Bladeren bron

various small improvements and fixes

kub 5 jaren geleden
bovenliggende
commit
d40a5af495

+ 3 - 3
Makefile

@@ -195,10 +195,10 @@ LDFLAGS += -Wl,-Map=$(TARGET).map
 endif
 
 
-target_: pico/pico_int_o32.h $(TARGET)
+target_: pico/pico_int_offs.h $(TARGET)
 
 clean:
-	$(RM) $(TARGET) $(OBJS) pico/pico_int_o32.h
+	$(RM) $(TARGET) $(OBJS) pico/pico_int_offs.h
 	$(RM) -r .opk_data
 
 $(TARGET): $(OBJS)
@@ -211,7 +211,7 @@ endif
 pprof: platform/linux/pprof.c
 	$(CC) $(CFLAGS) -O2 -ggdb -DPPROF -DPPROF_TOOL -I../../ -I. $^ -o $@ $(LDFLAGS) $(LDLIBS)
 
-pico/pico_int_o32.h:: tools/mkoffsets.sh
+pico/pico_int_offs.h:: tools/mkoffsets.sh
 	make -C tools/ XCC="$(CC)" XCFLAGS="$(CFLAGS)"
 
 .s.o:

+ 3 - 3
config.caanoo

@@ -4,11 +4,11 @@ CC = arm-gph-linux-gnueabi-gcc
 CXX = arm-gph-linux-gnueabi-g++
 AS = arm-gph-linux-gnueabi-as
 STRIP = arm-gph-linux-gnueabi-strip
-CFLAGS += -mfloat-abi=soft -mcpu=arm920t -mtune=arm920t -fno-stack-protector -D__GP2X__ -DGPERF
+CFLAGS += -mfloat-abi=soft -mcpu=arm920t -mtune=arm920t -fno-stack-protector -D__GP2X__
 CFLAGS += -fno-gcse -funswitch-loops -fweb -ftree-loop-im #-fpredictive-commoning -ftree-loop-distribution -frename-registers
-CFLAGS += -I/home/build/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/include -I/home/build/src/gp2x/armroot-eabi/include
+CFLAGS += -I${HOME}/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/include -I${HOME}/src/gp2x/armroot-eabi/include
 ASFLAGS += -mfloat-abi=soft -mcpu=arm920t
-LDFLAGS += -B/home/build/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/lib/gcc/arm-gph-linux-gnueabi/4.2.4 -B/home/build/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/lib -L/home/build/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/lib -L/home/build/src/gp2x/armroot-eabi/lib -static
+LDFLAGS += -B${HOME}/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/lib/gcc/arm-gph-linux-gnueabi/4.2.4 -B${HOME}/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/lib -L${HOME}/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/lib -L${HOME}/src/gp2x/armroot-eabi/lib -static
 LDLIBS += -lpng -lm -ldl
 
 ARCH = arm

+ 2 - 2
config.caanoo47

@@ -6,9 +6,9 @@ AS = arm-linux-gnueabi-as
 STRIP = arm-linux-gnueabi-strip
 CFLAGS += -mfloat-abi=soft -mcpu=arm920t -mtune=arm920t -Wno-unused-result -fno-stack-protector -D__GP2X__
 CFLAGS += -fno-gcse -funswitch-loops -fweb -ftree-loop-im #-fpredictive-commoning -ftree-loop-distribution -frename-registers
-CFLAGS += -I/home/build/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/include -I/home/build/src/gp2x/armroot-eabi/include
+CFLAGS += -I${HOME}/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/include -I${HOME}/src/gp2x/armroot-eabi/include
 ASFLAGS += -mfloat-abi=soft -mcpu=arm920t
-LDFLAGS += -B/home/build/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/lib/gcc/arm-gph-linux-gnueabi/4.2.4 -B/home/build/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/lib -L/home/build/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/lib -static
+LDFLAGS += -B${HOME}/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/lib/gcc/arm-gph-linux-gnueabi/4.2.4 -B${HOME}/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/lib -L${HOME}/opt/GPH_SDK/tools/gcc-4.2.4-glibc-2.7-eabi/arm-gph-linux-gnueabi/sys-root/usr/lib -static
 LDLIBS += -lpng -lm -ldl
 
 ARCH = arm

+ 3 - 3
config.dingux

@@ -4,12 +4,12 @@ CC = mipsel-linux-gcc
 CXX = mipsel-linux-g++
 AS = mipsel-linux-as
 STRIP = mipsel-linux-strip
-CFLAGS +=  -I/home/build/opt/opendingux-toolchain/usr/include/
-CFLAGS +=  -I/home/build/opt/opendingux-toolchain/usr/include/SDL
+CFLAGS +=  -I${HOME}/opt/opendingux-toolchain/usr/include/
+CFLAGS +=  -I${HOME}/opt/opendingux-toolchain/usr/include/SDL
 CFLAGS += -D_GNU_SOURCE=1 -D_REENTRANT -Wno-unused-result -fno-stack-protector
 ASFLAGS += 
 LDFLAGS += 
-LDLIBS += -B/home/build/opt/opendingux-toolchain/usr/lib -Wl,-rpath-link=/home/build/opt/opendingux-toolchain/usr/lib -Wl,-rpath-link=/home/build/opt/opendingux-toolchain/lib -lSDL -lasound -lpng -lm -lstdc++ -ldl
+LDLIBS += -B${HOME}/opt/opendingux-toolchain/usr/lib -Wl,-rpath-link=${HOME}/opt/opendingux-toolchain/usr/lib -Wl,-rpath-link=${HOME}/opt/opendingux-toolchain/lib -lSDL -lasound -lpng -lm -lstdc++ -ldl
 
