Marcel Ullrich
Marcel Ullrich
> Was this the issue where several closures end up in the same slot? The issue might be related. At least the starting position is similar. But for the same...
> Was this the issue where several closures end up in the same slot? I opened #118 for the other issue.
This program should not need autodiff to run. I run it with ``` ./build/bin/thorin -o - -VVVV -d clos lit/autodiff/pow_autodiff_eval2.thorin.disabled ``` to get the issue. The behavior displayed in your...
Maybe helpful: `clos/pow_ad_eval_simpl.thorin` (only clos dialect) and `pow_autodiff.thorin` (also ad,direct) experience the same problem. (I am not if it was always that way or if this test worked at some...
> I'm pretty sure it's a duplicate of #126 but rn I get: > > ``` > error: cannot pass argument '(.insert (%autodiff.zero «2; .Idx 4294967296», 0:(.Idx 2), s_473244), _473271)'...
Related issue: File: lit/mem/closure/out_of_nothing2_simpl.thorin in ad_ptr_merge. `add_mem` removes code by wrongly updating memory operations: Original code: ```rust .con q_5577173_reshape_2446227 _2446228::[_2446268: .Idx 4294967296, _2446230: .Cn [%mem.M, %mem.Ptr (.Idx 4294967296, 0)]] @(1:(.Idx...
A further (possibly simpler) example: `lit/matrix/print_const_dyn_mat.thorin` in NeuralCoder3/thorin2/matrix_dialect Command: `./build/bin/thorin -d direct -d affine -d math -d matrix -d clos lit/matrix/print_const_dyn_mat.thorin -o - -VVVV` An additional mem is inserted in...
I do not know the inner workings of the closure conversion pass. But in general, it might be useful to solve this issue by adding a separate pass in mem...
> Together with a (fixed) version of reshape/scalerize that makes everything flat and a (maybe separate from the previous) pass that reorders arguments, this should be enough to fix this...
As note: An alternative (to #165) way how this issue manifests is during code generation (instead of application): `%core.trait.*` is expected to be resolved entirely before code generation. In some...