Rabbit implementation #1613
Rabbit implementation #1613
Conversation
|
|
| assert self.m == other.m | ||
| self.v.update(other.v) | ||
|
|
||
|
|
There was a problem hiding this comment.
This class will be removed from types.py as it's not needed. The rabbit comparison logic is added in non_linear.py
| dest="edabit", | ||
| help="mixing arithmetic and binary computation using edaBits", | ||
| ) | ||
| parser.add_option( |
There was a problem hiding this comment.
I tried setting the flag for comparison_rabbit with program.use_comparison_rabbit(True) in the .mpc file (similar to use_edabits), but it always remained false, and I only managed to set it true by adding --comparison_rabit in the compilation instruction. Do you see what I am doing wrong there?
| """ | ||
| print_str(str(s) + '\n', *args, **kwargs) | ||
|
|
||
|
|
| z = floatingpoint.PreOpL(floatingpoint.or_op, y[::-1])[::-1] + [0] | ||
| w = [z[i] - z[i + 1] for i in range(BIT_SIZE)] | ||
|
|
||
| return types.sintbit(1) - types.sintbit(sum((R_bits[i] & w[i]) for i in range(BIT_SIZE))) |
There was a problem hiding this comment.
This line was initially 1 - sum(...) but I was getting the error below, and it only worked when I converted 1 and the value of the sum. Could you explain this part?
File "/benchmarks/MP-SPDZ/Compiler/comparison.py", line 85, in LTZ
movs(s, program.non_linear.ltz(a, k))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/benchmarks/MP-SPDZ/Compiler/non_linear.py", line 81, in ltz
return self.rabbitLTZ(a, k)
^^^^^^^^^^^^^^^^^^^^
File "/benchmarks/MP-SPDZ/Compiler/non_linear.py", line 71, in rabbitLTZ
w[1] = self.LTBits(masked_a, r_bits, BIT_SIZE)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/benchmarks/MP-SPDZ/Compiler/non_linear.py", line 54, in LTBits
return_value = 1 - sum((R_bits[i] & w[i]) for i in range(BIT_SIZE))
~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
File "/benchmarks/MP-SPDZ/Compiler/types.py", line 229, in read_mem_operation
return operation(self, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/benchmarks/MP-SPDZ/Compiler/GC/types.py", line 523, in __add__
return self.xor_int(other)
^^^^^^^^^^^^^^^^^^^
File "/benchmarks/MP-SPDZ/Compiler/GC/types.py", line 585, in xor_int
self_bits = self.bit_decompose()
^^^^^^^^^^^^^^^^^^^^
File "/benchmarks/MP-SPDZ/Compiler/GC/types.py", line 86, in bit_decompose
suffix = [0] * (n - self.n)
~~^~~~~~~~
TypeError: unsupported operand type(s) for -: 'NoneType' and 'NoneType'
There was a problem hiding this comment.
Also, now the result of the comparison is no longer 0 or 1, but I had values like -1 and 2, so I think that there might be something wrong with the typing there.
There was a problem hiding this comment.
It might be unintuitive, but the description for variables needs to come after the variable to make the documentation work: https://mp-spdz.readthedocs.io/en/latest/Compiler.html#Compiler.program.Program.use_edabit
3 participants
No description provided.