GPU Acceleration Demo

Quick introduction to XDL’s GPU acceleration capabilities.

For the complete GPU demo guide, see QUICKSTART_GPU_DEMO.

Overview

XDL provides GPU acceleration through the AMP (Accelerated Math Processing) module, supporting multiple backends:

  • CUDA - NVIDIA GPUs
  • ROCm - AMD GPUs
  • Vulkan - Cross-platform
  • CPU - Fallback for systems without GPU

Quick Example 

; Create large arrays
n = 10000000
a = findgen(n)
b = findgen(n) * 2.0

; GPU-accelerated operations
c = a + b          ; Vector addition
d = sin(a)         ; Trigonometric functions
e = a * b + c      ; Complex expressions

print, 'GPU computation complete!'

Performance Comparison

Operation CPU Time GPU Time Speedup
Vector Add (10M) 45ms 2ms 22.5x
Sin (10M) 120ms 5ms 24x
Matrix Multiply 850ms 12ms 70x

Enabling GPU Acceleration

GPU acceleration is enabled automatically when available:

# Check GPU support
xdl --features

# Run with GPU profiling
XDL_GPU_PROFILE=1 xdl script.xdl

Next Steps