Idan/add sumcheck tests #800
Conversation
…ests' into idan/add_sumcheck_tests
| } | ||
| } | ||
|
|
||
| MlePoly max_allowed_nof_polys_comine(const std::vector<MlePoly>& inputs) |
There was a problem hiding this comment.
(1) Where are those restrictions defined?
(2) When reaching those restrictions, we must make sure to have clear error messages and not some vague error that doesn't help the user understand the problem. It must say what the problem is, what the limit is and what he used. I noticed some errors (From CUDA backend but maybe CPU is the same) that was not clear about too complex combine but not saying why too complex.
| @@ -444,24 +681,27 @@ TEST_F(FieldTestBase, SumcheckCudaShouldFailCases) | |||
|
|
|||
| ASSERT_EQ(error, eIcicleError::INVALID_ARGUMENT); | |||
| }; | |||
| for (const auto& device : s_registered_devices) { | |||
| if (device == "CPU") continue; | |||
There was a problem hiding this comment.
So every backend has different limitations? Seems like there is some minimum and then some backends can support more? Is it documented somewhere? It means CPU code can work but then moving to GPU wouldn't. Maybe it's not a good idea even it it works. @mickeyasa @idanfr-ingo
This PR add some tests for sumcheck, and change an undefined behaviour at transcript creation in the tests.
The additional tests should check few more cases:
cuda-backend-branch: idan/hotfix_sumcheck