Move enum registration / conditional shard Inlining

[guusw]

Couldn't help it, I spent some extra time on improving build iteration times & cleaning up some of the enum registration.

It ties in a bit to my #472 branch too because I can move the enum help strings the source file too.

[guusw]

Some statistics before/after

Full shards source

With

rm -rf src/core src/extra
time cmake --build . --target shards

devel (Debug/Windows)

cmake --build . --target shards  0.00s user 0.01s system 0% cpu 1:46.59 total

this branch (Debug/Windows)

cmake --build . --target shards  0.00s user 0.01s system 0% cpu 52.644 total
cmake --build . --target shards  0.01s user 0.00s system 0% cpu 52.190 total

Partial rebuild (gfx/gizmos)

With

pushd src/extra/CMakeFiles/shards-extra.dir; rm -rf gfx gizmos; popd
time cmake --build . --target shards

devel (Debug/Windows)

cmake --build . --target shards  0.00s user 0.00s system 0% cpu 46.512 total
cmake --build . --target shards  0.00s user 0.00s system 0% cpu 46.027 total

this branch (Debug/Windows)

cmake --build . --target shards  0.00s user 0.00s system 0% cpu 20.348 total

[codecov]

Codecov Report

Merging #473 (478a339) into devel (49da721) will decrease coverage by 0.02%.
The diff coverage is 91.02%.

@@            Coverage Diff             @@
##            devel     #473      +/-   ##
==========================================
- Coverage   81.11%   81.09%   -0.03%     
==========================================
  Files         215      218       +3     
  Lines       28694    28735      +41     
==========================================
+ Hits        23275    23302      +27     
- Misses       5419     5433      +14     

Impacted Files	Coverage Δ
src/core/shards/bigint.cpp	95.76% <ø> (ø)
src/core/shards/flow.hpp	88.00% <ø> (ø)
src/core/shards/genetic.hpp	81.35% <ø> (ø)
src/core/shards/process.cpp	38.46% <ø> (ø)
src/core/shards/wasm.cpp	49.84% <ø> (ø)
src/core/shards/wires.hpp	93.54% <ø> (ø)
src/core/shards/ws.cpp	85.06% <ø> (ø)
src/extra/gfx/shader/flow_shards.cpp	86.90% <ø> (ø)
src/extra/gfx/shader/translator_utils.hpp	56.91% <ø> (ø)
src/extra/gfx/shards_types.hpp	100.00% <ø> (ø)
... and 26 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[sinkingsugar]

Well first of all I don't think this was a priority at all, second while nice on paper, I don't think shards will be inlined, C++ works in a very rudimentary way, inline functions have to be headers, anything .cpp gets it's own compilation unit. LTO might work but that's another bet.
Check assembly, I'm pretty sure this doesn't work.

Also you could have split the work.. enum stuff easy to approve, core runtime nope.

[sinkingsugar]

Maybe you should revert just the inline shards bit, that require valgrind measuring (that's how I implemented them...) before any refactor. 👍

[sinkingsugar]

Alright let's review it as is since you checked asm and get it out of the way then 👍
Btw you are already in SG timezone I can see @guusw 😄

[sinkingsugar]

Let me know when ready to review, add me 🙏

[guusw]

Yeah should be good, ran some test script as well

(defpure inner
  = .i
  .i (Math.Add 2))

(defwire test
  0 >= .v
  [] >= .seq
  (ForRange 0 10000000 (->
                        >= .av
                        .v (Math.Add .av) > .v
                        .v (Do inner) > .v
                        .v >> .seq))
  .v)

(defmesh root)
(schedule root test)
(if (run root) nil (throw "Root tick failed"))

Same timings comparing to devel, there's not that many inline shards and this should at least touch some to indicate that they're still inlined

[sinkingsugar]

Looks good 👍