 ARCH = mipsel
 PLATFORM = opendingux

+ 3 - 3
config.dingux54

@@ -4,12 +4,12 @@ CC = mipsel-linux-gnu-gcc
 CXX = mipsel-linux-gnu-g++
 AS = mipsel-linux-gnu-as
 STRIP = mipsel-linux-gnu-strip
-CFLAGS += -I/home/build/opt/opendingux-toolchain/usr/include/
-CFLAGS += -I/home/build/opt/opendingux-toolchain/usr/include/SDL
+CFLAGS += -I${HOME}/opt/opendingux-toolchain/usr/include/
+CFLAGS += -I${HOME}/opt/opendingux-toolchain/usr/include/SDL
 CFLAGS += -D_GNU_SOURCE=1 -D_REENTRANT -Wno-unused-result -fno-stack-protector
 ASFLAGS += 
 LDFLAGS += 
-LDLIBS += -B/home/build/opt/opendingux-toolchain/usr/lib -B/home/build/opt/opendingux-toolchain/lib -Wl,-rpath-link=/home/build/opt/opendingux-toolchain/usr/lib -Wl,-rpath-link=/home/build/opt/opendingux-toolchain/lib -lSDL -lasound -lpng -lz -lm -lstdc++ -ldl
+LDLIBS += -B${HOME}/opt/opendingux-toolchain/usr/lib -B${HOME}/opt/opendingux-toolchain/lib -Wl,-rpath-link=${HOME}/opt/opendingux-toolchain/usr/lib -Wl,-rpath-link=${HOME}/opt/opendingux-toolchain/lib -lSDL -lasound -lpng -lz -lm -lstdc++ -ldl
 
 ARCH = mipsel
 PLATFORM = opendingux

+ 2 - 2
config.gp2x

@@ -5,10 +5,10 @@ CXX = arm-open2x-linux-g++
 AS = arm-open2x-linux-as
 STRIP = arm-open2x-linux-strip
 CFLAGS += -msoft-float -mcpu=arm920t -mtune=arm920t -D__GP2X__
-CFLAGS += -I/home/build/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/include -I/home/build/src/gp2x/armroot/include
+CFLAGS += -I${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/include -I${HOME}/src/gp2x/armroot/include
 CFLAGS += -fno-gcse -funswitch-loops -fweb -ftree-loop-im #-fpredictive-commoning -ftree-loop-distribution -frename-registers
 ASFLAGS += -mcpu=arm920t -mfloat-abi=soft
-LDFLAGS += --sysroot /home/build/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux -L/home/build/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib -L/home/build/src/gp2x/armroot/lib -static
+LDFLAGS += --sysroot ${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux -L${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib -L${HOME}/src/gp2x/armroot/lib -static
 LDLIBS += -lpng -lm -ldl
 
 ARCH = arm

+ 2 - 2
config.gp2x47

@@ -5,10 +5,10 @@ CXX = arm-linux-gnueabi-g++
 AS = arm-linux-gnueabi-as
 STRIP = arm-linux-gnueabi-strip
 CFLAGS += -mabi=apcs-gnu -mno-thumb-interwork -mfloat-abi=soft -mfpu=fpa -mcpu=arm920t -mtune=arm920t -Wno-unused-result -fno-stack-protector -D__GP2X__
-CFLAGS += -I/home/build/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/include -I/home/build/src/gp2x/armroot/include
+CFLAGS += -I${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/include -I${HOME}/src/gp2x/armroot/include
 CFLAGS += -fno-gcse -funswitch-loops -fweb -ftree-loop-im #-fpredictive-commoning -ftree-loop-distribution -frename-registers
 ASFLAGS += -mabi=apcs-gnu -mfloat-abi=soft -mfpu=fpa -mcpu=arm920t
-LDFLAGS += -mabi=apcs-gnu -mfpu=fpa -B/home/build/opt/open2x/gcc-4.1.1-glibc-2.3.6/lib/gcc/arm-open2x-linux/4.1.1 -B/home/build/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib -L/home/build/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib -L/home/build/src/gp2x/armroot/lib -static
+LDFLAGS += -mabi=apcs-gnu -mfpu=fpa -B${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/lib/gcc/arm-open2x-linux/4.1.1 -B${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib -L${HOME}/opt/open2x/gcc-4.1.1-glibc-2.3.6/arm-open2x-linux/lib -L${HOME}/src/gp2x/armroot/lib -static
 LDLIBS += -lpng -lm -ldl
 
