Centre Blaise Pascal » Bench4GPU

#!/usr/bin/env python

#

# Splutter-by-MonteCarlo using PyCUDA/PyOpenCL

#

# CC BY-NC-SA 2014 : <emmanuel.quemener@ens-lyon.fr> 

#

# Thanks to Andreas Klockner for PyCUDA:

# http://mathema.tician.de/software/pycuda

# 

# 2013-01-01 : problems with launch timeout

# http://stackoverflow.com/questions/497685/how-do-you-get-around-the-maximum-cuda-run-time

# Option "Interactive" "0" in /etc/X11/xorg.conf

# Common tools

import numpy

from numpy.random import randint as nprnd

import sys

import getopt

import time

import math

from socket import gethostname

# find prime factors of a number

# Get for WWW :

# http://pythonism.wordpress.com/2008/05/17/looking-at-factorisation-in-python/

def PrimeFactors(x):

  factorlist=numpy.array([]).astype('uint32')

  loop=2

  while loop<=x:

    if x%loop==0:

      x/=loop

      factorlist=numpy.append(factorlist,[loop])

    else:

      loop+=1

  return factorlist

# Try to find the best thread number in Hybrid approach (Blocks&Threads)

# output is thread number

def BestThreadsNumber(jobs):

  factors=PrimeFactors(jobs)

  matrix=numpy.append([factors],[factors[::-1]],axis=0)

  threads=1

  for factor in matrix.transpose().ravel():

    threads=threads*factor

    if threads*threads>jobs:

      break

  return(long(threads))

# Predicted Amdahl Law (Reduced with s=1-p)  

def AmdahlR(N, T1, p):

  return (T1*(1-p+p/N))

# Predicted Amdahl Law

def Amdahl(N, T1, s, p):

  return (T1*(s+p/N))

# Predicted Mylq Law with first order

def Mylq(N, T1,s,c,p):

  return (T1*(s+p/N)+c*N)

# Predicted Mylq Law with second order

def Mylq2(N, T1,s,c1,c2,p):

  return (T1*(s+p/N)+c1*N+c2*N*N)

KERNEL_CODE_CUDA="""

// Marsaglia RNG very simple implementation

#define znew  ((z=36969*(z&65535)+(z>>16))<<16)

#define wnew  ((w=18000*(w&65535)+(w>>16))&65535)

#define MWC   (znew+wnew)

#define SHR3  (jsr=(jsr=(jsr=jsr^(jsr<<17))^(jsr>>13))^(jsr<<5))

#define CONG  (jcong=69069*jcong+1234567)

#define KISS  ((MWC^CONG)+SHR3)

#define MWCfp MWC * 2.328306435454494e-10f

#define KISSfp KISS * 2.328306435454494e-10f

#define MAX 4294967296

uint rotl(uint value, int shift) {

   return (value << shift) | (value >> (sizeof(value) * 8 - shift));

}

uint rotr(uint value, int shift) {

   return (value >> shift) | (value << (sizeof(value) * 8 - shift));

}

__global__ void MainLoopBlocks(uint *s,uint size,ulong iterations,uint seed_w,uint seed_z)

{

   // uint z=rotl(seed_z,blockIdx.x);

   // uint w=rotr(seed_w,blockIdx.x);

   // uint jsr=rotl(seed_z,blockIdx.x);

   // uint jcong=rotr(seed_w,blockIdx.x);

   uint z=seed_z/(blockIdx.x+1);

   uint w=seed_w%(blockIdx.x+1);

   uint jsr=seed_z/(blockIdx.x+1);

   uint jcong=seed_w%(blockIdx.x+1);

   for (ulong i=0;i<iterations;i++) {

      s[(uint)(((ulong)size*(ulong)CONG)/(ulong)MAX)]+=1;

   }

   __threadfence_block();

}

__global__ void MainLoopThreads(uint *s,uint size,ulong iterations,uint seed_w,uint seed_z)

{ 

   // uint z=rotl(seed_z,threadIdx.x);

