University:
Boston University
Course:
MET CS 201 | Introduction to Programming
Academic year:

2022
Views:

351

Pages:

17
Author:

Lafuentedqri

Easy, Effective, Efficient: GPU Programming in Python with PyOpenCL and PyCUDA Outline 1 PyCUDA 2 Automatic GPU Programming 3 GPU-DG: Challenges and Solutions Outline 1 PyCUDA 2 Automatic GPU Programming 3 GPU-DG: Challenges and Solutions Whetting your appetite 1 2 3 4 5 6 7 import pycuda.driver as cuda import pycuda.autoinit , pycuda.compiler import numpy a = numpy.random.randn(4,4).astype(numpy.float32) a gpu = cuda.mem alloc(a.nbytes) cuda.memcpy htod(a gpu, a) [This is examples/demo.py in the PyCUDA distribution.] Whetting your appetite 1 mod = pycuda.compiler.SourceModule(””” 2 global void twice( float ∗a) 3 { 4 int idx = threadIdx.x + threadIdx.y∗4; 5 a[ idx ] ∗= 2; 6 } 7 ”””) 8 9 func = mod.get function(”twice”) 10 func(a gpu, block=(4,4,1)) 11 12 a doubled = numpy.empty like(a) 13 cuda.memcpy dtoh(a doubled, a gpu) 14 print a doubled 15 print a Whetting your appetite 1 mod = pycuda.compiler.SourceModule(””” 2 global void twice( float ∗a) 3 { 4 int idx = threadIdx.x + threadIdx.y∗4; 5 a[ idx ] ∗= 2; 6 } 7 ”””) 8 9 func = mod.get function(”twice”) 10 func(a gpu, block=(4,4,1)) 11 12 a doubled = numpy.empty like(a) 13 cuda.memcpy dtoh(a doubled, a gpu) 14 print a doubled 15 print a Compute kernel Whetting your appetite, Part II Did somebody say “Abstraction is good”? Whetting your appetite, Part II 1 2 3 4 5 6 7 8 9 import numpy import pycuda.autoinit import pycuda.gpuarray as gpuarray a gpu = gpuarray.to gpu( numpy.random.randn(4,4).astype(numpy.float32)) a doubled = (2∗a gpu).get() print a doubled print a gpu gpuarray: Simple Linear Algebra pycuda.gpuarray: Meant to look and feel just like numpy. gpuarray.to gpu(numpy array) numpy array = gpuarray.get() +, -, ∗, /, fill, sin, exp, rand, basic indexing, norm, inner product, . . . Mixed types (int32 + float32 = float64) print gpuarray for debugging. Allows access to raw bits Use as kernel arguments, textures, etc. Sparse Matrix-Vector on the GPU New feature in 0.94: Sparse matrix-vector multiplication Uses “packeted format” by Garland and Bell (also includes parts of their code) Integrates with scipy.sparse. Conjugate-gradients solver included Deferred convergence checking PyOpenCL ↔ PyCUDA: A (rough) dictionary PyOpenCL Context CommandQueue Buffer Program Kernel Event (eg. enqueue marker) PyCUDA Context Stream mem alloc / DeviceAllocation SourceModule Function Event Scripting: Interpreted, not Compiled Program creation workflow: Edit Compile Link Run Scripting: Interpreted, not Compiled Program creation workflow: Edit Compile Link Run Scripting: Interpreted, not Compiled Program creation workflow: Edit Compile Link Run PyCUDA: Workflow Edit Cache? no Run nvcc SourceModule("...") Upload to GPU .cubin PyCUDA Run on GPU PyCUDA in the CUDA ecosystem C/C++ Python Runtime API PyCuda Driver API Kernel Driver Hardware CUDA has two Programming Interfaces: “Runtime” high-level (separate install) “Driver” low-level (libcuda.so, comes with GPU driver) PyCUDA: Vital Information http://mathema.tician.de/ software/pycuda Complete documentation X Consortium License (no warranty, free for all use) Convenient abstractions Array, Fast Vector Math, Reductions Requires: numpy, Python 2.4+ (Win/OS X/Linux)