Fix easyflash placement of auto copied main section

README.md

@@ -558,6 +558,11 @@ This sample reserves a fixed area of several banks for a common code segment.  A
 
 An alternative to calling the function in the cartridge ROM itself is to copy it to a different location in RAM and execute it there.  This sample creates functions in different ROM banks that are linked for a memory area in RAM at 0x7000.  The code is copied from the ROM to RAM and then executed.
 
+#### Easyflash low memory usage "easyflash.c"
+
+This sample will use the memory area starting from 0x0400 for the main code section when copying the code, using the stack page 0x100 for the startup itself, thus wasting small amount of RAM space.
+
+
 #### Terminate stay resident "tsr.c"
 
 A common usage for the RAM area in 0xc000..0xcfff which is visible but not usable from BASIC is to install a terminate stay resident BASIC extension, which can then be called using sys 49152.

include/c64/memmap.h

@@ -8,6 +8,7 @@
 #define MMAP_NO_ROM		0x35
 #define MMAP_RAM		0x30
 #define MMAP_CHAR_ROM	0x31
+#define MMAP_ALL_ROM	0x33
 
 // Install an IRQ an NMI trampoline, that routes the kernal interrupts
 // through an intermediate trampoline when the kernal ROM is not paged

include/crt.c

@@ -14,8 +14,9 @@
 #define	regs		__regs
 
 
-void StackStart, StackEnd, BSSStart, BSSEnd;
+void StackStart, StackEnd, BSSStart, BSSEnd, CodeStart, CodeEnd;
 
+#pragma section(code, 0x0000, CodeStart, CodeEnd)
 #pragma section(stack, 0x0000, StackStart, StackEnd)
 #pragma section(bss, 0x0000, BSSStart, BSSEnd)
 
@@ -26,6 +27,7 @@ int main(void);
 
 __asm startup
 {
+st0:
 #ifdef OSCAR_TARGET_CRT16
 		byt 0x09
 		byt 0x80
@@ -39,23 +41,88 @@ __asm startup
 
 // Copy cartridge rom to ram
 
-		ldx	#$40
+		ldx	#0
-		ldy	#$00
+lc0:	lda $8000, x
-		lda #$80
+		sta st0, x
+		inx	
+		bne	lc0
+
+// Set up address for decompress
+
+		lda #$00
+		sta ip
+		lda #$81
 		sta ip + 1
-		lda #$08
+
+		lda #<CodeStart
+		sta sp
+		lda #>CodeStart		
 		sta sp + 1		
-		sty ip
+
-		sty sp
+		ldy	#0
-l0:		lda (ip), y
+		lda (ip),y
+lx0:
+		sta accu
+		bmi wx1
+
+		clc
+		lda ip + 0
+		adc #1
+		sta addr + 0
+		lda ip + 1
+		adc #0
+		sta addr + 1
+
+		clc
+		lda addr + 0
+		adc accu
+		sta ip + 0
+		lda addr + 1
+		adc #0
+		sta ip + 1
+lx2:
+		ldy #0
+lx1:	lda (addr), y
 		sta (sp), y
 		iny
-		bne l0
+		cpy accu
-		inc ip + 1
+		bne lx1
-		inc sp + 1
+
-		dex
+		clc
-		bne l0
+		lda sp
+		adc accu
+		sta sp
+		lda sp + 1
+		adc #0
+		sta sp + 1
+
+		ldy #0
+		lda (ip), y
+		bne lx0
 		jmp w0
+wx1:
+		lda sp + 0
+		sec
+		ldy #1
+		sbc (ip), y
+		sta addr
+		lda sp + 1
+		sbc #0
+		sta addr + 1
+
+		clc
+		lda ip + 0
+		adc #2
+		sta ip + 0
+		lda ip + 1
+		adc #0
+		sta ip + 1
+
+		lda accu
+		and #$7f
+		sta accu
+		jmp lx2
+
 w0:
 		lda #$3f
 		sta $00

oscar64/InterCode.cpp

@@ -5365,6 +5365,8 @@ void InterCodeBasicBlock::UpdateLocalIntegerRangeSets(const GrowingVariableArray
 					{
 						vr.mMaxState = IntegerValueRange::S_BOUND;
 						vr.mMaxValue = 255;
+						vr.mMinState = IntegerValueRange::S_BOUND;
+						vr.mMinValue = 0;
 					}
 					if (vr.mMinState != IntegerValueRange::S_BOUND || vr.mMinValue < 0 || vr.mMinValue > 255)
 					{
@@ -11779,6 +11781,15 @@ void InterCodeBasicBlock::Disassemble(FILE* file, bool dumpSets)
 					fprintf(file, "-");
 			}
 			fprintf(file, "\n\n");
+			fprintf(file, "Exit  required temps : ");
+			for (i = 0; i < mExitRequiredTemps.Size(); i++)
+			{
+				if (mExitRequiredTemps[i])
+					fprintf(file, "#");
+				else
+					fprintf(file, "-");
+			}
+			fprintf(file, "\n\n");
 		}
 
 		for (i = 0; i < mInstructions.Size(); i++)
@@ -12958,6 +12969,8 @@ void InterCodeProcedure::Close(void)
 
 		DisassembleDebug("Peephole Temp Check");
 		
+		RemoveUnusedInstructions();
+
 		ReduceTemporaries();
 
 		MergeBasicBlocks();
@@ -13350,6 +13363,8 @@ void InterCodeProcedure::ReduceTemporaries(void)
 	for (i = 0; i < numTemps; i++)
 		collisionSet[i].Reset(numTemps);
 
+	Disassemble("ReduceTemps", true);
+
 	ResetVisited();
 	mEntryBlock->BuildCollisionTable(collisionSet);
 

oscar64/Linker.cpp

@@ -568,6 +568,62 @@ bool Linker::WritePrgFile(const char* filename)
 		return false;
 }
 
+static int memlzcomp(uint8 * dp, const uint8 * sp, int size)
+{
+	int	pos = 0, csize = 0;
+	while (pos < size)
+	{
+		int	pi = 0;
+		while (pi < 127 && pos < size)
+		{
+			int	bi = pi, bj = 0;
+			for (int i = 1; i < (pos < 255 ? pos : 255); i++)
+			{
+				int j = 0;
+				while (j < 127 && pos + j < size && sp[pos - i + j] == sp[pos + j])
+					j++;
+
+				if (j > bj)
+				{
+					bi = i;
+					bj = j;
+				}
+			}
+
+			if (bj >= 4)
+			{
+				if (pi > 0)
+				{
+					dp[csize++] = pi;
+					for (int i = 0; i < pi; i++)
+						dp[csize++] = sp[pos - pi + i];
+					pi = 0;
+				}
+
+				dp[csize++] = 128 + bj;
+				dp[csize++] = bi;
+				pos += bj;
+			}
+			else
+			{
+				pos++;
+				pi++;
+			}
+		}
+
+		if (pi > 0)
+		{
+			dp[csize++] = pi;
+			for (int i = 0; i < pi; i++)
+				dp[csize++] = sp[pos - pi + i];
+		}
+	}
+
+	dp[csize++] = 0;
+
+	return csize;
+}
+
 bool Linker::WriteCrtFile(const char* filename)
 {
 	FILE* file;
@@ -608,18 +664,27 @@ bool Linker::WriteCrtFile(const char* filename)
 		chipHeader.mBankNumber = 0;
 		chipHeader.mImageSize = 0x0020;
 
-		uint8 bootmem[8192];
+		uint8 bootmem[0x4000];
 
-		memset(bootmem, 0, 0x2000);
+		memset(bootmem, 0, 0x4000);
 
-		chipHeader.mLoadAddress = 0x0080;
+		LinkerRegion	* mainRegion = FindRegion(Ident::Unique("main"));
-		fwrite(&chipHeader, sizeof(chipHeader), 1, file);
+		LinkerRegion	* startupRegion = FindRegion(Ident::Unique("startup"));
-		fwrite(mMemory + 0x0800, 1, 0x2000, file);
 
-		memcpy(bootmem, mMemory + 0x2800, 0x1f00);
+		memcpy(bootmem, mMemory + startupRegion->mStart, startupRegion->mNonzero - startupRegion->mStart);
+		int usedlz = memlzcomp(bootmem + 0x0100, mMemory + mainRegion->mStart, mainRegion->mNonzero - mainRegion->mStart);
 
-		bootmem[0x1ffc] = 0x00;
+		Location	loc;
-		bootmem[0x1ffd] = 0xff;
+
+		if (usedlz > 0x03e00)
+		{
+			mErrors->Error(loc, ERRR_INSUFFICIENT_MEMORY, "Can not fit main region into first ROM bank");
+			fclose(file);
+			return false;
+		}
+
+		bootmem[0x3ffc] = 0x00;
+		bootmem[0x3ffd] = 0xff;
 
 		uint8	bootcode[] = {
 			0xa9, 0x87,
@@ -629,7 +694,7 @@ bool Linker::WriteCrtFile(const char* filename)
 			0x6c, 0xfc, 0xff
 		};
 
-		int j = 0x1f00;
+		int j = 0x3f00;
 		for (int i = 0; i < sizeof(bootcode); i++)
 		{
 			bootmem[j++] = 0xa9;
@@ -642,10 +707,14 @@ bool Linker::WriteCrtFile(const char* filename)
 		bootmem[j++] = 0x00;
 		bootmem[j++] = 0x04;
 
-		chipHeader.mLoadAddress = 0x00e0;
+		chipHeader.mLoadAddress = 0x0080;
 		fwrite(&chipHeader, sizeof(chipHeader), 1, file);
 		fwrite(bootmem, 1, 0x2000, file);
 
+		chipHeader.mLoadAddress = 0x00e0;
+		fwrite(&chipHeader, sizeof(chipHeader), 1, file);
+		fwrite(bootmem + 0x2000, 1, 0x2000, file);
+
 		for (int i = 1; i < 64; i++)
 		{
 			if (mCartridgeBankUsed[i])

oscar64/NativeCodeGenerator.cpp

@@ -1178,6 +1178,25 @@ bool NativeCodeInstruction::IsSame(const NativeCodeInstruction& ins) const
 		return false;
 }
 
+bool NativeCodeInstruction::MayBeMovedBefore(const NativeCodeInstruction& ins)
+{
+	if ((ChangesAddress() || ins.ChangesAddress()) && MayBeSameAddress(ins))
+		return false;
+	if (RequiresXReg() && ins.ChangesXReg() || ins.RequiresXReg() && ChangesXReg())
+		return false;
+	if (RequiresYReg() && ins.ChangesYReg() || ins.RequiresYReg() && ChangesYReg())
+		return false;
+	if (RequiresAccu() && ins.ChangesAccu() || ins.RequiresAccu() && ChangesAccu())
+		return false;
+	if (RequiresCarry() && ins.ChangesCarry() || ins.RequiresCarry() && ChangesCarry())
+		return false;
+
+	if ((ins.mLive & LIVE_CPU_REG_Z) && ChangesZFlag())
+		return false;
+
+	return true;
+}
+
 bool NativeCodeInstruction::MayBeSameAddress(const NativeCodeInstruction& ins, bool sameXY) const
 {
 	if (ins.mMode == ASMIM_ZERO_PAGE)
@@ -11692,7 +11711,7 @@ bool NativeCodeBasicBlock::RegisterValueForwarding(void)
 			{
 				if (mIns[i].mType == ASMIT_LDX)
 				{
-					if (xreg[mIns[i].mAddress])
+					if (xreg[mIns[i].mAddress] && !(mIns[i].mLive & LIVE_CPU_REG_Z))
 					{
 						mIns[i].mType = ASMIT_NOP; mIns[i].mMode = ASMIM_IMPLIED;
 						changed = true;
@@ -11719,7 +11738,7 @@ bool NativeCodeBasicBlock::RegisterValueForwarding(void)
 				}
 				else if (mIns[i].mType == ASMIT_LDY)
 				{
-					if (yreg[mIns[i].mAddress])
+					if (yreg[mIns[i].mAddress] && !(mIns[i].mLive & LIVE_CPU_REG_Z))
 					{
 						mIns[i].mType = ASMIT_NOP; mIns[i].mMode = ASMIM_IMPLIED;
 						changed = true;
@@ -11746,7 +11765,7 @@ bool NativeCodeBasicBlock::RegisterValueForwarding(void)
 				}
 				else if (mIns[i].mType == ASMIT_LDA)
 				{
-					if (areg[mIns[i].mAddress])
+					if (areg[mIns[i].mAddress] && !(mIns[i].mLive & LIVE_CPU_REG_Z))
 					{
 						mIns[i].mType = ASMIT_NOP; mIns[i].mMode = ASMIM_IMPLIED;
 						changed = true;
@@ -12333,6 +12352,36 @@ bool NativeCodeBasicBlock::AlternateXYUsage(void)
 	return changed;
 }
 
+bool NativeCodeBasicBlock::SimplifyLoopEnd(NativeCodeProcedure* proc)
+{
+	bool	changed = false;
+
+	if (!mVisited)
+	{
+		mVisited = true;
+
+		if (mTrueJump && !mFalseJump && mTrueJump->mLoopHead)
+		{
+			if (mTrueJump->mIns.Size() == 1 && mTrueJump->mFalseJump)
+			{
+				mIns.Push(mTrueJump->mIns[0]);
+				mBranch = mTrueJump->mBranch;
+				mFalseJump = mTrueJump->mFalseJump;
+				mTrueJump = mTrueJump->mTrueJump;
+
+				changed = true;
+			}
+		}
+
+		if (mTrueJump && mTrueJump->SimplifyLoopEnd(proc))
+			changed = true;
+		if (mFalseJump && mFalseJump->SimplifyLoopEnd(proc))
+			changed = true;
+	}
+
+	return changed;
+}
+
 bool NativeCodeBasicBlock::SimplifyDiamond(NativeCodeProcedure* proc)
 {
 	bool	changed = false;
@@ -25330,6 +25379,22 @@ bool NativeCodeBasicBlock::PeepHoleOptimizer(NativeCodeProcedure* proc, int pass
 						progress = true;
 					}
 #endif
+					else if (
+						mIns[i + 0].ChangesAccuAndFlag() &&
+						mIns[i + 1].mType == ASMIT_STA &&
+						mIns[i + 2].MayBeMovedBefore(mIns[i + 1]) && mIns[i + 2].MayBeMovedBefore(mIns[i + 0]) &&
+						mIns[i + 3].mType == ASMIT_ORA && mIns[i + 3].mMode == ASMIM_IMMEDIATE && mIns[i + 3].mAddress == 0)
+					{
+						NativeCodeInstruction	ins(mIns[i + 2]);
+						mIns.Remove(i + 2);
+						mIns.Insert(i, ins);
+						mIns[i + 0].mLive |= LIVE_CPU_REG_A | LIVE_CPU_REG_C | LIVE_CPU_REG_X | LIVE_CPU_REG_Y;
+						mIns[i + 1].mLive |= mIns[i + 3].mLive;
+						mIns[i + 2].mLive |= mIns[i + 3].mLive;
+						mIns[i + 3].mType = ASMIT_NOP; mIns[i + 3].mMode = ASMIM_IMPLIED;
+						progress = true;
+					}
+
 					if (
 						mIns[i + 0].mType == ASMIT_LDY && mIns[i + 0].mMode == ASMIM_IMMEDIATE && mIns[i + 0].mAddress == 0 &&
 						mIns[i + 1].mType == ASMIT_LDA && mIns[i + 1].mMode == ASMIM_INDIRECT_Y &&
@@ -28065,6 +28130,10 @@ void NativeCodeProcedure::Optimize(void)
 			ResetVisited();
 			if (mEntryBlock->SimplifyDiamond(this))
 				changed = true;
+
+			ResetVisited();
+			if (mEntryBlock->SimplifyLoopEnd(this))
+				changed = true;
 		}
 
 #if 1
@@ -28086,12 +28155,13 @@ void NativeCodeProcedure::Optimize(void)
 #if 1
 		if (!changed && step < 8)
 		{
+			cnt = 0;
 			step++;
 			changed = true;
 
 		}
 #endif
-
+		else
 			cnt++;
 
 	} while (changed);
@@ -28100,7 +28170,10 @@ void NativeCodeProcedure::Optimize(void)
 	ResetVisited();
 	mEntryBlock->ReduceLocalYPressure();
 #endif
+
+#if 1
 	CompressTemporaries();
+#endif
 