   // uint w=rotr(seed_w,threadIdx.x);

   // uint jsr=rotl(seed_z,threadIdx.x);

   // uint jcong=rotr(seed_w,threadIdx.x);

   uint z=seed_z;

   uint w=seed_w;

   uint jsr=seed_z;

   uint jcong=seed_w;

   for (ulong i=0;i<iterations;i++) {

      s[(uint)(((ulong)size*(ulong)CONG)/(ulong)MAX)]+=1;

   }

   __syncthreads();

}

__global__ void MainLoopHybrid(uint *s,uint size,ulong iterations,uint seed_w,uint seed_z)

{

   uint z=seed_z;

   uint w=seed_w;

   uint jsr=seed_z;

   uint jcong=seed_w;

   for (ulong i=0;i<iterations;i++) {

      s[(uint)(((ulong)size*(ulong)CONG)/(ulong)MAX)]+=1;

   }

   __syncthreads();

}

"""

KERNEL_CODE_OPENCL="""

// Marsaglia RNG very simple implementation

#define znew  ((z=36969*(z&65535)+(z>>16))<<16)

#define wnew  ((w=18000*(w&65535)+(w>>16))&65535)

#define MWC   (znew+wnew)

#define SHR3  (jsr=(jsr=(jsr=jsr^(jsr<<17))^(jsr>>13))^(jsr<<5))

#define CONG  (jcong=69069*jcong+1234567)

#define KISS  ((MWC^CONG)+SHR3)

#define CONGfp CONG * 2.328306435454494e-10f

#define SHR3fp SHR3 * 2.328306435454494e-10f

#define MWCfp MWC * 2.328306435454494e-10f

#define KISSfp KISS * 2.328306435454494e-10f

#define MAX (ulong)4294967296

uint rotl(uint value, int shift) {

    return (value << shift) | (value >> (sizeof(value) * CHAR_BIT - shift));

}

uint rotr(uint value, int shift) {

    return (value >> shift) | (value << (sizeof(value) * CHAR_BIT - shift));

}

__kernel void MainLoopGlobal(__global uint *s,const uint space,const ulong iterations,const uint seed_w,const uint seed_z)

{

   //__private const float id=(float)get_global_id(0);

   //__private const float size=(float)get_global_size(0);

   //__private const float block=space/size;

   __private const ulong id=(ulong)get_global_id(0);

   __private const ulong size=(ulong)get_global_size(0);

   __private const ulong block=(ulong)space/(ulong)size;

   __private uint z=seed_z-(uint)id;

   __private uint w=seed_w+(uint)id;

   __private uint jsr=seed_z;

   __private uint jcong=seed_w;

   for (__private ulong i=0;i<iterations;i++) {

      // Standard version does not work for several processes (some lost!)

      //__private uint position=(uint)(((ulong)space*(ulong)MWC)/(ulong)MAX);

      // Dense version

      //__private uint position=(uint)( ((ulong)CONG+MAX*(ulong)id)/(ulong)size*(ulong)space/(ulong)MAX );

      __private size_t position=(size_t)( ((ulong)MWC+(ulong)MAX*(ulong)id)*(ulong)block/(ulong)MAX );

      // Float version seems to be the best...

      //__private uint position=(uint)( block*(CONGfp+id) );

      s[position]++;

   }

   barrier(CLK_LOCAL_MEM_FENCE | CLK_GLOBAL_MEM_FENCE);

}

__kernel void MainLoopLocal(__global uint *s,uint size,ulong iterations,uint seed_w,uint seed_z)

{

   uint z=rotl(seed_z,get_local_id(0));

   uint w=rotr(seed_w,get_local_id(0));

   uint jsr=rotl(seed_z,get_local_id(0));

   uint jcong=rotr(seed_w,get_local_id(0));

   for (ulong i=0;i<iterations;i++) {

      s[(int)(((ulong)size*(ulong)CONG)/(ulong)MAX)]+=(uint)1;

