API Reference
Complete reference for the Parallax runtime API
Overview
The Parallax runtime provides a simple C API for GPU-accelerated computing. The API is designed to be minimal and easy to integrate into existing C++ applications.
Header:
Library:
#include <parallax/runtime.h>Library:
-lparallax-runtime
Core Functions
Memory Management
C
parallax/runtime.h
void* parallax_umalloc(size_t size, unsigned flags);
void parallax_ufree(void* ptr);parallax_umalloc - Allocate unified memory accessible from both CPU and GPU
size: Size in bytes to allocateflags: Reserved for future use (pass 0)- Returns: Pointer to allocated memory, or NULL on failure
parallax_ufree - Free unified memory
ptr: Pointer previously returned by parallax_umalloc
Automatic Coherence: Memory allocated with
parallax_umalloc uses
software-defined unified memory with block-level dirty tracking. Coherence is maintained
automatically - no explicit synchronization needed!
Example Usage
C++
example.cpp
#include <iostream>
#include <vector>
#include <algorithm>
#include <execution>
int main() {
std::vector<float> data(1'000'000, 1.0f);
// GPU acceleration - automatic memory coherence
std::for_each(std::execution::par,
data.begin(), data.end(),
[](float& x) { x *= 2.0f; });
// Use results immediately - no sync needed!
std::cout << "Result: " << data[0] << std::endl;
return 0;
}