Saner patching of static loads

3 files changed, 71 insertions, 105 deletions

diff --git a/compiler/elymasAsm.ey b/compiler/elymasAsm.ey
index fcb584a..15ed57b 100644
--- a/compiler/elymasAsm.ey
+++ b/compiler/elymasAsm.ey
@@ -12,16 +12,16 @@
   } "%" defq
 
   # registers
-  < [ /rax /rcx /rdx /rbx /rsp /rbp /rsi /rdi /r8 /r9 /r10 /r11 /r12 /r13 /r14 /r15 ] { 1 -01 defv }' each > ==bit64table
+  < [ /rax /rcx /rdx /rbx /rsp /rbp /rsi /rdi /r8 /r9 /r10 /r11 /r12 /r13 /r14 /r15 ] { 1 -01 defv }' each > ==:bit64table
   { bit64table -01 . -- } /bit64assert deff
 
-  < [ /eax /ecx /edx /ebx /esp /ebp /esi /edi /r8d /r9d /r10d /r11d /r12d /r13d /r14d /r15d ] { 1 -01 defv }' each > ==bit32table
+  < [ /eax /ecx /edx /ebx /esp /ebp /esi /edi /r8d /r9d /r10d /r11d /r12d /r13d /r14d /r15d ] { 1 -01 defv }' each > ==:bit32table
   { bit32table -01 . -- } /bit32assert deff
 
-  < [ /ax /cx /dx /bx /sp /bp /si /di /r8w /r9w /r10w /r11w /r12w /r13w /r14w /r15w ] { 1 -01 defv }' each > ==bit16table
+  < [ /ax /cx /dx /bx /sp /bp /si /di /r8w /r9w /r10w /r11w /r12w /r13w /r14w /r15w ] { 1 -01 defv }' each > ==:bit16table
   { bit16table -01 . -- } /bit16assert deff
 
-  < [ /al /cl /dl /bl /spl /ah /bpl /ch /sil /dh /dil /bh /r8b /r9b /r10b /r11b /r12b /r13b /r14b /r15b ] { 1 -01 defv }' each > ==bit8table
+  < [ /al /cl /dl /bl /spl /ah /bpl /ch /sil /dh /dil /bh /r8b /r9b /r10b /r11b /r12b /r13b /r14b /r15b ] { 1 -01 defv }' each > ==:bit8table
   { bit8table -01 . -- } /bit8assert deff
   
   <
@@ -41,7 +41,7 @@
     [ /r13b /r13w /r13d /r13 ] { 13 -01 defv }' each
     [ /r14b /r14w /r14d /r14 ] { 14 -01 defv }' each
     [ /r15b /r15w /r15d /r15 ] { 15 -01 defv }' each
-  > ==regnoTable
+  > ==:regnoTable
 
   { regnoTable -01 . } /regno deff
 
@@ -49,7 +49,7 @@
     [ /al /cl /dl /bl ] { 0 -01 defv }' each
     [ /spl /bpl /sil /dil ] { 1 -01 defv }' each
     [ /r8b /r9b /r10b /r11b /r12b /r13b /r14b /r15b ] { 1 -01 defv }' each
-  > ==rexreqbyteTable
+  > ==:rexreqbyteTable
   
   { rexreqbyteTable -01 . } /rexreqbyte deff
 
@@ -228,13 +228,13 @@
       } /encode deff >
     } ==Mem
     # base register plus 8 bit displacement
-    { < _ bit64assert ==base _ 128 lt assert ==disp /none ==idx { ==r
+    { < _ bit64assert ==base /none ==idx _ 128 lt assert ==disp { ==r
         r base modrm01
         disp imm8
       } /encode deff >
     } ==MemDisp8
     # base register plus 32 bit displacement
-    { < _ bit64assert ==base ==disp /none ==idx { ==r
+    { < _ bit64assert ==base /none ==idx ==disp { ==r
         r base modrm10
         disp imm32
       } /encode deff >
@@ -1196,10 +1196,10 @@
   }' each
 
   STACKSIZE sys .asm .alloc _ globalAllocations .register
-                              .base ==quoteEncodingBufferCode
+                              .base ==:quoteEncodingBufferCode
 
   STACKSIZE sys .asm .alloc _ globalAllocations .register
-                              .base ==quoteEncodingBufferObjects
+                              .base ==:quoteEncodingBufferObjects
 
   { ==opcodes
     opcodes len 1 sub PAGESIZE div 1 add PAGESIZE mul sys .asm .alloc /codearea defv
diff --git a/compiler/elymasGlobal.ey b/compiler/elymasGlobal.ey
index 1820725..1b36b1b 100644
--- a/compiler/elymasGlobal.ey
+++ b/compiler/elymasGlobal.ey
@@ -1079,7 +1079,7 @@
 
       %40 /dl :testbImmReg
       /patchConstant :jnzLbl8
-      %10 /dl :testbImmReg
+      %30 /dl :testbImmReg # TODO separate cases for static and typed
       /patchStatic :jnzLbl32
 