 ARCH = arm

+ 14 - 0
config.i386

@@ -0,0 +1,14 @@
+# Automatically generated by configure
+# Configured with: './configure' '--platform=generic'
+CC = gcc
+CXX = g++
+AS = as
+STRIP = strip
+CFLAGS +=  -I/usr/include/SDL -D_GNU_SOURCE=1 -D_REENTRANT -Wno-unused-result -m32 # -pg
+ASFLAGS += 
+LDFLAGS += -m32 #-pg
+LDLIBS += -L/usr/lib/i386-linux-gnu -L${HOME}/opt/lib32 -lSDL-1.2 -lasound -lpng -lz -lm -ldl
+
+ARCH = i386
+PLATFORM = generic
+SOUND_DRIVERS =  oss alsa sdl

+ 4 - 4
config.x86

@@ -4,11 +4,11 @@ CC = gcc
 CXX = g++
 AS = as
 STRIP = strip
-CFLAGS +=  -I/usr/include/SDL -D_GNU_SOURCE=1 -D_REENTRANT -Wno-unused-result -m32 # -pg
+CFLAGS +=  -I/usr/include/SDL -D_GNU_SOURCE=1 -D_REENTRANT -Wno-unused-result # -pg
 ASFLAGS += 
-LDFLAGS += -m32 #-pg
-LDLIBS += -L/usr/lib/i386-linux-gnu/debug -L/home/build/opt/lib32 -lSDL-1.2 -lasound -lpng -lz -lm -ldl
+LDFLAGS += #-pg
+LDLIBS += -L/usr/lib/x86_64-linux-gnu -lSDL-1.2 -lasound -lpng -lz -lm -ldl
 
-ARCH = x86
+ARCH = x86_64
 PLATFORM = generic
 SOUND_DRIVERS =  oss alsa sdl

+ 1 - 0
cpu/cz80/cz80.c

@@ -14,6 +14,7 @@
 #include "cz80.h"
 
 #if PICODRIVE_HACKS
+#include <pico/pico_int.h>
 #include <pico/memory.h>
 #endif
 

+ 0 - 6
cpu/drc/cmn.h

@@ -1,9 +1,3 @@
-typedef unsigned char  u8;
-typedef signed char    s8;
-typedef unsigned short u16;
-typedef signed short   s16;
-typedef unsigned int   u32;
-typedef signed int     s32;
 
 #define DRC_TCACHE_SIZE         (4*1024*1024)
 

+ 40 - 16
cpu/drc/emit_arm.c

@@ -177,26 +177,25 @@
 #define EOP_C_AM3_REG(cond,u,l,rn,rd,s,h,rm)       EOP_C_AM3(cond,u,0,l,rn,rd,s,h,rm)
 
 /* ldr and str */
-#define EOP_LDR_IMM2(cond,rd,rn,offset_12)  EOP_C_AM2_IMM(cond,1,0,1,rn,rd,offset_12)
-#define EOP_LDRB_IMM2(cond,rd,rn,offset_12) EOP_C_AM2_IMM(cond,1,1,1,rn,rd,offset_12)
+#define EOP_LDR_IMM2(cond,rd,rn,offset_12)  EOP_C_AM2_IMM(cond,(offset_12) >= 0,0,1,rn,rd,abs(offset_12))
+#define EOP_LDRB_IMM2(cond,rd,rn,offset_12) EOP_C_AM2_IMM(cond,(offset_12) >= 0,1,1,rn,rd,abs(offset_12))
 #define EOP_STR_IMM2(cond,rd,rn,offset_12)  EOP_C_AM2_IMM(cond,(offset_12) >= 0,0,0,rn,rd,abs(offset_12))
 
-#define EOP_LDR_IMM(   rd,rn,offset_12) EOP_C_AM2_IMM(A_COND_AL,1,0,1,rn,rd,offset_12)
-#define EOP_LDR_NEGIMM(rd,rn,offset_12) EOP_C_AM2_IMM(A_COND_AL,0,0,1,rn,rd,offset_12)
+#define EOP_LDR_IMM(   rd,rn,offset_12) EOP_C_AM2_IMM(A_COND_AL,(offset_12) >= 0,0,1,rn,rd,abs(offset_12))
 #define EOP_LDR_SIMPLE(rd,rn)           EOP_C_AM2_IMM(A_COND_AL,1,0,1,rn,rd,0)
-#define EOP_STR_IMM(   rd,rn,offset_12) EOP_C_AM2_IMM(A_COND_AL,1,0,0,rn,rd,offset_12)
+#define EOP_STR_IMM(   rd,rn,offset_12) EOP_C_AM2_IMM(A_COND_AL,(offset_12) >= 0,0,0,rn,rd,abs(offset_12))
 #define EOP_STR_SIMPLE(rd,rn)           EOP_C_AM2_IMM(A_COND_AL,1,0,0,rn,rd,0)
 
