from IST import *

I = imread('./lenna.jpg')
sample = zeros_like(I)
for i in arange(3):
    x, ys = ISTreal(I[:,:,i], its=100, p=0.5)
    # takes around 20 minutes on this machine
    I[:,:,i] = idwt2_full(x)
    sample[:,:,i] = ys

figure(figsize=(14,14))
subplot(121)
imshow(sample)
axis('off')
title('The sampled image')

subplot(122)
imshow(I)
axis('off')
title('The reconstructed image')

show()

%%capture string
import ISTpython as ISTp
#%timeit ISTp.IST()
# it takes a long time for this to finish

#print string.stdout
#python = float(string.stdout.split()[-4])
string = '1 loops, best of 3: 283 s per loop'
print string
python = float(string.split()[-4])

%%capture string
import IST as ISTnumpy
%timeit ISTnumpy.IST()

print string.stdout
numpy = float(string.stdout.split()[-4])

%%capture string
import ISTnumba
%timeit ISTnumba.IST()

print string.stdout
numba = float(string.stdout.split()[-4])

string = !matlab -nodisplay -nosplash -r "run IST.m; exit"
string = string[-1]
print string

matlab = float(string.split()[-2])

!gcc IST.c -o IST.o 
string = !time ./IST.o
c = float(string[1].split()[1][2:-1])
print "Total time in seconds: ", c

!gcc -O2 IST.c -o IST_O2.o 
string = !time ./IST.o
c_O2 = float(string[1].split()[1][2:-1])
print "Total time in seconds: ", c_O2

!gcc -O3 IST.c -o IST_O3.o
string = !time ./IST_O3.o
c_O3 = float(string[1].split()[1][2:-1])
print "Total time in seconds: ", c_O3

!llvm-gcc -O3 IST.c -o IST_llvm.o
string = !time ./IST_llvm.o
c_llvm = float(string[1].split()[1][2:-1])
print "Total time in seconds: ", c_llvm 

!mpicc -O3 ISt.c -o ISTparallel.o
string = !time ./ISTparallel.o
mpicc = float(string[1].split()[1][2:-1])
print "Total time in seconds: ", mpicc

from pylab import *

labels = ['python',  'numpy', 'numba', \
    'matlab', 'gcc', 'gcc --O2', \
    'gcc --O3', 'llvm --O3']
timings = [python, numpy, numba, matlab, c, c_O2, c_O3, c_llvm]
x = np.arange(len(labels))

# plotting from slowest to fastest
i = argsort(timings)
timings = sort(timings)[::-1]
labels = asarray(labels)
labels = labels[i][::-1]
fastest = timings[-1]

figure(figsize=(10,10))
matplotlib.rcParams.update({'font.size': 15})
ax = plt.axes(xticks=x, yscale='log')
ax.bar(x - 0.3, timings, width=0.6, alpha=0.4, bottom=1E-6)
ax.grid()
ax.set_xlim(-0.5, len(labels) - 0.5)
ax.set_ylim(1E-1, python*1.3)
ax.xaxis.set_major_formatter(plt.FuncFormatter(lambda i, loc: labels[int(i)]))
ax.set_ylabel('\\textrm{time (s)}')
ax.set_title("\\textrm{Iterative Soft Thresholding}")
show()


labels  = ['numpy', 'numba', 'matlab']
timings = [ numpy,   numba,   matlab]
x = np.arange(len(labels))

# plotting from max to min times
i = argsort(timings)
timings = sort(timings)[::-1]
labels = asarray(labels)
labels = labels[i][::-1]

ax = plt.axes(xticks=x)#, yscale='log')
ax.bar(x - 0.3, timings, width=0.6, alpha=0.4, bottom=1E-6)
ax.grid()
ax.set_xlim(-0.5, len(labels) - 0.5)
ax.set_ylim(0, 1.05*max(timings))
ax.xaxis.set_major_formatter(plt.FuncFormatter(lambda i, loc: labels[int(i)]))
ax.set_ylabel('\\textrm{time (s)}')
ax.set_title("\\textrm{Iterative Soft Thresholding}")
show()

print "C comparisions:"
print "NumPy / C  :", numpy  / fastest
print "Numba / C  :", numba  / fastest
print "Matlab / C :", matlab / fastest

print "\nThe comparision that we care about, NumPy and Matlab:"
print "Matlab / NumPy: ", matlab / numpy
print "Matlab / Numba: " , matlab / numba