def fft2d(A):
"""
Compute the 2-dimensional discrete Fourier Transform
This function computes the *n*-dimensional discrete Fourier Transform
over any axes in an *M*-dimensional array by means of the
Fast Fourier Transform (FFT).
Parameters
----------
a : array_like
Input array, can be complex
Returns
-------
out : complex ndarray
"""
if not A.dist():
raise Exception("The array must be distributed")
for d in A.shape:
if (d % np.BLOCKSIZE != 0):
raise Exception("The array dimensions must be divisible "\
"with np.BLOCKSIZE(%d)"%np.BLOCKSIZE)
#Find an axis that is not distributed.
localaxis = -1
for i, p in enumerate(A.pgrid()):
if p == 1:
localaxis = i
break
if localaxis == -1:
raise Exception("One dimension in the process grid must not "
"be distributed.")
#Convert to a complex array
B = np.empty(A.shape, dtype=np.complex, dist=True)
B[:] = A
## 1-D FFT on the X dimension
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A, axis=localaxis) ## local FFTs
## Transpose
A = pyHPC.transpose(A)
## 1-D FFt on the Y dimension
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A, axis=localaxis) ## local FFTs
## Transpose
A = pyHPC.transpose(A)
return A
def fft2d(A):
"""
Compute the 2-dimensional discrete Fourier Transform
This function computes the *n*-dimensional discrete Fourier Transform
over any axes in an *M*-dimensional array by means of the
Fast Fourier Transform (FFT).
Parameters
----------
a : array_like
Input array, can be complex
Returns
-------
out : complex ndarray
"""
if not A.dist():
raise Exception("The array must be distributed")
for d in A.shape:
if(d % np.BLOCKSIZE != 0):
raise Exception("The array dimensions must be divisible "\
"with np.BLOCKSIZE(%d)"%np.BLOCKSIZE)
#Find an axis that is not distributed.
localaxis = -1
for i,p in enumerate(A.pgrid()):
if p == 1:
localaxis = i
break
if localaxis == -1:
raise Exception("One dimension in the process grid must not "
"be distributed.")
#Convert to a complex array
B = np.empty(A.shape, dtype=np.complex, dist=True)
B[:] = A
## 1-D FFT on the X dimension
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A,axis=localaxis) ## local FFTs
## Transpose
A = pyHPC.transpose(A)
## 1-D FFt on the Y dimension
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A,axis=localaxis) ## local FFTs
## Transpose
A = pyHPC.transpose(A)
return A
def run():
if not np.SPMD_MODE:
print "[rank %d] Warning - ignored in non-SPMD mode\n" % (np.RANK),
return
try: #This test requires the pyHPC module
import pyHPC
except:
print "[rank %d] Warning - ignored pyHPC not found\n" % (np.RANK),
return
if np.BLOCKSIZE > 10:
print "[rank %d] Warning - ignored np.BLOCKSIZE too high\n" % (
np.RANK),
return
niter = 5
for m in range(np.BLOCKSIZE, niter * np.BLOCKSIZE, np.BLOCKSIZE):
for n in range(np.BLOCKSIZE, niter * np.BLOCKSIZE, np.BLOCKSIZE):
for k in range(np.BLOCKSIZE, niter * np.BLOCKSIZE, np.BLOCKSIZE):
for axis in permutations(xrange(3)):
Asrc = dnumpytest.random_list([m, n, k])
Ad = np.array(Asrc, dtype=float, dist=True)
Af = np.array(Asrc, dtype=float, dist=False)
Bd = pyHPC.transpose(Ad, axis)
Bf = np.transpose(Af, axis)
if not dnumpytest.array_equal(Bd, Bf):
raise Exception("Uncorrect result matrix\n")
def run():
if not np.SPMD_MODE:
print "[rank %d] Warning - ignored in non-SPMD mode\n"%(np.RANK),
return
try:#This test requires the pyHPC module
import pyHPC
except:
print "[rank %d] Warning - ignored pyHPC not found\n"%(np.RANK),
return
if np.BLOCKSIZE > 10:
print "[rank %d] Warning - ignored np.BLOCKSIZE too high\n"%(np.RANK),
return
niter = 5
for m in range(np.BLOCKSIZE,niter*np.BLOCKSIZE, np.BLOCKSIZE):
for n in range(np.BLOCKSIZE,niter*np.BLOCKSIZE, np.BLOCKSIZE):
for k in range(np.BLOCKSIZE,niter*np.BLOCKSIZE, np.BLOCKSIZE):
for axis in permutations(xrange(3)):
Asrc = dnumpytest.random_list([m,n,k])
Ad = np.array(Asrc, dtype=float, dist=True)
Af = np.array(Asrc, dtype=float, dist=False)
Bd = pyHPC.transpose(Ad, axis)
Bf = np.transpose(Af, axis)
if not dnumpytest.array_equal(Bd,Bf):
raise Exception("Uncorrect result matrix\n")
def fft3d(A):
"""
Compute the 3-dimensional discrete Fourier Transform
This function computes the *n*-dimensional discrete Fourier Transform
over any axes in an *M*-dimensional array by means of the
Fast Fourier Transform (FFT).
