More refactoring in preparation for inlining.

doc/Adam - working notes.txt

@@ -50,12 +50,6 @@ Remember to clean up the aaaaa things.
 
 ----
 
-I guess I've gotta actually do some inlining now, huh? Except I'm worried about reusability. So I've gotta make it smarter - instead of "Is this nmethod reusable for all receiver maps?", I've got to ask, "Is this nmethod reusable for *this* receiver map?"
-
-For now, I care most about inlining constant slot accesses. So I just need a reusabilityCondition that checks to make sure that the slot is a constant and contains the same object.
-
-----
-
 Stuff I want to do:
 
   - In the allocator, maybe initialize it right away with a vector of all 32 precolored register values. Then use those in the locationAssigner. (I think that's what the textbook wants me to do; maybe it makes a difference?)
@@ -64,14 +58,6 @@ Stuff I want to do:
 
 ----
 
-Oh, that's a good point about non-LIFO blocks - maybe I could convert them on-the-fly, since most blocks don't actually outlive their stack frame. So instead of zapping the home-frame pointer, convert it to a pointer to the heap-allocated thing.
-
-----
-
-OK, it frickin' works! Except for the test case that tries to compile a method, because I broke the stuff when I made compilers require a compilationRequester. So I've just gotta fix that and then my new compiler can be declared officially working!
-
-----
-
 http://selflanguage.org/_static/published/dynamic-deoptimization.pdf
 
 So it looks like the deal is:
@@ -89,8 +75,10 @@ First step is just to create the scopeDesc info.
 
 ----
 
-But I'd better finish cleaning up what I've done so far first:
-  - Fix the runtime compiler (the compilationRequester thing).
-  - Decide what to do about the aaaaa places.
-  - Maybe convert some primitives to IR nodes? I've got that stuff paged into my brain anyway, might as well take a look at it.
-  - Maybe go through the compiler and just clean it up a bit, make sure everything's really really clear.
+So now I've split up the machine-level and source-level allocators. Good! What next?
+
+OK, so when we inline a call to a method, what's different?
+
+  - Instead of moving stuff to the machine-level outgoing locations, we've gotta just move it to some new values that we create to hold the scope's incoming rcvr+args. (And those values have to be located in memory if the inlined scope has blocks.) And the scope has to have a valueForOutgoingResult too. Basically just the same as for primitives.
+  - The incoming args still need to be stored in memory for uplevel access, though, right?
+  - Just gotta call generateInitializationForLocals.

objects/applications/asmKit/asmFrame/asmFrame.self

@@ -25,20 +25,6 @@ See the LICENSE file for license information.
             | ) .
         } | ) 
 
- bootstrap addSlotsTo: bootstrap stub -> 'globals' -> 'assemblerSystems' -> 'framework' -> () From: ( | {
-         'Category: helpers\x7fCategory: iterating\x7fCategory: registers\x7fModuleInfo: Module: asmFrame InitialContents: FollowSlot\x7fVisibility: public'
-
-         allRegisterLocations = ( |
-            | 
-            cachedAllRegisterLocations ifNil: [| regLocProto |
-              [todo cleanup]. "I don't like that this says klein instead of vmKit or something.
-                               Maybe this cache should live elsewhere? -- Adam, Mar. 2009"
-              regLocProto: klein locations register.
-              cachedAllRegisterLocations: allRegisters copyMappedBy: [|:r| regLocProto copyForRegister: r].
-              cachedAllRegisterLocations
-            ]).
-        } | ) 
-
  bootstrap addSlotsTo: bootstrap stub -> 'globals' -> 'assemblerSystems' -> 'framework' -> () From: ( | {
          'Category: helpers\x7fCategory: iterating\x7fCategory: registers\x7fModuleInfo: Module: asmFrame InitialContents: FollowSlot\x7fVisibility: public'
 

objects/applications/asmKit/asmFrame/asmFrameRegs.self

@@ -1,6 +1,7 @@
  '$Revision: 30.14 $'
  '
-Copyright 2006 Sun Microsystems, Inc. All rights reserved. Use is subject to license terms.
+Copyright 1992-2006 Sun Microsystems, Inc. and Stanford University.
+See the LICENSE file for license information.
 '
 
 

objects/applications/asmKit/asmPPC/asmPPC.self

@@ -272,12 +272,6 @@ Used for testing.\x7fModuleInfo: Module: asmPPC InitialContents: FollowSlot\x7fV
          bigEndianMixin* = bootstrap stub -> 'globals' -> 'assemblerSystems' -> 'framework' -> 'bigEndianMixin' -> ().
         } | ) 
 
- bootstrap addSlotsTo: bootstrap stub -> 'globals' -> 'assemblerSystems' -> 'ppc' -> () From: ( | {
-         'Category: caches\x7fModuleInfo: Module: asmPPC InitialContents: InitializeToExpression: (nil)\x7fVisibility: private'
-
-         cachedAllRegisterLocations <- bootstrap stub -> 'globals' -> 'nil' -> ().
-        } | ) 
-
  bootstrap addSlotsTo: bootstrap stub -> 'globals' -> 'assemblerSystems' -> 'ppc' -> () From: ( | {
          'Category: caches\x7fModuleInfo: Module: asmPPC InitialContents: InitializeToExpression: (nil)\x7fVisibility: private'
 

objects/applications/asmKit/asmSPARC/asmSPARC.self

@@ -205,12 +205,6 @@ SlotsToOmit: parent.
          bigEndianMixin* = bootstrap stub -> 'globals' -> 'assemblerSystems' -> 'framework' -> 'bigEndianMixin' -> ().
         } | ) 
 
- bootstrap addSlotsTo: bootstrap stub -> 'globals' -> 'assemblerSystems' -> 'sparc' -> () From: ( | {
-         'Category: caches\x7fModuleInfo: Module: asmSPARC InitialContents: InitializeToExpression: (nil)\x7fVisibility: private'
-
-         cachedAllRegisterLocations.
-        } | ) 
-
  bootstrap addSlotsTo: bootstrap stub -> 'globals' -> 'assemblerSystems' -> 'sparc' -> () From: ( | {
          'Category: caches\x7fModuleInfo: Module: asmSPARC InitialContents: InitializeToExpression: (nil)\x7fVisibility: private'