   }

   barrier(CLK_LOCAL_MEM_FENCE | CLK_GLOBAL_MEM_FENCE);

}

__kernel void MainLoopHybrid(__global uint *s,uint size,ulong iterations,uint seed_w,uint seed_z)

{

   uint z=rotl(seed_z,get_group_id(0)*get_num_groups(0)+get_local_id(0));

   uint w=rotr(seed_w,get_group_id(0)*get_num_groups(0)+get_local_id(0));

   uint jsr=rotl(seed_z,get_group_id(0)*get_num_groups(0)+get_local_id(0));

   uint jcong=rotr(seed_w,get_group_id(0)*get_num_groups(0)+get_local_id(0));

   for (ulong i=0;i<iterations;i++) {

      s[(int)(((ulong)size*(ulong)CONG)/(ulong)MAX)]+=1;

      barrier(CLK_LOCAL_MEM_FENCE | CLK_GLOBAL_MEM_FENCE);

   }

}

"""

def MetropolisCuda(circle,iterations,steps,jobs,ParaStyle):

  # Avec PyCUDA autoinit, rien a faire !

  circleCU = cuda.InOut(circle)

  mod = SourceModule(KERNEL_CODE_CUDA)

  MetropolisBlocksCU=mod.get_function("MainLoopBlocks")

  MetropolisJobsCU=mod.get_function("MainLoopThreads")

  MetropolisHybridCU=mod.get_function("MainLoopHybrid")

  start = pycuda.driver.Event()

  stop = pycuda.driver.Event()

  MySplutter=numpy.zeros(steps)

  MyDuration=numpy.zeros(steps)

  if iterations%jobs==0:

    iterationsCL=numpy.uint64(iterations/jobs)

  else:

    iterationsCL=numpy.uint64(iterations/jobs+1)

  iterationsNew=iterationsCL*jobs

  for i in range(steps):

    Splutter=numpy.zeros(1024).astype(numpy.uint32)

    print Splutter

    SplutterCU = cuda.InOut(Splutter)

    start.record()

    start.synchronize()

    if ParaStyle=='Blocks':

      MetropolisBlocksCU(SplutterCU,

                         numpy.uint32(len(Splutter)),

                         numpy.uint64(iterationsCL),

                         numpy.uint32(nprnd(2**30/jobs)),

                         numpy.uint32(nprnd(2**30/jobs)),

                         grid=(jobs,1),

                         block=(1,1,1))

      print "%s with (WorkItems/Threads)=(%i,%i) %s method done" % \

            (Alu,jobs,1,ParaStyle)      

    elif ParaStyle=='Hybrid':

      threads=BestThreadsNumber(jobs)

      MetropolisHybridCU(SplutterCU,

                         numpy.uint32(len(Splutter)),

                         numpy.uint64(iterationsCL),

                         numpy.uint32(nprnd(2**30/jobs)),

                         numpy.uint32(nprnd(2**30/jobs)),

                         grid=(jobs,1),

                         block=(threads,1,1))

      print "%s with (WorkItems/Threads)=(%i,%i) %s method done" % \

            (Alu,jobs/threads,threads,ParaStyle)

    else:

      MetropolisJobsCU(SplutterCU,

                       numpy.uint32(len(Splutter)),

                       numpy.uint64(iterationsCL),

                       numpy.uint32(nprnd(2**30/jobs)),

                       numpy.uint32(nprnd(2**30/jobs)),

                       grid=(1,1),

                       block=(jobs,1,1))

      print "%s with (WorkItems/Threads)=(%i,%i) %s method done" % \

            (Alu,jobs,1,ParaStyle)

    stop.record()

    stop.synchronize()

    elapsed = start.time_till(stop)*1e-3

    print Splutter,sum(Splutter)

    MySplutter[i]=numpy.median(Splutter)

    print numpy.mean(Splutter),MySplutter[i],numpy.std(Splutter)