 #define EOP_LDR_REG_LSL(cond,rd,rn,rm,shift_imm) EOP_C_AM2_REG(cond,1,0,1,rn,rd,shift_imm,A_AM1_LSL,rm)
 #define EOP_LDRB_REG_LSL(cond,rd,rn,rm,shift_imm) EOP_C_AM2_REG(cond,1,1,1,rn,rd,shift_imm,A_AM1_LSL,rm);
 
-#define EOP_LDRH_IMM2(cond,rd,rn,offset_8)  EOP_C_AM3_IMM(cond,1,1,rn,rd,0,1,offset_8)
+#define EOP_LDRH_IMM2(cond,rd,rn,offset_8)  EOP_C_AM3_IMM(cond,(offset_8) >= 0,1,rn,rd,0,1,abs(offset_8))
 #define EOP_LDRH_REG2(cond,rd,rn,rm)        EOP_C_AM3_REG(cond,1,1,rn,rd,0,1,rm)
 
-#define EOP_LDRH_IMM(   rd,rn,offset_8)  EOP_C_AM3_IMM(A_COND_AL,1,1,rn,rd,0,1,offset_8)
+#define EOP_LDRH_IMM(   rd,rn,offset_8)  EOP_C_AM3_IMM(A_COND_AL,(offset_8) >= 0,1,rn,rd,0,1,abs(offset_8))
 #define EOP_LDRH_SIMPLE(rd,rn)           EOP_C_AM3_IMM(A_COND_AL,1,1,rn,rd,0,1,0)
 #define EOP_LDRH_REG(   rd,rn,rm)        EOP_C_AM3_REG(A_COND_AL,1,1,rn,rd,0,1,rm)
-#define EOP_STRH_IMM(   rd,rn,offset_8)  EOP_C_AM3_IMM(A_COND_AL,1,0,rn,rd,0,1,offset_8)
+#define EOP_STRH_IMM(   rd,rn,offset_8)  EOP_C_AM3_IMM(A_COND_AL,(offset_8) >= 0,0,rn,rd,0,1,abs(offset_8))
 #define EOP_STRH_SIMPLE(rd,rn)           EOP_C_AM3_IMM(A_COND_AL,1,0,rn,rd,0,1,0)
 #define EOP_STRH_REG(   rd,rn,rm)        EOP_C_AM3_REG(A_COND_AL,1,0,rn,rd,0,1,rm)
 
@@ -285,11 +284,29 @@ static void emith_op_imm2(int cond, int s, int op, int rd, int rn, unsigned int
 			imm = ~imm;
 			op = A_OP_MVN;
 		}
+#ifdef HAVE_ARMV7
+		for (v = imm, ror2 = 0; v && !(v & 3); v >>= 2)
+			ror2--;
+		if (v >> 8) {
+			/* 2+ insns needed - prefer movw/movt */
+			if (op == A_OP_MVN)
+				imm = ~imm;
+			EOP_MOVW(rd, imm);
+			if (imm & 0xffff0000)
+				EOP_MOVT(rd, imm);
+			return;
+		}
+#endif
 		break;
 
-	case A_OP_EOR:
 	case A_OP_SUB:
 	case A_OP_ADD:
+		// count bits in imm and swap ADD and SUB if more bits 1 than 0
+		if (s == 0 && count_bits(imm) > 16) {
+			imm = -imm;
+			op ^= (A_OP_ADD^A_OP_SUB);
+		}
+	case A_OP_EOR:
 	case A_OP_ORR:
 	case A_OP_BIC:
 		if (s == 0 && imm == 0 && rd == rn)
@@ -412,6 +429,8 @@ static int emith_xbranch(int cond, void *target, int is_call)
 
 #define emith_add_r_r_r_lsl(d, s1, s2, lslimm) \
 	EOP_ADD_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSL,lslimm)
+#define emith_add_r_r_r_lsl_ptr(d, s1, s2, lslimm) \
+	emith_add_r_r_r_lsl(d, s1, s2, lslimm)
 
 #define emith_addf_r_r_r_lsl(d, s1, s2, lslimm) \
 	EOP_ADD_REG(A_COND_AL,1,d,s1,s2,A_AM1_LSL,lslimm)
@@ -483,7 +502,7 @@ static int emith_xbranch(int cond, void *target, int is_call)
 	emith_add_r_r_r(d, d, s)
 
 #define emith_sub_r_r(d, s) \
-	EOP_SUB_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0)
+	emith_sub_r_r_r(d, d, s)
 
 #define emith_adc_r_r(d, s) \
 	EOP_ADC_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0)
@@ -529,6 +548,9 @@ static int emith_xbranch(int cond, void *target, int is_call)
 #define emith_move_r_imm(r, imm) \
 	emith_op_imm(A_COND_AL, 0, A_OP_MOV, r, imm)
 
+#define emith_move_r_ptr_imm(r, imm) \
+	emith_move_r_imm(r, (u32)(imm))
+
 #define emith_add_r_imm(r, imm) \
 	emith_op_imm(A_COND_AL, 0, A_OP_ADD, r, imm)
 
