Ensure determinism

[damip]

Discussed in #73

^{Originally posted by damip January 9, 2022}
To make sure execution is deterministic we need to:

• Disable wasm threads support (wasmer frontpage states that it supports wasm threads that are non deterministic )
• Canonicalize NaNs ( https://docs.rs/wasmer-compiler-singlepass-near/latest/wasmer_compiler_singlepass/struct.Singlepass.html#method.canonicalize_nans )
• Limit stack depth to a config defined value (I found this but dunno if it is related: https://docs.rs/wasmer-compiler-singlepass/latest/wasmer_compiler_singlepass/struct.Singlepass.html#method.enable_stack_check )
• Limit memory usage to a config defined value ( see https://github.com/wasmerio/wasmer/blob/c33b89c/examples/tunables_limit_memory.rs )
• uninitialized memory
• look for other sources of nondeterminism (?)

[gterzian]

Limit memory usage to a config defined value ( see https://github.com/wasmerio/wasmer/blob/c33b89c/examples/tunables_limit_memory.rs

So this seems to have been done with https://github.com/massalabs/massa-sc-runtime/blob/main/src/tunable_memory.rs. Correct? @adrien-zinger

[gterzian]

Canonicalize NaNs ( https://docs.rs/wasmer-compiler-singlepass-near/latest/wasmer_compiler_singlepass/struct.Singlepass.html#method.canonicalize_nans )

And this is done at

massa-sc-runtime/src/execution_impl.rs

Line 29 in 6e722ee

compiler_config.canonicalize_nans(true);

[gterzian]

Except when otherwise specified, when an arithmetic operator returns NaN, there is nondeterminism in determining the specific bits of the NaN. However, wasm does still provide the guarantee that NaN values returned from an operation will not have 1 bits in their fraction field that aren't set in any NaN values in the input operands, except for the most significant bit of the fraction field (which most operators set to 1).
Except when otherwise specified, when an arithmetic operator with a floating point result type receives no NaN input values and produces a NaN result value, the sign bit of the NaN result value is nondeterministic.

Are these two fixed by using canonical nans? Source: https://github.com/WebAssembly/design/blob/037c6fe94151eb13e30d174f5f7ce851be0a573e/Nondeterminism.md

[damip]

Are these two fixed by using canonical nans?

should be, yes

[damip]

The only remaining potential source of nondeterminism is the stack limit.
This needs more study.

[gterzian]

The only remaining potential source of nondeterminism is the stack limit.
This needs more study.

This seems like a good starting point: wasmerio/wasmer#2807

[damip]

Based on the discussion with Wasmer people, https://github.com/massalabs/massa-sc-runtime/pull/127/files adds stack checks. This should be enough to ensure determinism