merge

theano/compile/profilemode.py

@@ -7,6 +7,8 @@ from theano.gof.cc import OpWiseCLinker
 from theano import gof
 import theano.config as config
 
+import_time = time.time()
+
 class ProfileMode(Mode):
     def __init__(self, linker=default_linker, optimizer=default_optimizer):
         local_time = [0.0]
@@ -179,7 +181,7 @@ class ProfileMode(Mode):
         print '---------------------------'
         print ''
         
-        print 'local_time %fs (Time spent running thunks)'% local_time
+        print 'local_time %.3fs (Time spent running thunks)'% local_time
 
         if print_apply:
             print 'Apply-wise summary: <% of local_time spent at this position> <cumulative seconds> <apply time> <time per call> <nb_call> <Apply position> <Apply Op name>'
@@ -247,8 +249,15 @@ class ProfileMode(Mode):
                 %(max(0, len(sotimes)-n_ops_to_print),
                   sum(f for f, t, a, ci, nb_call in sotimes[n_ops_to_print:])*100,
                   sum(t for f, t, a, ci, nb_call in sotimes[n_ops_to_print:]))
+
         print '(*) Op is running a c implementation'
-        print 'compile time: %.3fs'%compile_time
+        print
+        total_time = time.time() - import_time
+        other_time = total_time - local_time - compile_time
+        print 'Time since import %.3fs'%(total_time)
+        print 'Local time %.3fs %.1f%%(Time spent running thunks)'% (local_time,local_time/total_time*100)
+        print 'Compile time: %.3fs %.1f%%'%(compile_time, compile_time/total_time*100)
+        print 'Other time since import %.3fs %.1f%%'%(other_time,other_time/total_time*100)
         
         if any([x[2].__name__.startswith("Gpu") for x in sotimes]):
             cpu=[]

theano/sandbox/cuda/Makefile

-CUDA_NDARRAY?=../../../../cuda_ndarray
-NUMPY_INCLUDE?=/usr/include/python2.6
-
-type_support.so : type_support.cu $(CUDA_NDARRAY)/cuda_ndarray.so
-	nvcc -g -shared -I$(CUDA_NDARRAY) -I$(CUDA_ROOT)/include -I$(NUMPY_INCLUDE) -o type_support.so -Xcompiler -fPIC type_support.cu -L$(CUDA_ROOT)/lib -L$(CUDA_NDARRAY) -lcuda_ndarray $(CFLAGS)
-
-clean : 
-	rm type_support.so *~ *.pyc */*~ */*.pyc

theano/sandbox/cuda/__init__.py

-from theano.sandbox.cuda.type import CudaNdarrayType
+import os, sys
+from theano.gof.compiledir import get_compiledir
+from theano.compile import optdb
+
+import logging, copy
+_logger_name = 'theano_cuda_ndarray'
+_logger = logging.getLogger(_logger_name)
+_logger.setLevel(logging.INFO)
+_logger.addHandler(logging.StreamHandler())
+def warning(*msg):
+    _logger.warning(_logger_name+'WARNING: '+' '.join(str(m) for m in msg))
+def info(*msg):
+    _logger.info(_logger_name+'INFO: '+' '.join(str(m) for m in msg))
+def debug(*msg):
+    _logger.debug(_logger_name+'DEBUG: '+' '.join(str(m) for m in msg))
+
+
+#compile type_support.cu
+#this need that nvcc(part of cuda) is installed
+
+old_file = os.path.join(os.path.split(__file__)[0],'type_support.so')
+if os.path.exists(old_file):
+    os.remove(old_file)
+
+try:
+    sys.path.append(get_compiledir())
+    from type_support.type_support import *
+
+except ImportError:
+
+    import nvcc_compiler
+
+    print __file__
+
+    cuda_path=os.path.split(old_file)[0]
+    code = open(os.path.join(cuda_path, "type_support.cu")).read()
 