@@ -536,7 +558,7 @@ static int emith_xbranch(int cond, void *target, int is_call)
 	emith_op_imm(A_COND_AL, 0, A_OP_ADC, r, imm)
 
 #define emith_adcf_r_imm(r, imm) \
-	emith_op_imm(A_COND_AL, 1, A_OP_ADC, r, (imm))
+	emith_op_imm(A_COND_AL, 1, A_OP_ADC, r, imm)
 
 #define emith_sub_r_imm(r, imm) \
 	emith_op_imm(A_COND_AL, 0, A_OP_SUB, r, imm)
@@ -610,13 +632,13 @@ static int emith_xbranch(int cond, void *target, int is_call)
 	emith_op_imm2(A_COND_AL, 0, A_OP_SUB, d, s, imm)
 
 #define emith_subf_r_r_imm(d, s, imm) \
-	emith_op_imm2(A_COND_AL, 1, A_OP_SUB, d, s, (imm))
+	emith_op_imm2(A_COND_AL, 1, A_OP_SUB, d, s, imm)
 
 #define emith_or_r_r_imm(d, s, imm) \
-	emith_op_imm2(A_COND_AL, 0, A_OP_ORR, d, s, (imm))
+	emith_op_imm2(A_COND_AL, 0, A_OP_ORR, d, s, imm)
 
 #define emith_eor_r_r_imm(d, s, imm) \
-	emith_op_imm2(A_COND_AL, 0, A_OP_EOR, d, s, (imm))
+	emith_op_imm2(A_COND_AL, 0, A_OP_EOR, d, s, imm)
 
 #define emith_neg_r_r(d, s) \
 	EOP_RSB_IMM(d, s, 0, 0)
@@ -758,7 +780,7 @@ static int emith_xbranch(int cond, void *target, int is_call)
 #define emith_clear_msb_c(cond, d, s, count) { \
 	u32 t; \
 	if ((count) <= 8) { \
-		t = (count) - 8; \
+		t = 8 - (count); \
 		t = (0xff << t) & 0xff; \
 		EOP_C_DOP_IMM(cond,A_OP_BIC,0,s,d,8/2,t); \
 	} else if ((count) >= 24) { \
@@ -880,7 +902,9 @@ static int emith_xbranch(int cond, void *target, int is_call)
 #define emith_sh2_rcall(a, tab, func, mask) { \
 	emith_lsr(mask, a, SH2_READ_SHIFT); \
 	EOP_ADD_REG_LSL(tab, tab, mask, 3); \
-	EOP_LDMIA(tab, (1<<func)|(1<<mask)); \
+	if (func < mask) EOP_LDMIA(tab, (1<<func)|(1<<mask)); /* ldm if possible */ \
+	else {	emith_read_r_r_offs(func, tab, 0); \
+		emith_read_r_r_offs(mask, tab, 4); } \
 	emith_addf_r_r_r(func,func,func); \
 }
 

+ 38 - 30
cpu/drc/emit_x86.c

@@ -194,6 +194,17 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI };
 	} \
 } while (0)
 
+#define emith_add_r_r_r_ptr(d, s1, s2) do { \
+	if (d == s1) { \
+		emith_add_r_r_ptr(d, s2); \
+	} else if (d == s2) { \
+		emith_add_r_r_ptr(d, s1); \
+	} else { \
+		emith_move_r_r_ptr(d, s1); \
+		emith_add_r_r_ptr(d, s2); \
+	} \
+} while (0)
+
 #define emith_sub_r_r_r(d, s1, s2) do { \
 	if (d == s1) { \
 		emith_sub_r_r(d, s2); \
@@ -268,9 +279,16 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI };
 	rcache_free_tmp(tmp_); \
 } while (0)
 
-#define emith_add_r_r_r_lsr(d, s1, s2, lslimm) do { \
+#define emith_add_r_r_r_lsl_ptr(d, s1, s2, lslimm) do { \
 	int tmp_ = rcache_get_tmp(); \
-	emith_lsr(tmp_, s2, lslimm); \
+	emith_lsl(tmp_, s2, lslimm); \
+	emith_add_r_r_r_ptr(d, s1, tmp_); \
+	rcache_free_tmp(tmp_); \
+} while (0)
+
+#define emith_add_r_r_r_lsr(d, s1, s2, lsrimm) do { \
+	int tmp_ = rcache_get_tmp(); \
+	emith_lsr(tmp_, s2, lsrimm); \
 	emith_add_r_r_r(d, s1, tmp_); \
 	rcache_free_tmp(tmp_); \
 } while (0)
@@ -297,6 +315,16 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI };
 	EMIT(imm, u32); \
 } while (0)
 
+#define emith_move_r_ptr_imm(r, imm) do { \
+	if ((uint64_t)(imm) <= UINT32_MAX) \
+		emith_move_r_imm(r, (uintptr_t)(imm)); \
+	else { \
+		EMIT_REX_IF(1, 0, r); \
+		EMIT_OP(0xb8 + (r)); \
+		EMIT((uint64_t)(imm), uint64_t); \
+	} \
+} while (0)
+
 #define emith_move_r_imm_s8(r, imm) \
 	emith_move_r_imm(r, (u32)(signed int)(signed char)(imm))
 