    MyDuration[i]=elapsed

    #AllPi=4./numpy.float32(iterationsCL)*circle.astype(numpy.float32)

    #MyPi[i]=numpy.median(AllPi)

    #print MyPi[i],numpy.std(AllPi),MyDuration[i]

  print jobs,numpy.mean(MyDuration),numpy.median(MyDuration),numpy.std(MyDuration)

  return(numpy.mean(MyDuration),numpy.median(MyDuration),numpy.std(MyDuration))

def MetropolisOpenCL(circle,iterations,steps,jobs,ParaStyle,Alu,Device):

  # Initialisation des variables en les CASTant correctement

  MaxMemoryXPU=0

  MinMemoryXPU=0

  if Device==0:

    print "Enter XPU selector based on ALU type: first selected"

    HasXPU=False

    # Default Device selection based on ALU Type

    for platform in cl.get_platforms():

      for device in platform.get_devices():

        deviceType=cl.device_type.to_string(device.type)

        deviceMemory=device.max_mem_alloc_size

        if deviceMemory>MaxMemoryXPU:

          MaxMemoryXPU=deviceMemory

        if deviceMemory<MinMemoryXPU or MinMemoryXPU==0:

          MinMemoryXPU=deviceMemory

        if deviceType=="GPU" and Alu=="GPU" and not HasXPU:

          XPU=device

          print "GPU selected with Allocable Memory %i: %s" % (deviceMemory,device.name)

          HasXPU=True

          MemoryXPU=deviceMemory

        if deviceType=="CPU" and Alu=="CPU" and not HasXPU:        

          XPU=device

          print "CPU selected with Allocable Memory %i: %s" % (deviceMemory,device.name)

          HasXPU=True

          MemoryXPU=deviceMemory

  else:

    print "Enter XPU selector based on device number & ALU type"

    Id=1

    HasXPU=False

    # Primary Device selection based on Device Id

    for platform in cl.get_platforms():

      for device in platform.get_devices():

        deviceType=cl.device_type.to_string(device.type)

        deviceMemory=device.max_mem_alloc_size

        if deviceMemory>MaxMemoryXPU:

          MaxMemoryXPU=deviceMemory

        if deviceMemory<MinMemoryXPU or MinMemoryXPU==0:

          MinMemoryXPU=deviceMemory

        if Id==Device and Alu==deviceType and HasXPU==False:

          XPU=device

          print "CPU/GPU selected with Allocable Memory %i: %s" % (deviceMemory,device.name)

          HasXPU=True

          MemoryXPU=deviceMemory

        Id=Id+1

    if HasXPU==False:

      print "No XPU #%i of type %s found in all of %i devices, sorry..." % \

          (Device,Alu,Id-1)

      return(0,0,0)

  print "Allocable Memory is %i, between %i and %i " % (MemoryXPU,MinMemoryXPU,MaxMemoryXPU)

  # Je cree le contexte et la queue pour son execution

  ctx = cl.Context([XPU])

  queue = cl.CommandQueue(ctx,

                          properties=cl.command_queue_properties.PROFILING_ENABLE)

  # Je recupere les flag possibles pour les buffers

  mf = cl.mem_flags

  MetropolisCL = cl.Program(ctx,KERNEL_CODE_OPENCL).build(options = "-cl-mad-enable -cl-fast-relaxed-math")

  MyDuration=numpy.zeros(steps)

  if iterations%jobs==0:

    iterationsCL=numpy.uint64(iterations/jobs)

  else:

    iterationsCL=numpy.uint64(iterations/jobs+1)

  iterationsNew=numpy.uint64(iterationsCL*jobs)

  MySplutter=numpy.zeros(steps)

  MaxWorks=2**(int)(numpy.log2(MinMemoryXPU/4))

  print MaxWorks,2**(int)(numpy.log2(MemoryXPU))

  Splutter=numpy.zeros((MaxWorks/jobs)*jobs).astype(numpy.uint32)

  for i in range(steps):