       /rdx /rcx :movqRegReg
@@ -1102,11 +1102,11 @@
       /rax :callqReg
       :ud2
 
-      @active
+      @active # TODO patch late resolve call away
       |ey* /rax :movqImmReg
       /rax :callqReg
 
-      @inactive
+      @inactive # TODO patch late resolve call away
       @done
 
       16 /r15 :addqImm8Reg
@@ -1301,18 +1301,38 @@
       :ud2
 
       @patchStaticPassive
+      0 /rax :movqImmReg
+      /rax :pushqReg
+      /patchStaticImplementation :jmpLbl8
+
+      @patchStaticActive
+      1 /rax :movqImmReg
+      /rax :pushqReg
+
+      @patchStaticImplementation
+
       /rdi /rcx :movqRegReg # save parent pointer count
       # FIXME: keep enough place for repeated parent dereference when creating the to-be-patched-code
 
       15 /rcx :cmpqImm8Reg
-      /patchStaticPassiveParentCountWithinLimits :jbeLbl8
+      /patchStaticParentCountWithinLimits :jbeLbl8
 
       "parent count out of save limits in internalExecuteIdentifierUnquotedAndPatchLateResolve" ::outputError
       :ud2
 
-      @patchStaticPassiveParentCountWithinLimits
+      @patchStaticParentCountWithinLimits
       8 /r15 /rdi :movqMemDisp8Reg
-      # TODO: could reset string reference here
+      # replace string reference with example object (for later optimizer inspection)
+      8 /rdi /rdx :movqMemDisp8Reg
+      16 1 /rdx /rdi /rdx :leaqMemIndexScaleDisp8Reg
+      /rdi /eax :movlMemReg
+      /rdi /rax :addqRegReg
+      /rdx /rax :cmpqRegReg
+      /patchStaticParentCountWithinLimitsSpaceExists :jaLbl8
+      "no space available for example object" ::outputError
+      :ud2
+      @patchStaticParentCountWithinLimitsSpaceExists
+      0 /rdx :andqImm8Mem # reset example object pointer
       16 /rdi :addqImm8Reg
 
       [
@@ -1321,125 +1341,67 @@
       ] ::loadToRdi
 
       /rcx /rcx :testqRegReg
-      /patchStaticPassiveNoFollowParents :jzLbl8
-      @patchStaticPassiveFollowParents
+      /patchStaticNoFollowParents :jzLbl8
+      @patchStaticFollowParents
       [
         16 /rax /rax :movqMemDisp8Reg
       ] ::loadToRdi
-      /patchStaticPassiveFollowParents :loopLbl8
-      @patchStaticPassiveNoFollowParents
+      /patchStaticFollowParents :loopLbl8
+      @patchStaticNoFollowParents
 
       %01 /bpl :testbImmReg
-      /patchStaticPassiveLoadFromExtensionArea :jnzLbl8
+      /patchStaticLoadFromExtensionArea :jnzLbl8
 
-      # @patchStaticPassiveLoadFromScope
+      # @patchStaticLoadFromScope
       32 /rsi :addqImm8Reg # now: rsi == offset of entry in scope
       [
         /rbx :popqReg
-        %EE /rax :pushqMemDisp32
-        /rbx :jmpqReg
+        %EE /rax /rax :movqMemDisp32Reg
       ] ::loadToRdi
-      6 /rdi :subqImm8Reg
-      /esi /rdi :movlRegMem # patch offset
-
-      8 /r15 /rdi :movqMemDisp8Reg
-      16 /rdi :addqImm8Reg
-
-      16 /r15 :addqImm8Reg
-      /r15 :pushqMem
-      8 /r15 :addqImm8Reg
-      /rdi :jmpqReg # continue with freshly patched code
+      /esi 4 neg /rdi :movlRegMemDisp8 # patch offset
+      /patchStaticLoadTail :jmpLbl8
 
-      @patchStaticPassiveLoadFromExtensionArea
+      @patchStaticLoadFromExtensionArea
       40 /rsi :addqImm8Reg # now: rsi == offset of entry in extension area + scope length
-
       [
         24 /rax /rcx :movqMemDisp8Reg # load extension area pointer
         /rax /edx :movlMemReg # load scope length
         /rdx :negqReg # prepare for substraction
         /rbx :popqReg
-        %EE 1 /rdx /rcx :pushqMemIndexScaleDisp32 # push loaded entry to stack
-        /rbx :jmpqReg
+        %EE 1 /rdx /rcx /rax :movqMemIndexScaleDisp32Reg # load entry
       ] ::loadToRdi
-      6 /rdi :subqImm8Reg
-      /esi /rdi :movlRegMem # patch offset
-
-      8 /r15 /rdi :movqMemDisp8Reg
-      16 /rdi :addqImm8Reg
-
-      16 /r15 :addqImm8Reg
-      /r15 :pushqMem
-      8 /r15 :addqImm8Reg
-      /rdi :jmpqReg # continue with freshly patched code
-
-      @patchStaticActive
-      /rdi /rcx :movqRegReg # save parent pointer count
-      # FIXME: keep enough place for repeated parent dereference when creating the to-be-patched-code
-
-      15 /rcx :cmpqImm8Reg
-      /patchStaticActiveParentCountWithinLimits :jbeLbl8
-
-      "parent count out of save limits in internalExecuteIdentifierUnquotedAndPatchLateResolve" ::outputError
-      :ud2
-
-      @patchStaticActiveParentCountWithinLimits
-      8 /r15 /rdi :movqMemDisp8Reg
-      # TODO: could reset string reference here
-      16 /rdi :addqImm8Reg
+      /esi 4 neg /rdi :movlRegMemDisp8 # patch offset
 