@@ -421,27 +449,28 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI };
 #define emith_sub_r_r_imm(d, s, imm) do { \
 	if (d != s) \
 		emith_move_r_r(d, s); \
-	if (imm) \
+	if ((s32)(imm) != 0) \
 		emith_sub_r_imm(d, imm); \
 } while (0)
 
 #define emith_and_r_r_imm(d, s, imm) do { \
 	if (d != s) \
 		emith_move_r_r(d, s); \
-	emith_and_r_imm(d, imm); \
+	if ((s32)(imm) != -1) \
+		emith_and_r_imm(d, imm); \
 } while (0)
 
 #define emith_or_r_r_imm(d, s, imm) do { \
 	if (d != s) \
 		emith_move_r_r(d, s); \
-	if ((s32)imm != 0) \
+	if ((s32)(imm) != 0) \
 		emith_or_r_imm(d, imm); \
 } while (0)
 
 #define emith_eor_r_r_imm(d, s, imm) do { \
 	if (d != s) \
 		emith_move_r_r(d, s); \
-	if ((s32)imm != 0) \
+	if ((s32)(imm) != 0) \
 		emith_eor_r_imm(d, imm); \
 } while (0)
 
@@ -612,31 +641,17 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI };
 	EMIT_REX_IF(1, r, rs); \
 	emith_deref_op(0x89, r, rs, offs)
 
-// note: don't use prefixes on this
 #define emith_read8_r_r_offs(r, rs, offs) do { \
-	int r_ = r; \
-	if (!is_abcdx(r)) \
-		r_ = rcache_get_tmp(); \
 	EMIT(0x0f, u8); \
-	emith_deref_op(0xb6, r_, rs, offs); \
-	if ((r) != r_) { \
-		emith_move_r_r(r, r_); \
-		rcache_free_tmp(r_); \
-	} \
+	emith_deref_op(0xb6, r, rs, offs); \
 } while (0)
 
 #define emith_read8s_r_r_offs(r, rs, offs) do { \
-	int r_ = r; \
-	if (!is_abcdx(r)) \
-		r_ = rcache_get_tmp(); \
 	EMIT(0x0f, u8); \
-	emith_deref_op(0xbe, r_, rs, offs); \
-	if ((r) != r_) { \
-		emith_move_r_r(r, r_); \
-		rcache_free_tmp(r_); \
-	} \
+	emith_deref_op(0xbe, r, rs, offs); \
 } while (0)
 
+// note: don't use prefixes on this
 #define emith_write8_r_r_offs(r, rs, offs) do {\
 	int r_ = r; \
 	if (!is_abcdx(r)) { \
@@ -664,16 +679,9 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI };
 } while (0)
 
 #define emith_read8_r_r_r(r, rs, rm) do { \
-	int r_ = r; \
-	if (!is_abcdx(r)) \
-		r_ = rcache_get_tmp(); \
 	EMIT(0x0f, u8); \
 	EMIT_OP_MODRM(0xb6, 0, r, 4); \
 	EMIT_SIB(0, rs, rm); /* mov r, [rm + rs * 1] */ \
-	if ((r) != r_) { \
-		emith_move_r_r(r, r_); \
-		rcache_free_tmp(r_); \
-	} \
 } while (0)
 
 #define emith_read16_r_r_r(r, rs, rm) do { \

+ 140 - 54
cpu/sh2/compiler.c

@@ -56,9 +56,10 @@
 // 04 - asm
 // 08 - runtime block entry log
 // 10 - smc self-check
+// 20 - runtime block entry counter
 // 100 - write trace
 // 200 - compare trace
-// 400 - block entry backtraceA on exit
+// 400 - block entry backtrace on exit
 // 800 - state dump on exit
 // {
 #ifndef DRC_DEBUG
@@ -178,7 +179,7 @@ static char sh2dasm_buff[64];
 }
 