    #Splutter=numpy.zeros(2**(int)(numpy.log2(MemoryXPU/4))).astype(numpy.uint32)

    #Splutter=numpy.zeros(1024).astype(numpy.uint32)

    #Splutter=numpy.zeros(jobs).astype(numpy.uint32)

    print Splutter,len(Splutter)

    Splutter[:]=0

    SplutterCL = cl.Buffer(ctx, mf.WRITE_ONLY|mf.COPY_HOST_PTR,hostbuf=Splutter)

    if ParaStyle=='Blocks':

      # Call OpenCL kernel

      # (1,) is Global work size (only 1 work size)

      # (1,) is local work size

      # circleCL is lattice translated in CL format

      # SeedZCL is lattice translated in CL format

      # SeedWCL is lattice translated in CL format

      # step is number of iterations

      # CLLaunch=MetropolisCL.MainLoopGlobal(queue,(jobs,),None,

      #                                      SplutterCL,

      #                                      numpy.uint32(len(Splutter)),

      #                                      numpy.uint64(iterationsCL),

      #                                      numpy.uint32(nprnd(2**30/jobs)),

      #                                      numpy.uint32(nprnd(2**30/jobs)))

      CLLaunch=MetropolisCL.MainLoopGlobal(queue,(jobs,),None,

                                           SplutterCL,

                                           numpy.uint32(len(Splutter)),

                                           numpy.uint64(iterationsCL),

                                           numpy.uint32(521288629),

                                           numpy.uint32(362436069))

      print "%s with (WorkItems/Threads)=(%i,%i) %s method done" % \

            (Alu,jobs,1,ParaStyle)

    elif ParaStyle=='Hybrid':

      threads=BestThreadsNumber(jobs)

      # en OpenCL, necessaire de mettre un Global_id identique au local_id

      CLLaunch=MetropolisCL.MainLoopHybrid(queue,(jobs,),(threads,),

                                           SplutterCL,

                                           numpy.uint32(len(Splutter)),

                                           numpy.uint64(iterationsCL),

                                           numpy.uint32(nprnd(2**30/jobs)),

                                           numpy.uint32(nprnd(2**30/jobs)))

      print "%s with (WorkItems/Threads)=(%i,%i) %s method done" % \

            (Alu,jobs/threads,threads,ParaStyle)

    else:

      # en OpenCL, necessaire de mettre un Global_id identique au local_id

      CLLaunch=MetropolisCL.MainLoopLocal(queue,(jobs,),(jobs,),

                                          SplutterCL,

                                          numpy.uint32(len(Splutter)),

                                          numpy.uint64(iterationsCL),

                                          numpy.uint32(nprnd(2**30/jobs)),

                                          numpy.uint32(nprnd(2**30/jobs)))

      print "%s with %i %s done" % (Alu,jobs,ParaStyle)

    CLLaunch.wait()

    cl.enqueue_copy(queue, Splutter, SplutterCL).wait()

    elapsed = 1e-9*(CLLaunch.profile.end - CLLaunch.profile.start)

    MyDuration[i]=elapsed

    #print Splutter,sum(Splutter)

    #MySplutter[i]=numpy.median(Splutter)

    #print numpy.mean(Splutter)*len(Splutter),MySplutter[i]*len(Splutter),numpy.std(Splutter)

    SplutterCL.release()

  print jobs,numpy.mean(MyDuration),numpy.median(MyDuration),numpy.std(MyDuration)

  return(numpy.mean(MyDuration),numpy.median(MyDuration),numpy.std(MyDuration))

def FitAndPrint(N,D,Curves):

  from scipy.optimize import curve_fit

  import matplotlib.pyplot as plt

  try:

    coeffs_Amdahl, matcov_Amdahl = curve_fit(Amdahl, N, D)

