Should this be renamed to pick_batchsize? On the one hand, the type is BatchDuplicated, but on the other hand, jacobian and gradient call it chunk

Sure, batchsize sgtm.

Otherwise, this looks good to me once test pass

I think test failures are unrelated? The tests I introduced pass locally

is this acceptable? it's hard to know given all the failing tests

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 71.63%. Comparing base (037dfed) to head (a44ae9d).
Report is 134 commits behind head on main.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

@@            Coverage Diff             @@
##             main    #1545      +/-   ##
==========================================
+ Coverage   67.50%   71.63%   +4.12%     
==========================================
  Files          31       32       +1     
  Lines       12668    13816    +1148     
==========================================
+ Hits         8552     9897    +1345     
+ Misses       4116     3919     -197     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@gdalle CI here seems really just blocked by this PR.

e.g.

BatchMixedDuplicated vector{float64} call |    3      3
Test Summary: | Pass  Total
TypeTree      |   20     20
ERROR: LoadError: LoadError: ArgumentError: Package EnzymeCore not found in current path:
- Run `import Pkg; Pkg.add("EnzymeCore")` to install the EnzymeCore package.

Stacktrace:
 [1] require(into::Module, mod::Symbol)
   @ Base ./loading.jl:893
 [2] include(fname::String)
   @ Base.MainInclude ./client.jl:444
 [3] top-level scope
   @ ~/work/Enzyme.jl/Enzyme.jl/test/runtests.jl:87
 [4] include(fname::String)
   @ Base.MainInclude ./client.jl:444
 [5] top-level scope
   @ none:6
in expression starting at /home/runner/work/Enzyme.jl/Enzyme.jl/test/utils.jl:2

My bad, I thought EnzymeCore was available for testing. I removed it from the test file

@wsmoses this should be good to go?

@vchuravy this looks like an easy merge?

Sorry I inadvertently formatted the code, will fix the diff

Is this mergeable?

sorry for all the constant shenanigans, but sadly this now has a merge conflict. I'll merge right after and ping me if its not (sorry this functionality isn't currently used in Enzyme or fixes an issue so I keep forgetting to hit merge after walking away as CI starts)

Alternatively, or in complement to this PR, I've been thinking of adding a batch size parameter to the Enzyme AD type, so that it can be statically inferred like in AutoForwardDiff. But maybe I should read the Width from the mode object instead? Is it the same?

On the other hand maybe this should be an adaptive function which dispatches on the array type, so that static arrays can get special treatment?

Actually I'm gonna put this on hold until I get a better sense of what it means to choose a batch size. I'll try to write something up for the DI docs and then go from there. Please don't merge in the meantime.