 #if (DRC_DEBUG & (8|256|512|1024)) || defined(PDB)
-static SH2 csh2[2][4];
+static SH2 csh2[2][8];
 static void REGPARM(3) *sh2_drc_log_entry(void *block, SH2 *sh2, u32 sr)
 {
   if (block != NULL) {
@@ -190,7 +191,6 @@ static void REGPARM(3) *sh2_drc_log_entry(void *block, SH2 *sh2, u32 sr)
   {
     static FILE *trace[2];
     int idx = sh2->is_slave;
-if (sh2 != &sh2s[0] && sh2 != &sh2s[1]) printf("sh2 %p?\n",sh2);
     if (!trace[0]) {
       truncate("pico.trace", 0);
       trace[0] = fopen("pico.trace0", "wb");
@@ -199,7 +199,8 @@ if (sh2 != &sh2s[0] && sh2 != &sh2s[1]) printf("sh2 %p?\n",sh2);
     if (csh2[idx][0].pc != sh2->pc) {
       fwrite(sh2, offsetof(SH2, read8_map), 1, trace[idx]);
       fwrite(&sh2->pdb_io_csum, sizeof(sh2->pdb_io_csum), 1, trace[idx]);
-      memcpy(&csh2[idx][0], sh2, offsetof(SH2, icount));
+      memcpy(&csh2[idx][0], sh2, offsetof(SH2, poll_cnt)+4);
+      csh2[idx][0].is_slave = idx;
     }
   }
 #elif (DRC_DEBUG & 512)
@@ -234,9 +235,10 @@ if (sh2 != &sh2s[0] && sh2 != &sh2s[1]) printf("sh2 %p?\n",sh2);
 #elif (DRC_DEBUG & 1024)
   {
     int x = sh2->is_slave, i;
-    for (i = 0; i < ARRAY_SIZE(csh2[x]); i++)
-      memcpy(&csh2[x][i], &csh2[x][i+1], offsetof(SH2, icount));
-    memcpy(&csh2[x][3], sh2, offsetof(SH2, icount));
+    for (i = 0; i < ARRAY_SIZE(csh2[x])-1; i++)
+      memcpy(&csh2[x][i], &csh2[x][i+1], offsetof(SH2, poll_cnt)+4);
+    memcpy(&csh2[x][ARRAY_SIZE(csh2[x])-1], sh2, offsetof(SH2, poll_cnt)+4);
+    csh2[x][0].is_slave = x;
   }
 #endif
   }
@@ -252,9 +254,9 @@ if (sh2 != &sh2s[0] && sh2 != &sh2s[1]) printf("sh2 %p?\n",sh2);
 // and can be discarded early
 // XXX: need to tune sizes
 static const int tcache_sizes[TCACHE_BUFFERS] = {
-  DRC_TCACHE_SIZE * 6 / 8, // ROM (rarely used), DRAM
-  DRC_TCACHE_SIZE / 8, // BIOS, data array in master sh2
-  DRC_TCACHE_SIZE / 8, // ... slave
+  DRC_TCACHE_SIZE * 14 / 16, // ROM (rarely used), DRAM
+  DRC_TCACHE_SIZE / 16, // BIOS, data array in master sh2
+  DRC_TCACHE_SIZE / 16, // ... slave
 };
 
 static u8 *tcache_bases[TCACHE_BUFFERS];
@@ -287,6 +289,9 @@ struct block_entry {
 #if (DRC_DEBUG & 2)
   struct block_desc *block;
 #endif
+#if (DRC_DEBUG & 32)
+  int entry_count;
+#endif
 };
 
 struct block_desc {
@@ -698,6 +703,14 @@ static void add_to_hashlist(struct block_entry *be, int tcache_id)
     (*head)->prev = be;
   be->next = *head;
   *head = be;
+
+#if (DRC_DEBUG & 2)
+  if (be->next != NULL) {
+    printf(" %08x: entry hash collision with %08x\n",
+      be->pc, be->next->pc);
+    hash_collisions++;
+  }
+#endif
 }
 
 static void rm_from_hashlist(struct block_entry *be, int tcache_id)
@@ -727,6 +740,14 @@ static void add_to_hashlist_unresolved(struct block_link *bl, int tcache_id)
   u32 tcmask = hash_table_sizes[tcache_id] - 1;
   struct block_link **head = &HASH_FUNC(unresolved_links[tcache_id], bl->target_pc, tcmask);
 
+#if DRC_DEBUG & 1
+  struct block_link *current = *head;
+  while (current != NULL && current != bl)
+    current = current->next;
+  if (current == bl)
+    dbg(1, "add_to_hashlist_unresolved @%p: bl %p %p %08x already in?", head, bl, bl->target, bl->target_pc);
+#endif
+
   bl->target = NULL; // marker for not resolved
   bl->prev = NULL;
   if (*head)
@@ -745,7 +766,7 @@ static void rm_from_hashlist_unresolved(struct block_link *bl, int tcache_id)
   while (current->prev != NULL)
     current = current->prev;
   if (current != *head)
-    dbg(1, "rm_from_hashlist unresolved @%p: bl %p %p %08x missing?", head, bl, bl->target, bl->target_pc);
+    dbg(1, "rm_from_hashlist_unresolved @%p: bl %p %p %08x missing?", head, bl, bl->target, bl->target_pc);
 #endif
 
   if (bl->prev != NULL)
@@ -980,10 +1001,12 @@ static void *dr_prepare_ext_branch(struct block_entry *owner, u32 pc, int is_sla
   struct block_entry *be = NULL;
   int target_tcache_id;
 
+  // get the target block entry
   be = dr_get_entry(pc, is_slave, &target_tcache_id);
   if (target_tcache_id && target_tcache_id != tcache_id)
     return sh2_drc_dispatcher;
 