+    loc = os.path.join(get_compiledir(),'type_support')
+    if not os.path.exists(loc):
+        os.makedirs(loc)
+ 
+    CUDA_NDARRAY=os.getenv('CUDA_NDARRAY')
+    include_dirs=[]
+    lib_dirs=[]
+    
+    if CUDA_NDARRAY:
+        include_dirs.append(CUDA_NDARRAY)
+        lib_dirs.append(CUDA_NDARRAY)
+    else:
+        import theano.sandbox
+        path = os.path.split(os.path.split(os.path.split(theano.sandbox.__file__)[0])[0])[0]
+        path2 = os.path.join(path,'cuda_ndarray')
+        if os.path.isdir(path2):
+            include_dirs.append(path2)
+            lib_dirs.append(path2)
+        else:
+            path = os.path.split(path)[0]
+            path2 = os.path.join(path,'cuda_ndarray')
+            include_dirs.append(path2)
+            lib_dirs.append(path2)
+
+    nvcc_compiler.nvcc_module_compile_str('type_support', code, location = loc, include_dirs=include_dirs, lib_dirs=lib_dirs, libs=['cuda_ndarray'])
+
+    from type_support.type_support import *
+
+
+
+from theano.sandbox.cuda.type import CudaNdarrayType
 from theano.sandbox.cuda.var import (CudaNdarrayVariable,
     CudaNdarrayConstant,
     CudaNdarraySharedVariable,
@@ -12,22 +78,7 @@ from basic_ops import (GpuFromHost, HostFromGpu, GpuElemwise,
 import opt
 import cuda_ndarray
 
-import os
 import theano.config as config
-from theano.compile import optdb
-
-import logging, copy
-_logger_name = 'theano_cuda_ndarray'
-_logger = logging.getLogger(_logger_name)
-_logger.setLevel(logging.WARN)
-_logger.addHandler(logging.StreamHandler())
-def warning(*msg):
-    _logger.warning(_logger_name+'WARNING: '+' '.join(str(m) for m in msg))
-def info(*msg):
-    _logger.info(_logger_name+'INFO: '+' '.join(str(m) for m in msg))
-def debug(*msg):
-    _logger.debug(_logger_name+'DEBUG: '+' '.join(str(m) for m in msg))
-
 
 def use(device=config.THEANO_GPU):
     if use.device_number is None:

theano/sandbox/cuda/basic_ops.py

@@ -5,7 +5,7 @@ from theano import Op, Type, Apply, Variable, Constant
 from theano import tensor, scalar, config
 
 from theano.sandbox.cuda.type import CudaNdarrayType
-from theano.sandbox.cuda.type_support import filter as type_support_filter
+from theano.sandbox.cuda import filter as type_support_filter
 
 from theano.sandbox.cuda.elemwise import NaiveAlgo
 

theano/sandbox/cuda/tests/test_bench_loopfusion.py

@@ -14,7 +14,7 @@ import logging, StringIO, time, sys
 import numpy
 
 import theano
-from theano.compile.sandbox import shared, pfunc
+from theano.compile import shared, pfunc
 from theano import tensor
 from theano.tensor.nnet import softplus
 from theano.sandbox.softsign import softsign

theano/sandbox/cuda/tests/test_blas.py

 import sys, time
-from theano.compile.sandbox.sharedvalue import shared
-from theano.compile.sandbox.pfunc import pfunc
+from theano.compile.sharedvalue import shared
+from theano.compile.pfunc import pfunc
 from theano import tensor
 
 import numpy

theano/sandbox/cuda/tests/test_opt.py

 import sys, time
-from theano.compile.sandbox.sharedvalue import shared
-from theano.compile.sandbox.pfunc import pfunc
+from theano.compile.sharedvalue import shared
+from theano.compile.pfunc import pfunc
 from theano import tensor
 import theano
 import numpy

theano/sandbox/cuda/tests/walltime.py

 import sys, time
-from theano.compile.sandbox.sharedvalue import shared
-from theano.compile.sandbox.pfunc import pfunc
+from theano.compile.sharedvalue import shared
+from theano.compile.pfunc import pfunc
 from theano import tensor
 
 import numpy

theano/sandbox/cuda/type.py

@@ -9,7 +9,7 @@ import theano.config as config
 
 import cuda_ndarray
 
-from theano.sandbox.cuda.type_support import filter as type_support_filter
+from theano.sandbox.cuda import filter as type_support_filter
 
 from theano.sandbox.cuda.nvcc_compiler import nvcc_module_compile_str
 

theano/sandbox/cuda/var.py

@@ -5,7 +5,7 @@ from theano import tensor
 from theano.compile import shared, SharedVariable, shared_constructor
 
 from theano.sandbox.cuda.type import CudaNdarrayType
-from theano.sandbox.cuda.type_support import filter as type_support_filter
+from theano.sandbox.cuda import filter as type_support_filter
 
 from theano.sandbox.cuda.basic_ops import HostFromGpu, GpuFromHost
 

theano/sandbox/scan.py

@@ -30,7 +30,7 @@ __docformat__ = 'restructedtext en'
 import traceback
 import numpy 
 import theano
-import theano.compile.sandbox
+import theano.compile
 from theano.tensor import opt
 from theano import gof
 from theano.compile import optdb