Platforms & Flavors Overview

The numerous outputs of the Hybridizer are called Flavors. Flavors define the language or jargon in which the Hybridizer outputs code. This output is then compiled using the appropriate compiler to generate machine code for the target architecture.

Available Flavors

Flavor	Target	Compiler	Best For
CUDA	NVIDIA GPUs	nvcc	Massive parallel throughput
OMP	Multi-core CPUs	gcc/clang with OpenMP	Portability, debugging
AVX	Intel/AMD CPUs	gcc/clang	High-perf CPU SIMD
AVX512	Intel Xeon/HPC	gcc/clang	Wide vector operations
NEON	ARM processors	gcc/clang	Embedded, mobile
POWER	IBM POWER	xlc/gcc	HPC systems

Choosing a Flavor

Flavor Details

CUDA Flavor

For NVIDIA GPUs. Generates .cu files with __device__ and __global__ attributes. Best for massive parallel throughput.

→ CUDA Backend Details

OMP Flavor

Plain C/C++ with OpenMP support. Useful for:

• Testing: Disambiguating issues between input code and flavor source-code generation
• Portability: Sequential fallback if compiled without OpenMP
• Performance: Optimizing compilers may generate better machine code than the JIT

→ OMP Backend Details

Vector Flavors (AVX/NEON/POWER)

C++ code using a micro-library for vectorization. The implementation is specific to each hardware architecture, but the interface is common.

→ Vector Backends Details

Next Steps

• CUDA Backend — Deep dive into GPU targeting
• OMP Backend — CPU multi-threading
• Vector Backends — SIMD on CPU