Anton Korobeynikov
Anton Korobeynikov
`svds_real_gen` fall-backs to full `svd` if all singular values are requested: ```R # If all singular values are requested, call svd() instead, # and give a warning if (k ==...
The spec contains some samples that cannot be compiled as of now. For example, https://p4.org/p4-spec/docs/p4-16-working-draft.html#sec-ops-on-hdrs mentions: ```p4 header H { bit x; bit y; } H h; h = {#};...
I am constantly seeing use-def to consume lots of memory. In many cases I'm seeing behavior like: ```log heap after P4::ComputeWriteSet: in use 6.4GB, max 6.4GB ProcessDefUse invoking P4::(anonymous namespace)::FindUninitialized...
In case the compile-time impact is high, how does this change compare to enabling LTO in the build? What I remember from Tofino days, we had rather significant speedup by...
Recently I was profiling the memory consumption of some large apps and it turns out that the memory allocation / deallocation pattern is very concerning. I collected the GC used...
> There should never be unparsed annotations actually used in the compiler -- it is just a matter of adding one `PARSE_`_type_ line to parseAnnotations.cpp (or in a backend file)....
Building e.g. a `frontend` target actually builds Tofino backend, which does not look correct to me.
We are having the following code in `lib/indent.h`: ```c++ namespace IndentCtl { inline std::ostream &endl(std::ostream &out) { return out
Each partial_apply consumes its arguments, therefore we should never release intermediate ones in the sequence of closures. Fixes #78847