Merge pull request #44 from quantumgizmos/issue40

Issue40

setup.py

@@ -2,7 +2,7 @@
 from Cython.Build import cythonize
 import numpy
 
-VERSION="0.1.51"
+VERSION="0.1.53"
 f=open("src/ldpc/VERSION","w+")
 f.write(VERSION)
 f.close()

src/ldpc/bp_decoder.pxd

@@ -29,7 +29,7 @@ cdef class bp_decoder:
     cdef int MEM_ALLOCATED
     cdef int input_vector_type
 
-    cpdef np.ndarray[np.int_t, ndim=1] decode(self, input_vector)
+    cpdef np.ndarray[np.uint8_t, ndim=1] decode(self, input_vector)
 
     cdef char* bp_decode_cy(self)
 

src/ldpc/bp_decoder.pyx

@@ -119,7 +119,7 @@ cdef class bp_decoder:
             for j in range(self.n): self.channel_probs[j]=error_rate
             self.error_rate=error_rate
 
-    cpdef np.ndarray[np.int_t, ndim=1] decode(self, input_vector):
+    cpdef np.ndarray[np.uint8_t, ndim=1] decode(self, input_vector):
 
         """
         Runs the BP decoder for a given input_vector.

src/ldpc/c_util.pxd

@@ -4,12 +4,12 @@ import numpy as np
 cimport numpy as np
 cimport cython
 
-# cdef char* numpy2char(np.ndarray[np.int_t, ndim=1] np_array,char* char_array)
+# cdef char* numpy2char(np.ndarray[np.uint8_t, ndim=1] np_array,char* char_array)
 cdef char* numpy2char(np_array,char* char_array)
 cdef char* spmatrix2char(matrix,char* char_array)
-cdef double* numpy2double(np.ndarray[np.float_t, ndim=1] np_array,double* double_array)
-cdef np.ndarray[np.int_t, ndim=1] char2numpy(char* char_array, int n)
-cdef np.ndarray[np.float_t, ndim=1] double2numpy(double* char_array, int n)
+cdef double* numpy2double(np.ndarray[dtype = np.float_t, ndim=1] np_array,double* double_array)
+cdef np.ndarray[dtype = np.uint8_t, ndim=1] char2numpy(char* char_array, int n)
+cdef np.ndarray[dtype = np.float_t, ndim=1] double2numpy(double* char_array, int n)
 
 
 cdef extern from "binary_char.h":

src/ldpc/c_util.pyx

@@ -3,7 +3,7 @@
 import numpy as np
 from scipy.sparse import spmatrix
 
-# cdef char* numpy2char(np.ndarray[np.int_t, ndim=1] np_array,char* char_array):
+# cdef char* numpy2char(np.ndarray[np.uint8_t, ndim=1] np_array,char* char_array):
 
 #     cdef int n = np_array.shape[0]
 #     for i in range(n): char_array[i]=np_array[i]
@@ -30,15 +30,15 @@ cdef double* numpy2double(np.ndarray[np.float_t, ndim=1] np_array,double* double
     for i in range(n): double_array[i]=np_array[i]
     return double_array
 
-cdef np.ndarray[np.int_t, ndim=1] char2numpy(char* char_array, int n):
+cdef np.ndarray[np.uint8_t, ndim=1] char2numpy(char* char_array, int n):
 
-    cdef np.ndarray[np.int_t, ndim=1] np_array=np.zeros(n).astype(int)
+    cdef np.ndarray[np.uint8_t, ndim=1] np_array=np.zeros(n).astype(np.uint8)
     for i in range(n):np_array[i]=char_array[i]
     return np_array
 
 cdef np.ndarray[np.float_t, ndim=1] double2numpy(double* char_array, int n):
 
-    cdef np.ndarray[np.float_t, ndim=1] np_array=np.zeros(n)
+    cdef np.ndarray[np.float_t, ndim=1] np_array=np.zeros(n, dtype=np.float64)
     for i in range(n):np_array[i]=char_array[i]
     return np_array
 

src/ldpc/experimental/bp_decoder2.pxd

@@ -27,7 +27,7 @@ cdef class bp_decoder:
     cdef double ms_scaling_factor
     cdef int MEM_ALLOCATED
 
-    cpdef np.ndarray[np.int_t, ndim=1] bp_decode(self, np.ndarray[np.int_t, ndim=1] syndrome)
+    cpdef np.ndarray[np.uint8_t, ndim=1] bp_decode(self, np.ndarray[np.uint8_t, ndim=1] syndrome)
 
     cdef char* bp_decode_cy(self)
 

src/ldpc/experimental/bp_decoder2.pyx

@@ -83,7 +83,7 @@ cdef class bp_decoder:
         else:
             print("Decoder not called")
 
-    cpdef np.ndarray[np.int_t, ndim=1] bp_decode(self, np.ndarray[np.int_t, ndim=1] syndrome):
+    cpdef np.ndarray[np.uint8_t, ndim=1] bp_decode(self, np.ndarray[np.uint8_t, ndim=1] syndrome):
         self.synd=numpy2char(syndrome,self.synd)
         self.bp_decode_cy()
         return char2numpy(self.bp_decoding,self.n)

src/ldpc/experimental/pymod2sparse.pxd

@@ -9,5 +9,5 @@ cdef class pymod2sparse():
 
     cpdef iter_row(self,int row_index,int reverse_iterate=False)
     cpdef iter_col(self,int col_index,int reverse_iterate=False)
-    cpdef np.ndarray[np.int_t, ndim=1] mul(self, np.ndarray[np.int_t, ndim=1] vector)
+    cpdef np.ndarray[np.uint8_t, ndim=1] mul(self, np.ndarray[np.uint8_t, ndim=1] vector)
 

src/ldpc/experimental/pymod2sparse.pyx

@@ -87,7 +87,7 @@ cdef class pymod2sparse():
 
         return self
 
-    cpdef np.ndarray[np.int_t, ndim=1] mul(self, np.ndarray[np.int_t, ndim=1] vector):
+    cpdef np.ndarray[np.uint8_t, ndim=1] mul(self, np.ndarray[np.uint8_t, ndim=1] vector):
 
         if len(vector)!=self.n:
             raise ValueError(f'Dimension mismatch. The supplied vector of length {len(vector)} cannot be multiplied by a matrix with dimesnions ({self.m},{self.n})!')

src/ldpc/mod2sparse.pxd

@@ -77,6 +77,6 @@ cdef class pymod2sparse():
 
     cpdef iter_row(self,int row_index,int reverse_iterate)
     cpdef iter_col(self,int col_index,int reverse_iterate)
-    cpdef np.ndarray[np.int_t, ndim=1] mul(self, np.ndarray[np.int_t, ndim=1] vector)
+    cpdef np.ndarray[np.uint8_t, ndim=1] mul(self, np.ndarray[np.uint8_t, ndim=1] vector)
 
 

src/ldpc/mod2sparse.pyx

@@ -173,7 +173,7 @@ cdef class pymod2sparse():
 
         return self
 
-    cpdef np.ndarray[np.int_t, ndim=1] mul(self, np.ndarray[np.int_t, ndim=1] vector):
+    cpdef np.ndarray[np.uint8_t, ndim=1] mul(self, np.ndarray[np.uint8_t, ndim=1] vector):
 
         if len(vector)!=self.n:
             raise ValueError(f'Dimension mismatch. The supplied vector of length {len(vector)} cannot be multiplied by a matrix with dimesnions ({self.m},{self.n})!')

src/ldpc/osd.pyx

@@ -154,7 +154,7 @@ cdef class bposd_decoder(bp_decoder):
             return self.osdw_decoding
 
 
-    cpdef np.ndarray[np.int_t, ndim=1] decode(self, input_vector):
+    cpdef np.ndarray[np.uint8_t, ndim=1] decode(self, input_vector):
 
         """
         Runs the BP+OSD decoder for a given syndrome.

test.py

@@ -0,0 +1,15 @@
+import numpy as np
+from ldpc.codes import rep_code
+from ldpc.bp_decoder import bp_decoder
+
+H = rep_code(3)
+print(H)
+
+bpd = bp_decoder(H, error_rate = 0.1, max_iter=0, bp_method='ps', ms_scaling_factor=0.625)
+
+syndrome = np.array([1,0])
+
+decoding = bpd.decode(syndrome)
+print(decoding)
+print(bpd.bp_decoding)
+print(bpd.log_prob_ratios)