Parameters
----------
a : array_like
Input array, can be complex
Returns
-------
out : complex ndarray
"""
if not A.dist():
raise Exception("The array must be distributed")
for d in A.shape:
if (d % np.BLOCKSIZE != 0):
raise Exception("The array dimensions must be divisible "\
"with np.BLOCKSIZE(%d)"%np.BLOCKSIZE)
#Find an axis that is not distributed.
localaxis = 0
if A.pgrid()[0] != 1:
raise Exception("The first dimension in the process grid must not "
"be distributed.")
#Convert to a complex array
B = np.empty(A.shape, dtype=np.complex, dist=True)
B[:] = A
## 1-D FFT on data along the X dimension, for which the data
## should be local and contiguous
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A, axis=localaxis) ## local FFTs
## Transpose the X,Y planes: X'=Y, Y'=X, Z'=Z, (Y,X,Z)
A = pyHPC.transpose(A, (1, 0, 2))
## 1-D FFT on data along the Y dimension (now is X', for which the
## data should be local and contiguous)
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A, axis=localaxis) ## local FFTs
## Transpose the X',Z' planes: X''=Z'=Z, Y''=Y'=X, Z''=X'=Y, (Z,X,Y)
A = pyHPC.transpose(A, (2, 1, 0))
## 1-D FFt on data along the Z dimension (now is X'', for which
## data should be local and contiguous)
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A, axis=localaxis) ## local FFTs
## Transpose the X',Z' planes: X''=Z'=Z, Y''=Y'=X, Z''=X'=Y, (Z,X,Y)
A = pyHPC.transpose(A, (1, 2, 0))
return A
def fft3d(A):
"""
Compute the 3-dimensional discrete Fourier Transform
This function computes the *n*-dimensional discrete Fourier Transform
over any axes in an *M*-dimensional array by means of the
Fast Fourier Transform (FFT).
Parameters
----------
a : array_like
Input array, can be complex
Returns
-------
out : complex ndarray
"""
if not A.dist():
raise Exception("The array must be distributed")
for d in A.shape:
if(d % np.BLOCKSIZE != 0):
raise Exception("The array dimensions must be divisible "\
"with np.BLOCKSIZE(%d)"%np.BLOCKSIZE)
#Find an axis that is not distributed.
localaxis=0
if A.pgrid()[0] != 1:
raise Exception("The first dimension in the process grid must not "
"be distributed.")
#Convert to a complex array
B = np.empty(A.shape, dtype=np.complex, dist=True)
B[:] = A
## 1-D FFT on data along the X dimension, for which the data
## should be local and contiguous
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A,axis=localaxis) ## local FFTs
## Transpose the X,Y planes: X'=Y, Y'=X, Z'=Z, (Y,X,Z)
A = pyHPC.transpose(A,(1,0,2))
## 1-D FFT on data along the Y dimension (now is X', for which the
## data should be local and contiguous)
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A,axis=localaxis) ## local FFTs
## Transpose the X',Z' planes: X''=Z'=Z, Y''=Y'=X, Z''=X'=Y, (Z,X,Y)
A = pyHPC.transpose(A,(2,1,0))
## 1-D FFt on data along the Z dimension (now is X'', for which
## data should be local and contiguous)
l_A = A.local()
if len(l_A) > 0:
l_A[:] = np.fft.fft(l_A,axis=localaxis) ## local FFTs
## Transpose the X',Z' planes: X''=Z'=Z, Y''=Y'=X, Z''=X'=Y, (Z,X,Y)
A = pyHPC.transpose(A,(1,2,0))
return A
