http://d.puremagic.com/issues/show_bug.cgi?id=6658



--- Comment #10 from Vladimir Panteleev <thecybershadow@gmail.com> 2013-04-02 12:50:44 EEST ---
I don't know what dt_t is, but judging from all 3 of its mentions in the 5381-line e2ir.c, I can only suppose that it is a leaky abstraction poking out of the DMD backend.

Anyway, I don't see how this applies to e2ir.c, since, as I've mentioned, dt_t only occurs 3 times in the file.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=6658



--- Comment #11 from Iain Buclaw <ibuclaw@ubuntu.com> 2013-04-02 03:01:26 PDT ---
(In reply to comment #10)
> I don't know what dt_t is, but judging from all 3 of its mentions in the 5381-line e2ir.c, I can only suppose that it is a leaky abstraction poking out of the DMD backend.
> 
> Anyway, I don't see how this applies to e2ir.c, since, as I've mentioned, dt_t only occurs 3 times in the file.

(In reply to comment #10)
> I don't know what dt_t is, but judging from all 3 of its mentions in the 5381-line e2ir.c, I can only suppose that it is a leaky abstraction poking out of the DMD backend.
> 
> Anyway, I don't see how this applies to e2ir.c, since, as I've mentioned, dt_t only occurs 3 times in the file.

In brief, why define all these dmd backend symbols (OPcall) that are of no use
to gcc, and when you can just build gcc trees directly (CALL_EXPR)?

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=6658



--- Comment #12 from Vladimir Panteleev <thecybershadow@gmail.com> 2013-04-02 21:57:32 EEST ---
I don't understand the argument.

Because... choosing a lower (and simpler) layer of abstraction would mean that less code would need to be reimplemented?

As I understand it, every time DMD implements a new expression type (for example, dot multiply), GDC and LDC would need to be updated. However, DMD already has a glue layer that lowers all of the D-specific expressions to something closer to abstract machine code, which I'd think is what alternative backends would be more interested in.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------