[sim/mips] Proposed change to mips_core_signal

[sim/mips] Proposed change to mips_core_signal

[Ben Elliston]

I am working on a simulator port to a MIPS-like core that has a vector
processing unit.  To produce more helpful diagnostic messages, it
would useful if mips_core_signal() could differentiate between scalar
memory transfers and vector memory transfers.

Are there any objections to adding two new elements to the
_transfer_type enum in sim-basics.h to identify such transfers?

Ben

Re: [sim/mips] Proposed change to mips_core_signal

[Chris G. Demetriou]

"Ben Elliston" <bje@redhat.com> writes:
> I am working on a simulator port to a MIPS-like core that has a vector
> processing unit.  To produce more helpful diagnostic messages, it
> would useful if mips_core_signal() could differentiate between scalar
> memory transfers and vector memory transfers.
> 
> Are there any objections to adding two new elements to the
> _transfer_type enum in sim-basics.h to identify such transfers?

I have no problem with it, but it's not clear that I know enough to
object.

(It doesn't really seem like this is a MIPS-specific issue either,
except that your port is MIPS-ish, right?)


chris

Re: [sim/mips] Proposed change to mips_core_signal

[Ben Elliston]

Chris G. Demetriou writes:

 > I have no problem with it, but it's not clear that I know enough to
 > object.  (It doesn't really seem like this is a MIPS-specific issue
 > either, except that your port is MIPS-ish, right?)

It's more than a MIPS-specific issue.  It potentially impacts all
simulators, since it alters the _transfer_type enum.  Something like:

--- sim-basics.h        2000/03/02 09:10:00     1.23
+++ sim-basics.h        2001/02/13 03:37:56
@@ -111,6 +111,8 @@
 typedef enum _transfer_type {
   read_transfer,
   write_transfer,
+  read_vector_transfer,
+  write_vector_transfer,
 } transfer_type;

Ben

Re: [sim/mips] Proposed change to mips_core_signal

[Ben Elliston]

Chris G. Demetriou writes:

 > > I am working on a simulator port to a MIPS-like core that has a vector
 > > processing unit.  To produce more helpful diagnostic messages, it
 > > would useful if mips_core_signal() could differentiate between scalar
 > > memory transfers and vector memory transfers.

 > > Are there any objections to adding two new elements to the
 > > _transfer_type enum in sim-basics.h to identify such transfers?

 > I have no problem with it, but it's not clear that I know enough to
 > object.

I've found a cleaner way of implementing this that does not affect
sim/common code.  Forget this thread ever existed!

Cheers, Ben