Divide two MemValue objects?

[imtiyazuddin]

How can I divide two MemValue objects? if not can I convert them to some other objects? what options would I need when compiling?

[mkskeller]

You should be to use operators just like for basic types. Futhermore, calling read() on a MemValue object returns the object "inside".

[imtiyazuddin]

I tried but I got following error :
File "/home/imtiyazuddin/Ubuntu/MP-SPDZ-master/Compiler/instructions.py", line 1226, in init
raise CompilerError('random inverse in ring not implemented')
Compiler.exceptions.CompilerError: random inverse in ring not implemented

I compiled with ./compile.py fairness -M -CDR 64 -l -X .

I tried to do compile without -R 64, it gets compiled and exectued but the result is not a floating point number?

[mkskeller]

This is because you're calling field division. What version are you using? The behaviour has changed in version 0.3.8 to be more intuitive: https://github.com/data61/MP-SPDZ/releases/tag/v0.3.8
In any case, you can convert to sfix or sfloat before the division to get the desired result.

[imtiyazuddin]

I am testing in 0.3.7, will try with 0.3.8, thanks!

[imtiyazuddin]

getting some error when trying to install 0.3.8 version:

In file included from /usr/include/openssl/conf.h:19:
/usr/include/openssl/macros.h:155:4: error: "OPENSSL_API_COMPAT expresses an impossible API compatibility level"

error "OPENSSL_API_COMPAT expresses an impossible API compatibility level"

^
1 error generated.
make: *** [Makefile:83: Processor/Program.o] Error 1

I have openssl 1.1.1h and in your documentation it says OpenSSL, tested against 1.1.1, so it should have worked right?

[mkskeller]

What system are you using, and what version of boost in particular? MP-SPDZ itself doesn't touch the variable that causes this error message?

[imtiyazuddin]

ubuntu 22.04 with boost 1.80

[mkskeller]

Ubuntu 22.04 doesn't come with boost 1.80. Maybe try removing it. MP-SPDZ installs boost when running make setup.

[imtiyazuddin]

how can I convet MemValue to sfix or sfloat ?

[mkskeller]

By construction, i.e., sfix(a) where = a = MemValue(...).

[imtiyazuddin]

getting same error
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/instructions.py", line 1298, in init
raise CompilerError('random inverse in ring not implemented')
Compiler.exceptions.CompilerError: random inverse in ring not implemented

does this also only work in 0.3.8? (having some issues installing it)

[mkskeller]

Can you post the full trace?

[imtiyazuddin]

Default bit length for compilation: 63
Default security parameter for compilation: 40
Compiling file Programs/Source/fairness.mpc
Traceback (most recent call last):
File "./compile.py", line 41, in
main(compiler)
File "./compile.py", line 36, in main
compilation(compiler)
File "./compile.py", line 19, in compilation
prog = compiler.compile_file()
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/compilerLib.py", line 411, in compile_file
exec(compile(infile.read(), infile.name, "exec"), self.VARS)
File "/tmp/tmpp96rasg2", line 94, in
w_p_unfav = (nunfav * np) / (n * npunfav)
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/types.py", line 141, in vectorized_operation
res = operation(self, *args, **kwargs)
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/types.py", line 2244, in rtruediv
a,b = self.get_random_inverse()
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/types.py", line 177, in vectorized_function
res = function(cls, *args, **kwargs)
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/types.py", line 209, in instruction_typed_operation
res = operation(self, *args, **kwargs)
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/types.py", line 2035, in get_random_inverse
inverse(*res)
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/instructions_base.py", line 406, in maybe_gf2n_instruction
return instruction(*args, **kwargs)
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/instructions_base.py", line 313, in maybe_vectorized_instruction
return instruction(*args, **kwargs)
File "/home/imtiyazuddin/Ubuntu/mpspdz-0.3.7/MP-SPDZ-master/Compiler/instructions.py", line 1298, in init
raise CompilerError('random inverse in ring not implemented')
Compiler.exceptions.CompilerError: random inverse in ring not implemented

[mkskeller]

What are the types of the variables in (nunfav * np) / (n * npunfav)?

[imtiyazuddin]

n is sint and remaining all MemValue()

[mkskeller]

MemValue is just a container type, so how have they been initialized?

[imtiyazuddin]

I am using them as counters for certain triggers
so they will have some int values

[mkskeller]

I see. If you want to make this work in the older version, you will need to convert any of them to sfix or sfloat manually.

[imtiyazuddin]

thanks!! will give it a shot!!