Fix bugs in matrix multiplication with binary circuits.

Compiler/GC/types.py

@@ -17,6 +17,7 @@ from Compiler import instructions_base
 import Compiler.GC.instructions as inst
 import operator
 import math
+import itertools
 from functools import reduce
 
 class bits(Tape.Register, _structure, _bit):
@@ -1182,9 +1183,20 @@ class sbitintvec(sbitvec, _number, _bitint, _sbitintbase):
             return self.from_vec(other * x for x in self.v)
         elif isinstance(other, sbitfixvec):
             return NotImplemented
+        other_bits = util.bit_decompose(other)
+        m = float('inf')
+        for x in itertools.chain(self.v, other_bits):
+            try:
+                m = min(m, x.n)
+            except:
+                pass
+        if m == 1:
+            op = operator.mul
+        else:
+            op = operator.and_
         matrix = []
-        for i, b in enumerate(util.bit_decompose(other)):
-            matrix.append([x & b for x in self.v[:len(self.v)-i]])
+        for i, b in enumerate(other_bits):
+            matrix.append([op(x, b) for x in self.v[:len(self.v)-i]])
         v = sbitint.wallace_tree_from_matrix(matrix)
         return self.from_vec(v[:len(self.v)])
     __rmul__ = __mul__

Compiler/types.py

@@ -5263,7 +5263,8 @@ class Array(_vectorizable):
                 # length can be None for single-element arrays
                 length = 0
             base = self.address + index * self.value_type.mem_size()
-            if size is not None and isinstance(base, _register):
+            if size is not None and isinstance(base, _register) \
+               and not issubclass(self.value_type, _vec):
                 base = regint._expand_address(base, size)
             self.address_cache[program.curr_block, key] = \
                 util.untuplify([base + i * length \
@@ -6063,6 +6064,7 @@ class SubMultiArray(_vectorizable):
                     assert n_threads is None
                     if max(res_matrix.sizes) > 1000:
                         raise AttributeError()
+                    self.value_type.matrix_mul
                     A = self.get_vector()
                     B = other.get_vector()
                     res_matrix.assign_vector(

GC/TinySecret.h

@@ -146,7 +146,7 @@ public:
         if (this != &res)
             res.get_regs().assign(this->get_regs().begin(),
                     this->get_regs().begin()
-                            + max(size_t(n_bits), this->get_regs().size()));
+                            + min(size_t(n_bits), this->get_regs().size()));
 
         res.resize_regs(n_bits);
     }

Processor/Instruction.hpp

@@ -666,6 +666,21 @@ unsigned BaseInstruction::get_max_reg(int reg_type) const
           return r[1] + size;
       else
           return 0;
+  case TRANS:
+      if (reg_type == SBIT)
+      {
+          int n_outputs = n;
+          auto& args = start;
+          int n_inputs = args.size() - n_outputs;
+          long long res = 0;
+          for (int i = 0; i < n_outputs; i++)
+              res = max(res, args[i] + DIV_CEIL(n_inputs, 64));
+          for (int j = 0; j < n_inputs; j++)
+              res = max(res, args[n_outputs] + DIV_CEIL(n_outputs, 64));
+          return res;
+      }
+      else
+          return 0;
   default:
       if (get_reg_type() != reg_type)
           return 0;
@@ -731,7 +746,6 @@ unsigned BaseInstruction::get_max_reg(int reg_type) const
   case ANDM:
   case NOTS:
   case NOTCB:
-  case TRANS:
       size = DIV_CEIL(n, 64);
       break;
   case CONVCBIT2S: