diff --git a/src/test/java/dk/camelot64/kickc/test/TestProgramsFast.java b/src/test/java/dk/camelot64/kickc/test/TestProgramsFast.java index b2fc789cd..78160aa69 100644 --- a/src/test/java/dk/camelot64/kickc/test/TestProgramsFast.java +++ b/src/test/java/dk/camelot64/kickc/test/TestProgramsFast.java @@ -3485,6 +3485,11 @@ public class TestProgramsFast extends TestPrograms { compileAndCompare("consolidate-constant-problem.c"); } + @Test + public void testConsolidateConstantProblem2() throws IOException { + compileAndCompare("consolidate-constant-problem-2.c"); + } + @Test public void testConsolidateArrayIndexProblem() throws IOException { compileAndCompare("consolidate-array-index-problem.c"); diff --git a/src/test/kc/consolidate-constant-problem-2.c b/src/test/kc/consolidate-constant-problem-2.c new file mode 100644 index 000000000..3dd4de8ad --- /dev/null +++ b/src/test/kc/consolidate-constant-problem-2.c @@ -0,0 +1,13 @@ +// Constant consolidation produces a byte* + byte* error + +char* const SCREEN = (char*)0x400; + +void main() { + + for(char COLS=0;COLS<2;COLS++) { + char* sc = SCREEN + COLS*4 + 2; + for(char i=0;i<4;i++) + sc[i] = COLS; + } + +} \ No newline at end of file diff --git a/src/test/ref/consolidate-constant-problem-2.asm b/src/test/ref/consolidate-constant-problem-2.asm new file mode 100644 index 000000000..398ce82da --- /dev/null +++ b/src/test/ref/consolidate-constant-problem-2.asm @@ -0,0 +1,57 @@ +// Constant consolidation produces a byte* + byte* error + // Commodore 64 PRG executable file +.file [name="consolidate-constant-problem-2.prg", type="prg", segments="Program"] +.segmentdef Program [segments="Basic, Code, Data"] +.segmentdef Basic [start=$0801] +.segmentdef Code [start=$80d] +.segmentdef Data [startAfter="Code"] +.segment Basic +:BasicUpstart(main) + .label SCREEN = $400 +.segment Code +main: { + .label __2 = 2 + .label sc = 2 + ldx #0 + __b1: + // for(char COLS=0;COLS<2;COLS++) + cpx #2 + bcc __b2 + // } + rts + __b2: + // COLS*4 + txa + asl + asl + // SCREEN + COLS*4 + clc + adc #SCREEN + adc #0 + sta.z __2+1 + // char* sc = SCREEN + COLS*4 + 2 + lda #2 + clc + adc.z sc + sta.z sc + bcc !+ + inc.z sc+1 + !: + ldy #0 + __b3: + // for(char i=0;i<4;i++) + cpy #4 + bcc __b4 + // for(char COLS=0;COLS<2;COLS++) + inx + jmp __b1 + __b4: + // sc[i] = COLS + txa + sta (sc),y + // for(char i=0;i<4;i++) + iny + jmp __b3 +} diff --git a/src/test/ref/consolidate-constant-problem-2.cfg b/src/test/ref/consolidate-constant-problem-2.cfg new file mode 100644 index 000000000..ae73bc7f6 --- /dev/null +++ b/src/test/ref/consolidate-constant-problem-2.cfg @@ -0,0 +1,28 @@ + +void main() +main: scope:[main] from + [0] phi() + to:main::@1 +main::@1: scope:[main] from main main::@5 + [1] main::COLS#2 = phi( main/0, main::@5/main::COLS#1 ) + [2] if(main::COLS#2<2) goto main::@2 + to:main::@return +main::@return: scope:[main] from main::@1 + [3] return + to:@return +main::@2: scope:[main] from main::@1 + [4] main::$1 = main::COLS#2 << 2 + [5] main::$2 = SCREEN + main::$1 + [6] main::sc#0 = main::$2 + 2 + to:main::@3 +main::@3: scope:[main] from main::@2 main::@4 + [7] main::i#2 = phi( main::@2/0, main::@4/main::i#1 ) + [8] if(main::i#2<4) goto main::@4 + to:main::@5 +main::@5: scope:[main] from main::@3 + [9] main::COLS#1 = ++ main::COLS#2 + to:main::@1 +main::@4: scope:[main] from main::@3 + [10] main::sc#0[main::i#2] = main::COLS#2 + [11] main::i#1 = ++ main::i#2 + to:main::@3 diff --git a/src/test/ref/consolidate-constant-problem-2.log b/src/test/ref/consolidate-constant-problem-2.log new file mode 100644 index 000000000..ebc047654 --- /dev/null +++ b/src/test/ref/consolidate-constant-problem-2.log @@ -0,0 +1,431 @@ + +CONTROL FLOW GRAPH SSA + +void main() +main: scope:[main] from __start + main::COLS#0 = 0 + to:main::@1 +main::@1: scope:[main] from main main::@5 + main::COLS#2 = phi( main/main::COLS#0, main::@5/main::COLS#1 ) + main::$0 = main::COLS#2 < 2 + if(main::$0) goto main::@2 + to:main::@return +main::@2: scope:[main] from main::@1 + main::COLS#3 = phi( main::@1/main::COLS#2 ) + main::$1 = main::COLS#3 * 4 + main::$2 = SCREEN + main::$1 + main::$3 = main::$2 + 2 + main::sc#0 = main::$3 + main::i#0 = 0 + to:main::@3 +main::@3: scope:[main] from main::@2 main::@4 + main::sc#2 = phi( main::@2/main::sc#0, main::@4/main::sc#1 ) + main::COLS#6 = phi( main::@2/main::COLS#3, main::@4/main::COLS#4 ) + main::i#2 = phi( main::@2/main::i#0, main::@4/main::i#1 ) + main::$4 = main::i#2 < 4 + if(main::$4) goto main::@4 + to:main::@5 +main::@4: scope:[main] from main::@3 + main::i#3 = phi( main::@3/main::i#2 ) + main::sc#1 = phi( main::@3/main::sc#2 ) + main::COLS#4 = phi( main::@3/main::COLS#6 ) + main::sc#1[main::i#3] = main::COLS#4 + main::i#1 = ++ main::i#3 + to:main::@3 +main::@5: scope:[main] from main::@3 + main::COLS#5 = phi( main::@3/main::COLS#6 ) + main::COLS#1 = ++ main::COLS#5 + to:main::@1 +main::@return: scope:[main] from main::@1 + return + to:@return + +void __start() +__start: scope:[__start] from + call main + to:__start::@1 +__start::@1: scope:[__start] from __start + to:__start::@return +__start::@return: scope:[__start] from __start::@1 + return + to:@return + +SYMBOL TABLE SSA +constant byte* const SCREEN = (byte*)$400 +void __start() +void main() +bool~ main::$0 +number~ main::$1 +byte*~ main::$2 +byte*~ main::$3 +bool~ main::$4 +byte main::COLS +byte main::COLS#0 +byte main::COLS#1 +byte main::COLS#2 +byte main::COLS#3 +byte main::COLS#4 +byte main::COLS#5 +byte main::COLS#6 +byte main::i +byte main::i#0 +byte main::i#1 +byte main::i#2 +byte main::i#3 +byte* main::sc +byte* main::sc#0 +byte* main::sc#1 +byte* main::sc#2 + +Adding number conversion cast (unumber) 2 in main::$0 = main::COLS#2 < 2 +Adding number conversion cast (unumber) 4 in main::$1 = main::COLS#3 * 4 +Adding number conversion cast (unumber) main::$1 in main::$1 = main::COLS#3 * (unumber)4 +Adding number conversion cast (unumber) 2 in main::$3 = main::$2 + 2 +Adding number conversion cast (unumber) 4 in main::$4 = main::i#2 < 4 +Successful SSA optimization PassNAddNumberTypeConversions +Simplifying constant pointer cast (byte*) 1024 +Simplifying constant integer cast 2 +Simplifying constant integer cast 4 +Simplifying constant integer cast 2 +Simplifying constant integer cast 4 +Successful SSA optimization PassNCastSimplification +Finalized unsigned number type (byte) 2 +Finalized unsigned number type (byte) 4 +Finalized unsigned number type (byte) 2 +Finalized unsigned number type (byte) 4 +Successful SSA optimization PassNFinalizeNumberTypeConversions +Inferred type updated to byte in main::$1 = main::COLS#3 * 4 +Alias main::COLS#2 = main::COLS#3 +Alias main::sc#0 = main::$3 +Alias main::COLS#4 = main::COLS#6 main::COLS#5 +Alias main::sc#1 = main::sc#2 +Alias main::i#2 = main::i#3 +Successful SSA optimization Pass2AliasElimination +Identical Phi Values main::COLS#4 main::COLS#2 +Identical Phi Values main::sc#1 main::sc#0 +Successful SSA optimization Pass2IdenticalPhiElimination +Simple Condition main::$0 [3] if(main::COLS#2<2) goto main::@2 +Simple Condition main::$4 [10] if(main::i#2<4) goto main::@4 +Successful SSA optimization Pass2ConditionalJumpSimplification +Constant main::COLS#0 = 0 +Constant main::i#0 = 0 +Successful SSA optimization Pass2ConstantIdentification +Removing unused procedure __start +Removing unused procedure block __start +Removing unused procedure block __start::@1 +Removing unused procedure block __start::@return +Successful SSA optimization PassNEliminateEmptyStart +Rewriting multiplication to use shift [2] main::$1 = main::COLS#2 * 4 +Successful SSA optimization Pass2MultiplyToShiftRewriting +Inlining constant with var siblings main::COLS#0 +Inlining constant with var siblings main::i#0 +Constant inlined main::COLS#0 = 0 +Constant inlined main::i#0 = 0 +Successful SSA optimization Pass2ConstantInlining +Adding NOP phi() at start of main +CALL GRAPH + +Created 2 initial phi equivalence classes +Coalesced [10] main::COLS#7 = main::COLS#1 +Coalesced [13] main::i#4 = main::i#1 +Coalesced down to 2 phi equivalence classes +Adding NOP phi() at start of main + +FINAL CONTROL FLOW GRAPH + +void main() +main: scope:[main] from + [0] phi() + to:main::@1 +main::@1: scope:[main] from main main::@5 + [1] main::COLS#2 = phi( main/0, main::@5/main::COLS#1 ) + [2] if(main::COLS#2<2) goto main::@2 + to:main::@return +main::@return: scope:[main] from main::@1 + [3] return + to:@return +main::@2: scope:[main] from main::@1 + [4] main::$1 = main::COLS#2 << 2 + [5] main::$2 = SCREEN + main::$1 + [6] main::sc#0 = main::$2 + 2 + to:main::@3 +main::@3: scope:[main] from main::@2 main::@4 + [7] main::i#2 = phi( main::@2/0, main::@4/main::i#1 ) + [8] if(main::i#2<4) goto main::@4 + to:main::@5 +main::@5: scope:[main] from main::@3 + [9] main::COLS#1 = ++ main::COLS#2 + to:main::@1 +main::@4: scope:[main] from main::@3 + [10] main::sc#0[main::i#2] = main::COLS#2 + [11] main::i#1 = ++ main::i#2 + to:main::@3 + + +VARIABLE REGISTER WEIGHTS +void main() +byte~ main::$1 22.0 +byte*~ main::$2 22.0 +byte main::COLS +byte main::COLS#1 22.0 +byte main::COLS#2 16.11111111111111 +byte main::i +byte main::i#1 202.0 +byte main::i#2 134.66666666666666 +byte* main::sc +byte* main::sc#0 22.4 + +Initial phi equivalence classes +[ main::COLS#2 main::COLS#1 ] +[ main::i#2 main::i#1 ] +Added variable main::$1 to live range equivalence class [ main::$1 ] +Added variable main::$2 to live range equivalence class [ main::$2 ] +Added variable main::sc#0 to live range equivalence class [ main::sc#0 ] +Complete equivalence classes +[ main::COLS#2 main::COLS#1 ] +[ main::i#2 main::i#1 ] +[ main::$1 ] +[ main::$2 ] +[ main::sc#0 ] +Allocated zp[1]:2 [ main::COLS#2 main::COLS#1 ] +Allocated zp[1]:3 [ main::i#2 main::i#1 ] +Allocated zp[1]:4 [ main::$1 ] +Allocated zp[2]:5 [ main::$2 ] +Allocated zp[2]:7 [ main::sc#0 ] +REGISTER UPLIFT POTENTIAL REGISTERS +Statement [4] main::$1 = main::COLS#2 << 2 [ main::COLS#2 main::$1 ] ( [ main::COLS#2 main::$1 ] { } ) always clobbers reg byte a +Removing always clobbered register reg byte a as potential for zp[1]:2 [ main::COLS#2 main::COLS#1 ] +Statement [5] main::$2 = SCREEN + main::$1 [ main::COLS#2 main::$2 ] ( [ main::COLS#2 main::$2 ] { } ) always clobbers reg byte a +Statement [6] main::sc#0 = main::$2 + 2 [ main::COLS#2 main::sc#0 ] ( [ main::COLS#2 main::sc#0 ] { } ) always clobbers reg byte a +Statement [10] main::sc#0[main::i#2] = main::COLS#2 [ main::COLS#2 main::sc#0 main::i#2 ] ( [ main::COLS#2 main::sc#0 main::i#2 ] { } ) always clobbers reg byte a +Removing always clobbered register reg byte a as potential for zp[1]:3 [ main::i#2 main::i#1 ] +Statement [4] main::$1 = main::COLS#2 << 2 [ main::COLS#2 main::$1 ] ( [ main::COLS#2 main::$1 ] { } ) always clobbers reg byte a +Statement [5] main::$2 = SCREEN + main::$1 [ main::COLS#2 main::$2 ] ( [ main::COLS#2 main::$2 ] { } ) always clobbers reg byte a +Statement [6] main::sc#0 = main::$2 + 2 [ main::COLS#2 main::sc#0 ] ( [ main::COLS#2 main::sc#0 ] { } ) always clobbers reg byte a +Statement [10] main::sc#0[main::i#2] = main::COLS#2 [ main::COLS#2 main::sc#0 main::i#2 ] ( [ main::COLS#2 main::sc#0 main::i#2 ] { } ) always clobbers reg byte a +Potential registers zp[1]:2 [ main::COLS#2 main::COLS#1 ] : zp[1]:2 , reg byte x , reg byte y , +Potential registers zp[1]:3 [ main::i#2 main::i#1 ] : zp[1]:3 , reg byte x , reg byte y , +Potential registers zp[1]:4 [ main::$1 ] : zp[1]:4 , reg byte a , reg byte x , reg byte y , +Potential registers zp[2]:5 [ main::$2 ] : zp[2]:5 , +Potential registers zp[2]:7 [ main::sc#0 ] : zp[2]:7 , + +REGISTER UPLIFT SCOPES +Uplift Scope [main] 336.67: zp[1]:3 [ main::i#2 main::i#1 ] 38.11: zp[1]:2 [ main::COLS#2 main::COLS#1 ] 22.4: zp[2]:7 [ main::sc#0 ] 22: zp[1]:4 [ main::$1 ] 22: zp[2]:5 [ main::$2 ] +Uplift Scope [] + +Uplifting [main] best 3111 combination reg byte y [ main::i#2 main::i#1 ] reg byte x [ main::COLS#2 main::COLS#1 ] zp[2]:7 [ main::sc#0 ] reg byte a [ main::$1 ] zp[2]:5 [ main::$2 ] +Uplifting [] best 3111 combination +Coalescing zero page register [ zp[2]:5 [ main::$2 ] ] with [ zp[2]:7 [ main::sc#0 ] ] - score: 1 +Allocated (was zp[2]:5) zp[2]:2 [ main::$2 main::sc#0 ] + +ASSEMBLER