v2.1.0

Release

Updated VM and implement more thread safe methods along with some optimization.

Full Changelog: v2.0.0...v2.1.0

v2.0.0

Release

Release Note

This major release transforms HyperGL into a high-performance graphics runtime by introducing the HyperGL VM.

• New Paradigm: Added CommandBuffer, a C-side bytecode VM that replays rendering and compute logic at the limit of the CPU cache.
• Programmable ISA: The VM is now Turing-complete, supporting 8 registers, ALU operations, jumps (GOTO), and modular subroutines (CALL/RET).
• GPU-Driven Branching: Added skip_if_zero and skip_reg_* for culling and logic entirely within the C-loop.
• Binary Scenes: Introduced the .hgb format—serialize your command buffers to disk and load them with zero Python overhead.
• High-Level Compiler: Added SceneCompiler and @subroutine to make VM programming feel like native Python.
• Zero-GIL: Command Buffer submission is fully thread-safe and releases the GIL, enabling massive scaling on Python 3.13t/3.14t.
• Backward Compatibility: All v1.0.x immediate mode methods (pipe.render(), etc.) remain fully supported for prototyping and UI.

Known Limitations

• MacOS Support: OpenGL is deprecated to 4.1.

Full Changelog: v1.0.1...v2.0.0

v1.0.1

Release

Release Note

Some bug fixes and optimization in begin/end frames of the context

And some cleanup to satisfy the linter.

Tested Environments

• OS: Windows 10, Windows 11
• GPU: NVIDIA RTX 3050, RTX 2060, GT 730
• OpenGL Version: 4.5+ Core Profile

Known Limitations

• MacOS Support: OpenGL is deprecated to 4.1.

Full Changelog: v1.0.0...v1.0.1

v1.0.0

Release

Release Note

Core API and lifetime semantics are considered stable.

Future changes will be additive or clearly versioned.

Tested Environments

• OS: Windows 10
• GPU: NVIDIA RTX 3050, RTX 2060
• OpenGL Version: 4.5+ Core Profile

Known Limitations

• MacOS Support: OpenGL is deprecated to 4.1.

Full Changelog: v1.0.0-beta.5...v1.0.0

v1.0.0-beta.5

Beta Release

Fix

HyperGL v1.0.0-beta.5 introduces... fixes. And comment banners. And some basic CI/CD.

But more important, partial MacOS support (it compiles!)

Tested Environments

• OS: Windows 10
• GPU: NVIDIA RTX 3050, RTX 2060
• OpenGL Version: 4.5+ Core Profile

Known Limitations

• MacOS Support: OpenGL is deprecated to 4.1. Tough luck.

Full Changelog: v1.0.0-beta.4...v1.0.0-beta.5

v1.0.0-beta.4

Beta Release

AZDO capability and headless support

HyperGL v1.0.0-beta.4 introduces AZDO (Approaching Zero Driver Overhead) helpers. And headless support.

HyperGL can now train Reinforcement Learning agents at 12,000+ FPS with zero-copy NumPy readback and render scenes with 100,000+ unique textures in a single Python draw call.

Tested Environments

• OS: Windows 10
• GPU: NVIDIA RTX 3050, RTX 2060
• OpenGL Version: 4.5+ Core Profile

Key Features

• ctx.pack_indirect([...]): High-performance C-level packing of indirect draw commands (no more manual struct.pack).
• buffer.write_texture_handle(offset, image) Automatically retrieves and writes resident 64-bit handles to GPU memory.
• Headless Mode: HyperGL is now a first-class citizen on servers and CI/CD pipelines.

Known Limitations

• MacOS Support: unsupported, probably forever.
• OpenGL: 4.5+ is required for the advanced features this release has to offer. OpenGL 4.6 is explicitly requested and will fallback to older versions if unsupported.

v1.1.0-beta1

Beta Release

AZDO capability

HyperGL v1.1 introduces AZDO (Approaching Zero Driver Overhead) features, effectively removing the CPU bottleneck for Python rendering.

Tested Environments

• OS: Windows 10
• GPU: NVIDIA RTX 3050, RTX 2060
• OpenGL Version: 4.5+ Core Profile

Key Features

• Bindless Textures (GL_ARB_bindless_texture): Pass 64-bit texture handles directly to shaders via SSBOs. Access thousands of unique textures in a single draw call.
• Indirect Drawing (GL_ARB_multi_draw_indirect): Generate draw commands on the GPU or store them in buffers. Issue a single Python call to render millions of instances.
• Free-Threading Support (Python 3.14t): Still fully compatible with Python's experimental free-threaded (No-GIL) build, allowing render loops and game logic to run in true parallel. All according to PEP 703.

Known Limitations

• MacOS Support: unsupported, probably forever.
• OpenGL: 4.5+ is required for the advanced features this release has to offer.

v1.0.1-beta.1

Beta Release

This is the first public preview of HyperGL, a Pythonic wrapper for OpenGL. The goal of this project is to bridge the features of ModernGL and ZenGL while maintaining high performance, plus supporting the latest cutting edge versions of Python(3.13/3.14) and free threading.

Tested Environments

• OS: Windows 10
• GPU: NVIDIA RTX 3050, RTX 2060
• OpenGL Version: 4.5+ Core Profile

Key Features

• SSBOs and persistent mapping: Fixed. You just map in the main thread.

Known Limitations

• MacOS Support: unsupported, probably forever.

v1.0.0-beta.1

Beta Release

This is the first public preview of HyperGL, a Pythonic wrapper for OpenGL. The goal of this project is to bridge the features of ModernGL and ZenGL while maintaining high performance, plus supporting the latest cutting edge versions of Python(3.13/3.14) and free threading.

Tested Environments(so far)

As this is a one-man job, I have limited hardware to test on. This beta has been confirmed to work on:

• OS: Windows 10
• GPU: NVIDIA RTX 3050, RTX 2060
• OpenGL Version: 4.5+ Core Profile

Key Features

• Abstraction: No more manual glGenBuffers or glBufferData—handled via the Buffer class.
• Context Management: Integrated with [GLFW/PySDL2/PyGame/etc.] for easy window setup.
• Type Hinting: Fully typed for better IDE autocomplete support.

Known Limitations

• MacOS Support: Currently untested, and it wouldn't be prioritized. At all. As a matter of fact, probably never.