Segfault in `lower-jit` (`rewriteKernelCallABI`) lowering a KernelAbstractions launch to `gpu.LaunchFuncOp`

`raise=true` fails to lower a KA kernel that takes a struct argument containing a traced **scalar** (`raise_triton_custom_call` operand mismatch) on Julia 1.12 but not Julia 1.11

foreach optimizer pattern for much faster @compile with large number of parameter leafs

Missing gradient chlo.erf

Expose model logp and gradlogp

Extremely slow `@trace` for a stencil operation

Strange convert arguments for Comrade

Reactant precompilation 0.2.260

Missing complex erf

autodiff dynamic loops with conditional