BEFORE OPTIMIZATION + // File Comments +// Constant consolidation produces a byte* + byte* error + // Upstart + // Commodore 64 PRG executable file +.file [name="consolidate-constant-problem-2.prg", type="prg", segments="Program"] +.segmentdef Program [segments="Basic, Code, Data"] +.segmentdef Basic [start=$0801] +.segmentdef Code [start=$80d] +.segmentdef Data [startAfter="Code"] +.segment Basic +:BasicUpstart(main) + // Global Constants & labels + .label SCREEN = $400 +.segment Code + // main +main: { + .label __2 = 2 + .label sc = 2 + // [1] phi from main to main::@1 [phi:main->main::@1] + __b1_from_main: + // [1] phi main::COLS#2 = 0 [phi:main->main::@1#0] -- vbuxx=vbuc1 + ldx #0 + jmp __b1 + // main::@1 + __b1: + // [2] if(main::COLS#2<2) goto main::@2 -- vbuxx_lt_vbuc1_then_la1 + cpx #2 + bcc __b2 + jmp __breturn + // main::@return + __breturn: + // [3] return + rts + // main::@2 + __b2: + // [4] main::$1 = main::COLS#2 << 2 -- vbuaa=vbuxx_rol_2 + txa + asl + asl + // [5] main::$2 = SCREEN + main::$1 -- pbuz1=pbuc1_plus_vbuaa + clc + adc #SCREEN + adc #0 + sta.z __2+1 + // [6] main::sc#0 = main::$2 + 2 -- pbuz1=pbuz1_plus_vbuc1 + lda #2 + clc + adc.z sc + sta.z sc + bcc !+ + inc.z sc+1 + !: + // [7] phi from main::@2 to main::@3 [phi:main::@2->main::@3] + __b3_from___b2: + // [7] phi main::i#2 = 0 [phi:main::@2->main::@3#0] -- vbuyy=vbuc1 + ldy #0 + jmp __b3 + // main::@3 + __b3: + // [8] if(main::i#2<4) goto main::@4 -- vbuyy_lt_vbuc1_then_la1 + cpy #4 + bcc __b4 + jmp __b5 + // main::@5 + __b5: + // [9] main::COLS#1 = ++ main::COLS#2 -- vbuxx=_inc_vbuxx + inx + // [1] phi from main::@5 to main::@1 [phi:main::@5->main::@1] + __b1_from___b5: + // [1] phi main::COLS#2 = main::COLS#1 [phi:main::@5->main::@1#0] -- register_copy + jmp __b1 + // main::@4 + __b4: + // [10] main::sc#0[main::i#2] = main::COLS#2 -- pbuz1_derefidx_vbuyy=vbuxx + txa + sta (sc),y + // [11] main::i#1 = ++ main::i#2 -- vbuyy=_inc_vbuyy + iny + // [7] phi from main::@4 to main::@3 [phi:main::@4->main::@3] + __b3_from___b4: + // [7] phi main::i#2 = main::i#1 [phi:main::@4->main::@3#0] -- register_copy + jmp __b3 +} + // File Data + +ASSEMBLER OPTIMIZATIONS +Removing instruction jmp __b1 +Removing instruction jmp __breturn +Removing instruction jmp __b3 +Removing instruction jmp __b5 +Succesful ASM optimization Pass5NextJumpElimination +Removing instruction __b1_from_main: +Removing instruction __breturn: +Removing instruction __b3_from___b2: +Removing instruction __b5: +Removing instruction __b1_from___b5: +Removing instruction __b3_from___b4: +Succesful ASM optimization Pass5UnusedLabelElimination + +FINAL SYMBOL TABLE +constant byte* const SCREEN = (byte*) 1024 +void main() +byte~ main::$1 reg byte a 22.0 +byte*~ main::$2 zp[2]:2 22.0 +byte main::COLS +byte main::COLS#1 reg byte x 22.0 +byte main::COLS#2 reg byte x 16.11111111111111 +byte main::i +byte main::i#1 reg byte y 202.0 +byte main::i#2 reg byte y 134.66666666666666 +byte* main::sc +byte* main::sc#0 sc zp[2]:2 22.4 + +reg byte x [ main::COLS#2 main::COLS#1 ] +reg byte y [ main::i#2 main::i#1 ] +reg byte a [ main::$1 ] +zp[2]:2 [ main::$2 main::sc#0 ] + + +FINAL ASSEMBLER +Score: 2446 + + // File Comments +// Constant consolidation produces a byte* + byte* error + // Upstart + // Commodore 64 PRG executable file +.file [name="consolidate-constant-problem-2.prg", type="prg", segments="Program"] +.segmentdef Program [segments="Basic, Code, Data"] +.segmentdef Basic [start=$0801] +.segmentdef Code [start=$80d] +.segmentdef Data [startAfter="Code"] +.segment Basic +:BasicUpstart(main) + // Global Constants & labels + .label SCREEN = $400 +.segment Code + // main +main: { + .label __2 = 2 + .label sc = 2 + // [1] phi from main to main::@1 [phi:main->main::@1] + // [1] phi main::COLS#2 = 0 [phi:main->main::@1#0] -- vbuxx=vbuc1 + ldx #0 + // main::@1 + __b1: + // for(char COLS=0;COLS<2;COLS++) + // [2] if(main::COLS#2<2) goto main::@2 -- vbuxx_lt_vbuc1_then_la1 + cpx #2 + bcc __b2 + // main::@return + // } + // [3] return + rts + // main::@2 + __b2: + // COLS*4 + // [4] main::$1 = main::COLS#2 << 2 -- vbuaa=vbuxx_rol_2 + txa + asl + asl + // SCREEN + COLS*4 + // [5] main::$2 = SCREEN + main::$1 -- pbuz1=pbuc1_plus_vbuaa + clc + adc #SCREEN + adc #0 + sta.z __2+1 + // char* sc = SCREEN + COLS*4 + 2 + // [6] main::sc#0 = main::$2 + 2 -- pbuz1=pbuz1_plus_vbuc1 + lda #2 + clc + adc.z sc + sta.z sc + bcc !+ + inc.z sc+1 + !: + // [7] phi from main::@2 to main::@3 [phi:main::@2->main::@3] + // [7] phi main::i#2 = 0 [phi:main::@2->main::@3#0] -- vbuyy=vbuc1 + ldy #0 + // main::@3 + __b3: + // for(char i=0;i<4;i++) + // [8] if(main::i#2<4) goto main::@4 -- vbuyy_lt_vbuc1_then_la1 + cpy #4 + bcc __b4 + // main::@5 + // for(char COLS=0;COLS<2;COLS++) + // [9] main::COLS#1 = ++ main::COLS#2 -- vbuxx=_inc_vbuxx + inx + // [1] phi from main::@5 to main::@1 [phi:main::@5->main::@1] + // [1] phi main::COLS#2 = main::COLS#1 [phi:main::@5->main::@1#0] -- register_copy + jmp __b1 + // main::@4 + __b4: + // sc[i] = COLS + // [10] main::sc#0[main::i#2] = main::COLS#2 -- pbuz1_derefidx_vbuyy=vbuxx + txa + sta (sc),y + // for(char i=0;i<4;i++) + // [11] main::i#1 = ++ main::i#2 -- vbuyy=_inc_vbuyy + iny + // [7] phi from main::@4 to main::@3 [phi:main::@4->main::@3] + // [7] phi main::i#2 = main::i#1 [phi:main::@4->main::@3#0] -- register_copy + jmp __b3 +} + // File Data + diff --git a/src/test/ref/consolidate-constant-problem-2.sym b/src/test/ref/consolidate-constant-problem-2.sym new file mode 100644 index 000000000..20fcc9fd7 --- /dev/null +++ b/src/test/ref/consolidate-constant-problem-2.sym @@ -0,0 +1,17 @@ +constant byte* const SCREEN = (byte*) 1024 +void main() +byte~ main::$1 reg byte a 22.0 +byte*~ main::$2 zp[2]:2 22.0 +byte main::COLS +byte main::COLS#1 reg byte x 22.0 +byte main::COLS#2 reg byte x 16.11111111111111 +byte main::i +byte main::i#1 reg byte y 202.0 +byte main::i#2 reg byte y 134.66666666666666 +byte* main::sc +byte* main::sc#0 sc zp[2]:2 22.4 + +reg byte x [ main::COLS#2 main::COLS#1 ] +reg byte y [ main::i#2 main::i#1 ] +reg byte a [ main::$1 ] +zp[2]:2 [ main::$2 main::sc#0 ]