
 
-
 [build-system]
 requires = ["maturin>=1.0,<2.0"]
 build-backend = "maturin"
 
 [tool.maturin]
+python-source = "python"
 module-name = "rem_math._rem_math"
 
-[tool.poetry.group.dev.dependencies]
-pytest-benchmark = "^5.1.0"
-numpy = "^2.3.1"
-maturin = "^1.9.0"
-mkdocs = "^1.6.1"
+[dependency-groups]
+dev = [
+    "maturin>=1.10.2",
+    "numpy>=2.4.0",
+    "pytest-benchmark>=5.2.3",
+    "mkdocs>=1.6.1",
+]