    D_Amdahl=Amdahl(N,coeffs_Amdahl[0],coeffs_Amdahl[1],coeffs_Amdahl[2])

    coeffs_Amdahl[1]=coeffs_Amdahl[1]*coeffs_Amdahl[0]/D[0]

    coeffs_Amdahl[2]=coeffs_Amdahl[2]*coeffs_Amdahl[0]/D[0]

    coeffs_Amdahl[0]=D[0]

    print "Amdahl Normalized: T=%.2f(%.6f+%.6f/N)" % \

        (coeffs_Amdahl[0],coeffs_Amdahl[1],coeffs_Amdahl[2])

  except:

    print "Impossible to fit for Amdahl law : only %i elements" % len(D) 

  try:

    coeffs_AmdahlR, matcov_AmdahlR = curve_fit(AmdahlR, N, D)

    D_AmdahlR=AmdahlR(N,coeffs_AmdahlR[0],coeffs_AmdahlR[1])

    coeffs_AmdahlR[1]=coeffs_AmdahlR[1]*coeffs_AmdahlR[0]/D[0]

    coeffs_AmdahlR[0]=D[0]

    print "Amdahl Reduced Normalized: T=%.2f(%.6f+%.6f/N)" % \

        (coeffs_AmdahlR[0],1-coeffs_AmdahlR[1],coeffs_AmdahlR[1])

  except:

    print "Impossible to fit for Reduced Amdahl law : only %i elements" % len(D) 

  try:

    coeffs_Mylq, matcov_Mylq = curve_fit(Mylq, N, D)

    coeffs_Mylq[1]=coeffs_Mylq[1]*coeffs_Mylq[0]/D[0]

    # coeffs_Mylq[2]=coeffs_Mylq[2]*coeffs_Mylq[0]/D[0]

    coeffs_Mylq[3]=coeffs_Mylq[3]*coeffs_Mylq[0]/D[0]

    coeffs_Mylq[0]=D[0]

    print "Mylq Normalized : T=%.2f(%.6f+%.6f/N)+%.6f*N" % (coeffs_Mylq[0],

                                                            coeffs_Mylq[1],

                                                            coeffs_Mylq[3],

                                                            coeffs_Mylq[2])

    D_Mylq=Mylq(N,coeffs_Mylq[0],coeffs_Mylq[1],coeffs_Mylq[2],

                coeffs_Mylq[3])

  except:

    print "Impossible to fit for Mylq law : only %i elements" % len(D) 

  try:

    coeffs_Mylq2, matcov_Mylq2 = curve_fit(Mylq2, N, D)

    coeffs_Mylq2[1]=coeffs_Mylq2[1]*coeffs_Mylq2[0]/D[0]

    # coeffs_Mylq2[2]=coeffs_Mylq2[2]*coeffs_Mylq2[0]/D[0]

    # coeffs_Mylq2[3]=coeffs_Mylq2[3]*coeffs_Mylq2[0]/D[0]

    coeffs_Mylq2[4]=coeffs_Mylq2[4]*coeffs_Mylq2[0]/D[0]

    coeffs_Mylq2[0]=D[0]

    print "Mylq 2nd order Normalized: T=%.2f(%.6f+%.6f/N)+%.6f*N+%.6f*N^2" % \

        (coeffs_Mylq2[0],coeffs_Mylq2[1],

         coeffs_Mylq2[4],coeffs_Mylq2[2],coeffs_Mylq2[3])

  except:

    print "Impossible to fit for 2nd order Mylq law : only %i elements" % len(D) 

  if Curves:

    plt.xlabel("Number of Threads/work Items")

    plt.ylabel("Total Elapsed Time")

    Experience,=plt.plot(N,D,'ro') 

    try:

      pAmdahl,=plt.plot(N,D_Amdahl,label="Loi de Amdahl")    

      pMylq,=plt.plot(N,D_Mylq,label="Loi de Mylq")

    except:

      print "Fit curves seem not to be available"

    plt.legend()

    plt.show()

if __name__=='__main__':

  # Set defaults values

  # Alu can be CPU, GPU or ACCELERATOR

  Alu='CPU'