 #if 1
 	ResetVisited();

oscar64/NativeCodeGenerator.h

@@ -138,6 +138,7 @@ public:
 	bool IsCommutative(void) const;
 	bool IsShift(void) const;
 	bool IsSimpleJSR(void) const;
+	bool MayBeMovedBefore(const NativeCodeInstruction& ins);
 
 	bool ReplaceYRegWithXReg(void);
 	bool ReplaceXRegWithYReg(void);
@@ -383,6 +384,7 @@ public:
 	bool ExpandADCToBranch(NativeCodeProcedure* proc);
 	bool Split16BitLoopCount(NativeCodeProcedure* proc);
 	bool SimplifyDiamond(NativeCodeProcedure* proc);
+	bool SimplifyLoopEnd(NativeCodeProcedure* proc);
 
 	bool MoveAccuTrainUp(int at, int end);
 	bool MoveAccuTrainsUp(void);

oscar64setup/oscar64setup.vdproj

@@ -136,6 +136,12 @@
         }
         "Entry"
         {
+        "MsmKey" = "8:_23211A34D7544532ABFD8D17B647717F"
+        "OwnerKey" = "8:_UNDEFINED"
+        "MsmSig" = "8:_UNDEFINED"
+        }
+        "Entry"
+        {
         "MsmKey" = "8:_24F2466589154E2DAA108C2EDB3BB432"
         "OwnerKey" = "8:_UNDEFINED"
         "MsmSig" = "8:_UNDEFINED"
@@ -1395,6 +1401,26 @@
             "IsDependency" = "11:FALSE"
             "IsolateTo" = "8:"
             }
+            "{1FB2D0AE-D3B9-43D4-B9DD-F88EC61E35DE}:_23211A34D7544532ABFD8D17B647717F"
+            {
+            "SourcePath" = "8:..\\samples\\memmap\\easyflashlow.c"
+            "TargetName" = "8:easyflashlow.c"
+            "Tag" = "8:"
+            "Folder" = "8:_A62A71A6A08941C5964B90112D87731F"
+            "Condition" = "8:"
+            "Transitive" = "11:FALSE"
+            "Vital" = "11:TRUE"
+            "ReadOnly" = "11:FALSE"
+            "Hidden" = "11:FALSE"
+            "System" = "11:FALSE"
+            "Permanent" = "11:FALSE"
+            "SharedLegacy" = "11:FALSE"
+            "PackageAs" = "3:1"
+            "Register" = "3:1"
+            "Exclude" = "11:FALSE"
+            "IsDependency" = "11:FALSE"
+            "IsolateTo" = "8:"
+            }
             "{1FB2D0AE-D3B9-43D4-B9DD-F88EC61E35DE}:_24F2466589154E2DAA108C2EDB3BB432"
             {
             "SourcePath" = "8:..\\include\\oscar.c"

samples/memmap/easyflashlow.c

@@ -0,0 +1,38 @@
+#include <c64/memmap.h>
+#include <c64/charwin.h>
+#include <c64/cia.h>
+#include <c64/vic.h>
+#include <string.h>
+
+#pragma section( startup, 0 )
+
+#pragma region( startup, 0x0100, 0x0200, , , { startup } )
+
+#pragma region( main, 0x0400, 0x8000, , , { code, data, bss, heap, stack } )
+
+CharWin	cw;
+
+int main(void)
+{
+	// Copy Char ROM
+	mmap_set(MMAP_ALL_ROM);
+
+	memcpy((char *)0xd000, (char *)0xd000, 0x1000);	
+
+	// Enable ROM
+	mmap_set(MMAP_ROM);
+
+	// Init CIAs (no kernal rom was executed so far)
+	cia_init();
+
+	// Init VIC
+	vic_setmode(VICM_TEXT, (char *)0xc000, (char *)0xd800);
+
+	// Prepare output window
+	cwin_init(&cw, (char *)0xc000, 0, 0, 40, 25);
+	cwin_clear(&cw);
+
+	cwin_put_string(&cw, p"Hello World", 7);
+
+	while (true) ; 
+}

samples/memmap/make.bat

@@ -8,4 +8,5 @@ call ..\..\bin\oscar64 charsetload.c -d64=charsetload.d64 -fz=../resources/chars
 call ..\..\bin\oscar64 easyflash.c -n -tf=crt
 call ..\..\bin\oscar64 easyflashreloc.c -n -tf=crt
 call ..\..\bin\oscar64 easyflashshared.c -n -tf=crt
+call ..\..\bin\oscar64 easyflashlow.c -n -tf=crt
 call ..\..\bin\oscar64 tsr.c -n -dNOFLOAT -dNOLONG