[m-dev.] diff: fix reuse codegen problem

[Peter Ross]

On Fri, Mar 09, 2001 at 01:38:49AM +1100, Fergus Henderson wrote:
> On 08-Mar-2001, Peter Ross <peter.ross at miscrit.be> wrote:
> > On Thu, Mar 08, 2001 at 11:50:35PM +1100, Fergus Henderson wrote:
> > > On 08-Mar-2001, Peter Ross <peter.ross at miscrit.be> wrote:
> > > > 
> > > > compiler/ml_unify_gen.m:
> > > >     Change from using the depreceated body binop to the unmkbody unop.
> > > 
> > > s/depreceated/deprecated/
> > > 
> > > What makes you think that the `body' binop is deprecated?
> > > If it is, why is it deprecated?
> >
> > Because mlds_output_binop no longer handles this case,
> 
> Ah, I see.
> 
> > and it used to work at some previous point,
> 
> Hmm, I'm quite skeptical about that.
> It looks to me like it can never have worked.
> 
> > so I assumed that we were no longer using this operator.
> > 
> > > > +++ compiler/ml_unify_gen.m	2001/03/08 09:36:41
> > > > @@ -1065,8 +1065,8 @@
> > > >  			Var2Rval = lval(Var2Lval)
> > > >  		;
> > > >  			Var2Rval = mkword(PrimaryTag,
> > > > -					binop(body, lval(Var2Lval),
> > > > -					ml_gen_mktag(ReusePrimaryTag)))
> > > > +					unop(std_unop(unmkbody),
> > > > +					lval(Var2Lval)))
> > > 
> > > Could you explain that one for me in more detail?
> >
> > When reusing a cell, we must ensure that the tag is correct.  If the tag
> > on the cell we wish to reuse is different to the one which should be on
> > the new cell, then we replace the old tag with the new tag.
> 
> OK, that much I understood.  I understand what the code is intended to do.
> What I don't understand is how your modification helps it to do that.
> The old code there
> 
>   			Var2Rval = mkword(PrimaryTag,
> 					binop(body, lval(Var2Lval),
> 					ml_gen_mktag(ReusePrimaryTag)))
> 
> looks fine to me.  If mlds_to_c is not handling the `body' binop,
> then mlds_to_c should be fixed, not the code here.
> 
> The new code that you replaced it with, namely
> 
>   			Var2Rval = mkword(PrimaryTag,
> 					unop(std_unop(unmkbody),
> 					lval(Var2Lval)))
> 
> looks completely wrong.  You can't just replace `body'
> with `unmkbody', since they do opposite things.
> 
> In particular,
> 
> 	(1) the argument to `unmkbody' is supposed to be something that
> 	resulted from a call to `mkbody', i.e. something without a tag.
> 	But the argument that you pass to it, i.e. `Var2Lval', might have
> 	a tag on it.
> 
> 	(2) the second argument to `mkword' is supposed to be either
> 	a pointer or something that resulted from a call to `mkbody',
> 	but you pass it the result of a call to `unmkbody', which
> 	is neither of those two.
> 

Thanks for pointing out my stupid mistakes.  I have backed the change
out.  I think what the problem was that previously code with the body
binop was never generated by the MLDS backend (including my previous
incarnations of the reuse branch), yet I thought it was.  That
compounded by misread of the unmkbody macro and you end up with the
above.


===================================================================


Estimated hours taken: 0.5
Branches: main, reuse

Back out my previous change to ml_unify_gen, and fix mlds_to_c instead
to handle the body binop.

compiler/ml_unify_gen.m:
    Back out my previous broken change.

compiler/mlds_to_c.m:
    Handle the body binop.


Index: compiler/ml_unify_gen.m
===================================================================
RCS file: /home/mercury1/repository/mercury/compiler/ml_unify_gen.m,v
retrieving revision 1.16.2.9
diff -u -r1.16.2.9 ml_unify_gen.m
--- compiler/ml_unify_gen.m	2001/03/08 09:42:44	1.16.2.9
+++ compiler/ml_unify_gen.m	2001/03/09 14:07:14
@@ -1065,8 +1065,8 @@
 			Var2Rval = lval(Var2Lval)
 		;
 			Var2Rval = mkword(PrimaryTag,
-					unop(std_unop(unmkbody),
-					lval(Var2Lval)))
+					binop(body, lval(Var2Lval),
+					ml_gen_mktag(ReusePrimaryTag)))
 		},
 
 		{ MLDS_Statement = ml_gen_assign(Var1Lval, Var2Rval, Context) },
Index: compiler/mlds_to_c.m
===================================================================
RCS file: /home/mercury1/repository/mercury/compiler/mlds_to_c.m,v
retrieving revision 1.55.2.7
diff -u -r1.55.2.7 mlds_to_c.m
--- compiler/mlds_to_c.m	2001/03/09 12:49:22	1.55.2.7
+++ compiler/mlds_to_c.m	2001/03/09 14:07:16
@@ -2837,6 +2837,14 @@
 		mlds_output_rval(Y),
 		io__write_string("]")
 	;
+		{ Op = body }
+	->
+		io__write_string("MR_body("),
+		mlds_output_rval(X),
+		io__write_string(", "),
+		mlds_output_rval(Y),
+		io__write_string(")")
+	;
 		{ c_util__string_compare_op(Op, OpStr) }
 	->
 		io__write_string("(strcmp("),
--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------