[m-rev.] for post-commit review: group *int_least* types together

[Zoltan Somogyi]

On Sun, 18 Feb 2018 04:40:18 -0500 (EST), Julien Fischer <jfischer at opturion.com> wrote:

> 
> Hi,
> 
> On Sun, 18 Feb 2018, Zoltan Somogyi wrote:
> 
> > Julien, can you please have a look at the parts of the diff
> > to llds_out_global.m that are marked by XXX BUG?
> > I am not sure how I can write tests for the bounds of
> > {u,}int_least32 that will work on both 32 and 64 machines,
> > given the current state of bootstrapping 64 bit integers.
> 
> What's the issue with the current state of 64-bit integers?  Bootstrap
> compilers have been required to support them since the beginning of this
> month.  (Indeed, the Mercury compiler itself has been using them since
> then as well.)
> 
> The last major** uncommitted piece of work for 64-bit integers, is
> enabling unboxed 64-bit integers on 64-bit machines, but that only
> affect performance.  (I'll be committing the chnage that enables that
> tomorrow.)
> 
> Have I missed something here?

I phrased myself badly. What I want to do is to commit the following
diff:

--- a/compiler/llds_out_global.m
+++ b/compiler/llds_out_global.m
@@ -743,10 +743,11 @@ ok_int_const(N, LLDSType) :-
             IntLeastType = uint_least16,
             0 =< N, N < 65536
         ;
-            ( IntLeastType = int_least32
-            ; IntLeastType = uint_least32
-            )
-            % XXX BUG: The type int may be 64 bit.
+            IntLeastType = int_least32,
+            -2147483647 =< N, N < 2147483647
+        ;
+            IntLeastType = uint_least32,
+            0 =< N, N < 4294967295
         )
     ;
         LLDSType = lt_int(IntType),
@@ -764,10 +765,11 @@ ok_int_const(N, LLDSType) :-
             IntType = int_type_uint16,
             0 =< N, N < 65536
         ;
-            ( IntType = int_type_int32
-            ; IntType = int_type_uint32
-            )
-            % XXX BUG: The type int may be 64 bit.
+            IntType = int_type_int32,
+            -2147483647 =< N, N < 214748364
+        ;
+            IntType = int_type_uint32,
+            0 =< N, N < 4294967295
         ;
             ( IntType = int_type_int64
             ; IntType = int_type_uint64

I know that the references to 4294967295 will work
on 64 bit machines. What I want to know is: will they work
on 32 bit machines as well? (Where, ironically, they are
not needed.) I cannot test it, since I don't have
any 32 bit machines. I asked you because you were
the one to last touch the relevant parts of the compiler,
as part of adding support for 64 bit unboxed integers.

Zoltan.