  # Id of GPU : 1 is for first find !

  Device=0

  # GPU style can be Cuda (Nvidia implementation) or OpenCL

  GpuStyle='OpenCL'

  # Parallel distribution can be on Threads or Blocks

  ParaStyle='Blocks'

  # Iterations is integer

  Iterations=100000000

  # JobStart in first number of Jobs to explore

  JobStart=1

  # JobEnd is last number of Jobs to explore

  JobEnd=16

  # JobStep is the step of Jobs to explore

  JobStep=1

  # Redo is the times to redo the test to improve metrology

  Redo=1

  # OutMetrology is method for duration estimation : False is GPU inside

  OutMetrology=False

  Metrology='InMetro'

  # Curves is True to print the curves

  Curves=False

  # Fit is True to print the curves

  Fit=False

  try:

    opts, args = getopt.getopt(sys.argv[1:],"hoclfa:g:p:i:s:e:t:r:d:",["alu=","gpustyle=","parastyle=","iterations=","jobstart=","jobend=","jobstep=","redo=","device="])

  except getopt.GetoptError:

    print '%s -o (Out of Core Metrology) -c (Print Curves) -f (Fit to Amdahl Law) -a <CPU/GPU/ACCELERATOR> -d <DeviceId> -g <CUDA/OpenCL> -p <Threads/Hybrid/Blocks> -i <Iterations> -s <JobStart> -e <JobEnd> -t <JobStep> -r <RedoToImproveStats> ' % sys.argv[0]

    sys.exit(2)

  for opt, arg in opts:

    if opt == '-h':

      print '%s -o (Out of Core Metrology) -c (Print Curves) -f (Fit to Amdahl Law) -a <CPU/GPU/ACCELERATOR> -d <DeviceId> -g <CUDA/OpenCL> -p <Threads/Hybrid/Blocks> -i <Iterations> -s <JobStart> -e <JobEnd> -t <JobStep> -r <RedoToImproveStats>' % sys.argv[0]

      print "\nInformations about devices detected under OpenCL:"

      # For PyOpenCL import

      try:

        import pyopencl as cl

        Id=1

        for platform in cl.get_platforms():

          for device in platform.get_devices():

            deviceType=cl.device_type.to_string(device.type)

            deviceMemory=device.max_mem_alloc_size

            print "Device #%i of type %s with memory %i : %s" % (Id,deviceType,deviceMemory,device.name)

            Id=Id+1

        print

        sys.exit()

      except ImportError:

        print "Your platform does not seem to support OpenCL"

    elif opt == '-o':

      OutMetrology=True

      Metrology='OutMetro'

    elif opt == '-c':

      Curves=True

    elif opt == '-f':

      Fit=True

    elif opt in ("-a", "--alu"):

      Alu = arg

    elif opt in ("-d", "--device"):

      Device = int(arg)

    elif opt in ("-g", "--gpustyle"):

      GpuStyle = arg

    elif opt in ("-p", "--parastyle"):

      ParaStyle = arg

    elif opt in ("-i", "--iterations"):

      Iterations = numpy.uint64(arg)

    elif opt in ("-s", "--jobstart"):

      JobStart = int(arg)

    elif opt in ("-e", "--jobend"):

      JobEnd = int(arg)

    elif opt in ("-t", "--jobstep"):

      JobStep = int(arg)

    elif opt in ("-r", "--redo"):

      Redo = int(arg)

  if Alu=='CPU' and GpuStyle=='CUDA':

    print "Alu can't be CPU for CUDA, set Alu to GPU"

    Alu='GPU'

  if ParaStyle not in ('Blocks','Threads','Hybrid'):

    print "%s not exists, ParaStyle set as Threads !" % ParaStyle

    ParaStyle='Threads'

  print "Compute unit : %s" % Alu

  print "Device Identification : %s" % Device

  print "GpuStyle used : %s" % GpuStyle

  print "Parallel Style used : %s" % ParaStyle

  print "Iterations : %s" % Iterations

  print "Number of threads on start : %s" % JobStart

  print "Number of threads on end : %s" % JobEnd

  print "Number of redo : %s" % Redo

  print "Metrology done out of CPU/GPU : %r" % OutMetrology

  if GpuStyle=='CUDA':

    try:

      # For PyCUDA import

      import pycuda.driver as cuda

      import pycuda.gpuarray as gpuarray

      import pycuda.autoinit

      from pycuda.compiler import SourceModule

    except ImportError:

      print "Platform does not seem to support CUDA"

  if GpuStyle=='OpenCL':

    try:

      # For PyOpenCL import

      import pyopencl as cl

      Id=1

      for platform in cl.get_platforms():

        for device in platform.get_devices():

          deviceType=cl.device_type.to_string(device.type)

          print "Device #%i of type %s : %s" % (Id,deviceType,device.name)

          if Id == Device:

            # Set the Alu as detected Device Type

            Alu=deviceType

          Id=Id+1

    except ImportError:

      print "Platform does not seem to support CUDA"

  average=numpy.array([]).astype(numpy.float32)

  median=numpy.array([]).astype(numpy.float32)

  stddev=numpy.array([]).astype(numpy.float32)

  ExploredJobs=numpy.array([]).astype(numpy.uint32)

  Jobs=JobStart

  while Jobs <= JobEnd:

    avg,med,std=0,0,0

    ExploredJobs=numpy.append(ExploredJobs,Jobs)

    circle=numpy.zeros(Jobs).astype(numpy.uint64)

    if OutMetrology:

      duration=numpy.array([]).astype(numpy.float32)

      for i in range(Redo):

        start=time.time()

        if GpuStyle=='CUDA':

          try:

            a,m,s=MetropolisCuda(circle,Iterations,1,Jobs,ParaStyle)

          except:

            print "Problem with %i // computations on Cuda" % Jobs

        elif GpuStyle=='OpenCL':

          try:

            a,m,s=MetropolisOpenCL(circle,Iterations,1,Jobs,ParaStyle,

                                   Alu,Device)

          except:

            print "Problem with %i // computations on OpenCL" % Jobs            

        duration=numpy.append(duration,time.time()-start)

      if (a,m,s) != (0,0,0):

        avg=numpy.mean(duration)

        med=numpy.median(duration)

        std=numpy.std(duration)

      else:

        print "Values seem to be wrong..."

    else:

      if GpuStyle=='CUDA':

        try:

          avg,med,std=MetropolisCuda(circle,Iterations,Redo,Jobs,ParaStyle)

        except:

          print "Problem with %i // computations on Cuda" % Jobs

      elif GpuStyle=='OpenCL':

        try:

          avg,med,std=MetropolisOpenCL(circle,Iterations,Redo,Jobs,ParaStyle,Alu,Device)

        except:

          print "Problem with %i // computations on OpenCL" % Jobs            

    if (avg,med,std) != (0,0,0):

      print "jobs,avg,med,std",Jobs,avg,med,std

      average=numpy.append(average,avg)

      median=numpy.append(median,med)

      stddev=numpy.append(stddev,std)

    else:

      print "Values seem to be wrong..."

    #THREADS*=2

    if len(average)!=0:

      numpy.savez("Splutter_%s_%s_%s_%i_%i_%.8i_Device%i_%s_%s" % (Alu,GpuStyle,ParaStyle,JobStart,JobEnd,Iterations,Device,Metrology,gethostname()),(ExploredJobs,average,median,stddev))

      ToSave=[ ExploredJobs,average,median,stddev ]

      numpy.savetxt("Splutter_%s_%s_%s_%i_%i_%.8i_Device%i_%s_%s" % (Alu,GpuStyle,ParaStyle,JobStart,JobEnd,Iterations,Device,Metrology,gethostname()),numpy.transpose(ToSave))

    Jobs+=JobStep

  if Fit:

    FitAndPrint(ExploredJobs,median,Curves)