Re: sequences of length 1 in inline-threaded

On Thu, Dec 02, 2004 at 08:44:41AM -0500, Etienne Gagnon wrote:
> Hi David,
> 
> David Bélanger wrote:
> >Is there a reason of using these 1-instruction sequences?
> 
> There is no fundamental reason.  If I remember correctly, the original 
> reason
> had to do with measurements, but I'm not sure.  It might also have been
> pure lazyness; not wanting to handle 1-instruction sequences differently. 
> ;-)
> 

ok, it is more tricky to get it right than I first thought.

Some of the 1-instruction sequences are not quite equivalent to the
direct implementation, there is some extra.  :)

I think I got it right though I didn't test much yet.  I have included a
patch for people interested.

I don't think it will change much performance.

Slightly less memory used.  Also, at code preparation, a lookup in the
splay tree is not needed for some of these 1-inst sequence.  That's
about it.

David

---

David Bélanger
Web page:   http://www.cs.mcgill.ca/~dbelan2/
Public key: http://www.cs.mcgill.ca/~dbelan2/public_key.txt

_______________________________________________
SableVM-devel mailing list
[email protected]
http://sablevm.org/lists/control/listinfo/sablevm-devel

Index: src/libsablevm/prepare_code.c
===================================================================
--- src/libsablevm/prepare_code.c	(revision 3059)
+++ src/libsablevm/prepare_code.c	(working copy)
@@ -6705,8 +6705,10 @@
 	  if (_svmf_no_inlining
 	      (env, instruction->code, SVM_IS_INLINED_BYTECODE))
 #else
-	  if (vm->instructions[instruction->code].flag ==
-	      SVM_INTRP_FLAG_CONTAINS_BRANCH_OR_CALL)
+	    if ((vm->instructions[instruction->code].flag ==
+		 SVM_INTRP_FLAG_CONTAINS_BRANCH_OR_CALL) ||
+		((instruction->sequence_length == 1) &&
+		 (!instruction->sequence_contains_resolution)))
 #endif
 	    {
 	      assert (instruction->sequence_length == 1);