import numpy as np
import pycuda.driver as cuda
import pycuda.autoinit
from pycuda.compiler import SourceModule

a = np.linspace(1,16,16).astype(np.float32)
a

a_gpu = cuda.mem_alloc(a.nbytes)

cuda.memcpy_htod(a_gpu, a)

mod = SourceModule("""
  __global__ void dubluj(float *a)
  {
    int idx = threadIdx.x;
    a[idx] *= 2;
    
  }
  """)

func = mod.get_function("dubluj")

func(a_gpu, block=(16,1,1))

print a
cuda.memcpy_dtoh(a, a_gpu)
print a

func(cuda.InOut(a), block=(4, 4, 1))
print a

import pycuda.gpuarray as gpuarray

a = np.linspace(1,16,16).astype(np.float32)
a_gpu = gpuarray.to_gpu(a.astype(numpy.float32))
a_doubled = (a_gpu*2).get()

print a_gpu
print a_doubled

import pycuda.driver as cuda
import pycuda.autoinit
from pycuda.compiler import SourceModule

mod = SourceModule("""
    #include <stdio.h>

    __global__ void printf_test()
    {
      printf("GONZO %d.%d.%d\\n", threadIdx.x, threadIdx.y, threadIdx.z);
    }
    """)

func = mod.get_function("printf_test")
func(block=(2,2,1))



import pycuda.driver as cuda
import pycuda.autoinit
from pycuda.compiler import SourceModule

mod = SourceModule("""
    __global__ void kernel(float *a)
    {
      int idx = threadIdx.x + blockDim.x*blockIdx.x;
    
      if(threadIdx.x==0)
          a[idx] = 1.0f;
    }
    """)

a = np.zeros(10).astype(np.float32)
func = mod.get_function("kernel")
print np.linspace(0,9,10)
print "----------------"
print a
func(cuda.InOut(a),block=(5,1,1),grid=(2,1,1))
print a