Optimized int shifts ge eight

autotest/autotest.bat

@@ -108,6 +108,12 @@ if %errorlevel% neq 0 goto :error
 ..\release\oscar64 -e -n asmtest.c
 if %errorlevel% neq 0 goto :error
 
+..\release\oscar64 -e bitshifttest.c
+if %errorlevel% neq 0 goto :error
+
+..\release\oscar64 -e -n bitshifttest.c
+if %errorlevel% neq 0 goto :error
+
 exit /b 0
 :error
 echo Failed with error #%errorlevel%.

autotest/bitshifttest.c

@@ -1,4 +1,5 @@
 #include <stdio.h>
+#include <assert.h>
 
 unsigned shl1b(int n)
 {
@@ -73,22 +74,123 @@ unsigned shr8n(int n)
 
 #pragma native(shr8n)
 
+void shl16b(unsigned xu, int xi)
+{
+	unsigned	ua[16];
+	int			ia[16];
+#assign s 0
+#repeat
+	ua[s] = xu << s;
+	ia[s] = xi << s;
+#assign s s + 1
+#until s == 16	
+
+	for(int i=0; i<16; i++)
+	{
+		assert(ua[i] == xu << i);
+		assert(ia[i] == xi << i);
+	}
+}
+
+void shr16b(unsigned xu, int xi)
+{
+	unsigned	ua[16];
+	int			ia[16];
+#assign s 0
+#repeat
+	ua[s] = xu >> s;
+	ia[s] = xi >> s;
+#assign s s + 1
+#until s == 16	
+
+	for(int i=0; i<16; i++)
+	{
+		assert(ua[i] == xu >> i);
+		assert(ia[i] == xi >> i);
+	}
+}
+
+void shl16n(unsigned xu, int xi)
+{
+	unsigned	ua[16];
+	int			ia[16];
+#assign s 0
+#repeat
+	ua[s] = xu << s;
+	ia[s] = xi << s;
+#assign s s + 1
+#until s == 16	
+
+	for(int i=0; i<16; i++)
+	{
+		assert(ua[i] == xu << i);
+		assert(ia[i] == xi << i);
+	}
+}
+
+void shr16n(unsigned xu, int xi)
+{
+	unsigned	ua[16];
+	int			ia[16];
+#assign s 0
+#repeat
+	ua[s] = xu >> s;
+	ia[s] = xi >> s;
+#assign s s + 1
+#until s == 16	
+
+	for(int i=0; i<16; i++)
+	{
+		assert(ua[i] == xu >> i);
+		assert(ia[i] == xi >> i);
+	}
+}
+
+#pragma native(shl16n)
+#pragma native(shr16n)
+
 int main(void)
 {
 	for(int i=0; i<32; i++)
 	{
 		printf("1: %.4x : %.4x | %.4x : %.4x\n", shl1b(i), shl1n(i), shr1b(i), shr1n(i));
+		assert(shl1b(i) == shl1n(i));
+		assert(shr1b(i) == shr1n(i));
 	}
 
 	for(int i=0; i<32; i++)
 	{
 		printf("4: %.4x : %.4x | %.4x : %.4x\n", shl4b(i), shl4n(i), shr4b(i), shr4n(i));
+		assert(shl4b(i) == shl4n(i));
+		assert(shr4b(i) == shr4n(i));
 	}
 
 	for(int i=0; i<32; i++)
 	{
 		printf("8: %.4x : %.4x | %.4x : %.4x\n", shl8b(i), shl8n(i), shr8b(i), shr8n(i));
+		assert(shl8b(i) == shl8n(i));
+		assert(shr8b(i) == shr8n(i));
 	}
 
+	shl16b(0x0000, 0x0000);
+	shl16b(0xffff, 0xffff);
+	shl16b(0x1234, 0x1234);
+	shl16b(0xfedc, 0xfedc);
+
+	shr16b(0x0000, 0x0000);
+	shr16b(0xffff, 0xffff);
+	shr16b(0x1234, 0x1234);
+	shr16b(0xfedc, 0xfedc);
+
+	shl16n(0x0000, 0x0000);
+	shl16n(0xffff, 0xffff);
+	shl16n(0x1234, 0x1234);
+	shl16n(0xfedc, 0xfedc);
+
+	shr16n(0x0000, 0x0000);
+	shr16n(0xffff, 0xffff);
+	shr16n(0x1234, 0x1234);
+	shr16n(0xfedc, 0xfedc);
+
 	return 0;
 }

oscar64/NativeCodeGenerator.cpp

@@ -3790,6 +3790,15 @@ void NativeCodeBasicBlock::BinaryOperator(InterCodeProcedure* proc, NativeCodePr
 						mIns.Push(NativeCodeInstruction(ASMIT_ROL, ASMIM_ZERO_PAGE, treg + 1));
 					}
 				}
+				else if (shift >= 8)
+				{
+					mIns.Push(NativeCodeInstruction(ASMIT_LDA, ASMIM_ZERO_PAGE, BC_REG_TMP + proc->mTempOffset[ins->mSTemp[1]]));
+					for (int i = 8; i < shift; i++)
+						mIns.Push(NativeCodeInstruction(ASMIT_ASL, ASMIM_IMPLIED));
+					mIns.Push(NativeCodeInstruction(ASMIT_STA, ASMIM_ZERO_PAGE, treg + 1));
+					mIns.Push(NativeCodeInstruction(ASMIT_LDA, ASMIM_IMMEDIATE, 0x00));
+					mIns.Push(NativeCodeInstruction(ASMIT_STA, ASMIM_ZERO_PAGE, treg + 0));
+				}
 				else
 				{
 					if (ins->mSTemp[1] != ins->mTTemp)
@@ -3900,6 +3909,15 @@ void NativeCodeBasicBlock::BinaryOperator(InterCodeProcedure* proc, NativeCodePr
 						mIns.Push(NativeCodeInstruction(ASMIT_ROR, ASMIM_ZERO_PAGE, treg));
 					}
 				}
+				else if (shift >= 8)
+				{
+					mIns.Push(NativeCodeInstruction(ASMIT_LDA, ASMIM_ZERO_PAGE, BC_REG_TMP + proc->mTempOffset[ins->mSTemp[1]] + 1));
+					for (int i = 8; i < shift; i++)
+						mIns.Push(NativeCodeInstruction(ASMIT_LSR, ASMIM_IMPLIED));
+					mIns.Push(NativeCodeInstruction(ASMIT_STA, ASMIM_ZERO_PAGE, treg));
+					mIns.Push(NativeCodeInstruction(ASMIT_LDA, ASMIM_IMMEDIATE, 0x00));
+					mIns.Push(NativeCodeInstruction(ASMIT_STA, ASMIM_ZERO_PAGE, treg + 1));
+				}
 				else
 				{
 					if (ins->mSTemp[1] != ins->mTTemp)
@@ -3918,7 +3936,7 @@ void NativeCodeBasicBlock::BinaryOperator(InterCodeProcedure* proc, NativeCodePr
 					for (int i = 1; i < shift; i++)
 					{
 						mIns.Push(NativeCodeInstruction(ASMIT_LSR, ASMIM_ZERO_PAGE, treg + 1));
-						mIns.Push(NativeCodeInstruction(ASMIT_ROL, ASMIM_IMPLIED));
+						mIns.Push(NativeCodeInstruction(ASMIT_ROR, ASMIM_IMPLIED));
 					}
 					mIns.Push(NativeCodeInstruction(ASMIT_STA, ASMIM_ZERO_PAGE, treg ));
 				}
@@ -3990,6 +4008,18 @@ void NativeCodeBasicBlock::BinaryOperator(InterCodeProcedure* proc, NativeCodePr
 						mIns.Push(NativeCodeInstruction(ASMIT_STA, ASMIM_ZERO_PAGE, treg + 1));
 					}
 				}
+				else if (shift >= 8)
+				{
+					mIns.Push(NativeCodeInstruction(ASMIT_LDA, ASMIM_ZERO_PAGE, BC_REG_TMP + proc->mTempOffset[ins->mSTemp[1]] + 1));
+					for (int i = 8; i < shift; i++)
+						mIns.Push(NativeCodeInstruction(ASMIT_LSR, ASMIM_IMPLIED));
+					mIns.Push(NativeCodeInstruction(ASMIT_EOR, ASMIM_IMMEDIATE, 0x80 >> (shift - 8)));
+					mIns.Push(NativeCodeInstruction(ASMIT_SEC, ASMIM_IMPLIED));
+					mIns.Push(NativeCodeInstruction(ASMIT_SBC, ASMIM_IMMEDIATE, 0x80 >> (shift - 8)));
+					mIns.Push(NativeCodeInstruction(ASMIT_STA, ASMIM_ZERO_PAGE, treg));
+					mIns.Push(NativeCodeInstruction(ASMIT_SBC, ASMIM_ZERO_PAGE, treg));
+					mIns.Push(NativeCodeInstruction(ASMIT_STA, ASMIM_ZERO_PAGE, treg + 1));
+				}
 				else
 				{
 					if (ins->mSTemp[1] != ins->mTTemp)

oscar64/Preprocessor.cpp

@@ -207,11 +207,13 @@ bool Preprocessor::CloseSource(void)
 
 bool Preprocessor::PushSource(void)
 {
+	mSource->mLocation = mLocation;
 	return mSource->PushSource();
 }
 
 bool Preprocessor::PopSource(void)
 {
+	mLocation = mSource->mLocation;
 	return mSource->PopSource();
 }