meheff
meheff
Missing coverage (at least): * structs * constexpr * parameterization
The fuzzer has a built in timeout feature, where if the sample runs for longer than X, the sample is considered to fail. This could be used to fuzz for...
Our LEC flow has light usage and no fuzzing which means it probably has bitrotted some. To keep it fresh and bugfree we should fuzz it. Potential options of increasing...
I wonder if [Csmith](https://embed.cs.utah.edu/csmith/) would be useful for generating random input programs for fuzzing xlscc. It might need to be augmented with a wrapper which does I/O and feeds Csmith-generated...
Currrently XLS IR has two multiply operations, `kUMul` and `kSMul`, for unsigned and signed multiplication. They support arbitrary, mixed widths for the two operands and the result. It would be...
### Discussed in https://github.com/google/xls/discussions/646 Originally posted by **joachimschmidt557** June 27, 2022 Hi! I was wondering whether it would be possible to lower DSLX function invocations to module instantiations in the...
The JIT can be very slow to compile. I suspect it is because we are creating enormous functions. We should create a single function per node to speed things up.
Add an optimization which removes dead registers (and narrows registers too). We already have the necessary analysis in proc_state_optimization which removes state elements which are no observable. Same thing could...
Error is bit 51 is spuriously set in a tuple element of the return value. Minimized IR: ``` package sample file_number 0 "fake_file.x" top fn __sample__main(x0: bits[59], x1: bits[52]) ->...
For some XLS nodes (e.g., trace) the JIT emits calls to functions defined in ir_builder_visitor: https://github.com/google/xls/blob/df241840abdedfdf64443f18de41fd70abee3fcb/xls/jit/ir_builder_visitor.cc#L410 This will not work with AOT compilation. These calls need to be replaced with...