Removal of Targetint.OCaml, move to Target_imm.Imm#500
Removal of Targetint.OCaml, move to Target_imm.Imm#500mshinwell merged 4 commits intoocaml-flambda:flambda2.0-stablefrom
Conversation
|
I've realised this needs a check at toplevel that we're not running on 32 bit. I don't think it is worth supporting those platforms here at present (and we already do not in Un_cps anyway). (update: check added) |
Gbury
left a comment
There was a problem hiding this comment.
This looks good and should work.
While we're at it, maybe we might want to consider adding some more checks in some of the conversions functions to be more safe and more consistent in the handling of overflows ? For instance One_bit_fewer.of_int silently truncates the input int to make it fit, whereas we probably want to make it produce an error ?
|
CRs added. We also agreed a subsequent patch will rename |
|
On second thoughts, the wasm work will need 32-bit support, so let's go for |
The code around
Targetint.OCamlis not ideal at present. This PR goes some way towards tidying it up. It movesTargetint.OCamlinto theTarget_immmodule; it is now calledTarget_imm.Imm. This significantly reducesTargetint-related changes outside the Flambda 2 portion of the tree.In due course we should squash the
Immmodule to just haveTarget_imm, but one step at a time.