+      @patchStaticLoadTail
       [
-        ::currentScope /rax :movqImmReg
-        /rax /rax :movqMemReg
+        /rax :pushqReg
+        /rcx :movqImmOOBReg
       ] ::loadToRdi
-
-      /rcx /rcx :testqRegReg
-      /patchStaticActiveNoFollowParents :jzLbl8
-      @patchStaticActiveFollowParents
+      /rdx /rdi :movqRegMem
+      8 /rdi :addqImm8Reg
       [
-        16 /rax /rax :movqMemDisp8Reg
+        /rax /rcx :movqRegMem
       ] ::loadToRdi
-      /patchStaticActiveFollowParents :loopLbl8
-      @patchStaticActiveNoFollowParents
 
-      %01 /bpl :testbImmReg
-      /patchStaticActiveLoadFromExtensionArea :jnzLbl8
+      /rax :popqReg
+      /rax /rax :testqRegReg
+      /patchStaticLoadActive :jnzLbl8
 
-      # @patchStaticActiveLoadFromScope
-      32 /rsi :addqImm8Reg # now: rsi == offset of entry in scope
+      # @patchStaticLoadPassive
       [
-        /rbx :popqReg
-        %EE /rax :pushqMemDisp32
-        /rbx :pushqReg
-        |ey* /rax :movqImmReg
-        /rax :jmpqReg
+        /rbx :jmpqReg
       ] ::loadToRdi
-      17 /rdi :subqImm8Reg
-      /esi /rdi :movlRegMem # patch offset
-
-      8 /r15 /rdi :movqMemDisp8Reg
-      16 /rdi :addqImm8Reg
-
-      16 /r15 :addqImm8Reg
-      /r15 :pushqMem
-      8 /r15 :addqImm8Reg
-      /rdi :jmpqReg # continue with freshly patched code
 
-      @patchStaticActiveLoadFromExtensionArea
-      40 /rsi :addqImm8Reg # now: rsi == offset of entry in extension area + scope length
+      /patchStaticLoadTail2 :jmpLbl8
 
+      @patchStaticLoadActive
       [
-        24 /rax /rcx :movqMemDisp8Reg # load extension area pointer
-        /rax /edx :movlMemReg # load scope length
-        /rdx :negqReg # prepare for substraction
-        /rbx :popqReg
-        %EE 1 /rdx /rcx :pushqMemIndexScaleDisp32 # push loaded entry to stack
         /rbx :pushqReg
         |ey* /rax :movqImmReg
         /rax :jmpqReg
       ] ::loadToRdi
-      17 /rdi :subqImm8Reg
-      /esi /rdi :movlRegMem # patch offset
 
+      @patchStaticLoadTail2
       8 /r15 /rdi :movqMemDisp8Reg
       16 /rdi :addqImm8Reg
 
diff --git a/compiler/standardClient.ey b/compiler/standardClient.ey
index 643a805..e00279f 100644
--- a/compiler/standardClient.ey
+++ b/compiler/standardClient.ey
@@ -220,17 +220,20 @@
   }" /cloneThread deffd
 
   { #==thread ==newpc
+    0 -1201 =[]
+  }" /updateThread deffd
+
+  { #==thread ==newpc
     [ -0201 threadGetCaptures _ len dearray ] ]
   }" /fullCloneThread deffd
 
   |add /origadd deffd
+  str .|bitTest /bitTest deffd
+  str .|bitSet /bitSet deffd
+  str .|zero /zero deffd
 
   # TODO think about implementation efficiency
   { ==maxSize
-    str .|bitTest /bitTest deffd
-    str .|bitSet /bitSet deffd
-    str .|zero /zero deffd
-
     <
     0 ==size
     [ maxSize { 0 }" rep ] =*get
@@ -281,7 +284,7 @@
         clist .clear
       }" { # TERM
         position maxPosition lt {
-          position string * 1 code * { pc 1 add thread cloneThread nlist .add }" rep
+          position string * 1 code * { pc 1 add thread updateThread nlist .add }" rep
         }" rep
       }" { # JUMP
         pc 1 code add thread cloneThread iPush
@@ -302,6 +305,7 @@
     0 ==i
     { position maxPosition le }" {
       0 =i
+
       { i clist .size lt }" {
         i clist .get _ =thread
                        threadGetPC _ =pc