+  // get a block link
   if (blink_free[tcache_id] != NULL) {
     bl = blink_free[tcache_id];
     blink_free[tcache_id] = bl->next;
@@ -995,6 +1018,7 @@ static void *dr_prepare_ext_branch(struct block_entry *owner, u32 pc, int is_sla
     block_link_pool_counts[tcache_id] = cnt+1;
   }
 
+  // prepare link and add to ougoing list of owner
   bl->tcache_id = tcache_id;
   bl->target_pc = pc;
   bl->jump = tcache_ptr;
@@ -1940,6 +1964,7 @@ static void rcache_invalidate(void)
       cache_regs[i].type = HR_FREE;
     cache_regs[i].gregs = 0;
   }
+
   for (i = 0; i < ARRAY_SIZE(guest_regs); i++) {
     guest_regs[i].flags &= GRF_STATIC;
     if (!(guest_regs[i].flags & GRF_STATIC))
@@ -1953,7 +1978,8 @@ static void rcache_invalidate(void)
       cache_regs[guest_regs[i].sreg].gregs = 1 << i;
       guest_regs[i].vreg = guest_regs[i].sreg;
     }
-  };
+  }
+
   rcache_counter = 0;
   rcache_hint_soon = rcache_hint_late = 0;
 
@@ -2005,6 +2031,7 @@ static int emit_get_rbase_and_offs(SH2 *sh2, u32 a, u32 *offs)
   u32 mask = 0;
   int poffs;
   int hr;
+  unsigned long la;
 
   poffs = dr_ctx_get_mem_ptr(a, &mask);
   if (poffs == -1)
@@ -2014,15 +2041,16 @@ static int emit_get_rbase_and_offs(SH2 *sh2, u32 a, u32 *offs)
   if (mask < 0x1000) {
     // can't access data array or BIOS directly from ROM or SDRAM,
     // since code may run on both SH2s (tcache_id of translation block needed))
-    emith_ctx_read(hr, poffs);
+    emith_ctx_read_ptr(hr, poffs);
     if (a & mask & ~omask)
-      emith_add_r_imm(hr, a & mask & ~omask);
+      emith_add_r_r_ptr_imm(hr, hr, a & mask & ~omask);
+    *offs = a & omask;
   } else {
     // known fixed host address
-    a = (a & mask) + *(u32 *)((char *)sh2 + poffs);
-    emith_move_r_imm(hr, (a & ~omask));
+    la = (unsigned long)*(void **)((char *)sh2 + poffs) + (a & mask);
+    *offs = la & omask;
+    emith_move_r_ptr_imm(hr, la & ~omask);
   }
-  *offs = a & omask;
   return hr;
 }
 
@@ -2392,8 +2420,6 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
   void *branch_patch_ptr[MAX_LOCAL_BRANCHES];
   u32 branch_patch_pc[MAX_LOCAL_BRANCHES];
   int branch_patch_count = 0;
-  u32 literal_addr[MAX_LITERALS];
-  int literal_addr_count = 0;
   u8 op_flags[BLOCK_INSN_LIMIT];
   struct {
     u32 test_irq:1;
@@ -2473,7 +2499,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
   {
     u32 delay_dep_fw = 0, delay_dep_bk = 0;
     int tmp3, tmp4;
-    u32 sr;
+    int sr;
 
     opd = &ops[i];
     op = FETCH_OP(pc);
@@ -2487,7 +2513,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       pc, op, sh2dasm_buff);
 #endif
 
-    if ((op_flags[i] & OF_BTARGET) || pc == base_pc)
+    if (op_flags[i] & OF_BTARGET)
     {
       if (pc != base_pc)
       {
@@ -2517,6 +2543,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
         else {
           dbg(1, "too many entryp for block #%d,%d pc=%08x",
             tcache_id, blkid_main, pc);
+          break;
         }
       } else {
         entry = block->entryp;
@@ -2537,10 +2564,10 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
 
 #if (DRC_DEBUG & 0x10)
       rcache_get_reg_arg(0, SHR_PC, NULL);
-      tmp = emit_memhandler_read(2);
+      tmp = emit_memhandler_read(1);
       tmp2 = rcache_get_tmp();
       tmp3 = rcache_get_tmp();
-      emith_move_r_imm(tmp2, FETCH32(pc));
+      emith_move_r_imm(tmp2, (s16)FETCH_OP(pc));
       emith_move_r_imm(tmp3, 0);
       emith_cmp_r_r(tmp, tmp2);
       EMITH_SJMP_START(DCOND_EQ);
@@ -2556,9 +2583,20 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       emith_cmp_r_imm(sr, 0);
       emith_jump_cond(DCOND_LE, sh2_drc_exit);
 
+#if (DRC_DEBUG & 32)
+      // block hit counter
+      tmp  = rcache_get_tmp_arg(0);
+      tmp2 = rcache_get_tmp_arg(1);
+      emith_move_r_ptr_imm(tmp, (uptr)entry);
+      emith_read_r_r_offs(tmp2, tmp, offsetof(struct block_entry, entry_count));
+      emith_add_r_imm(tmp2, 1);
+      emith_write_r_r_offs(tmp2, tmp, offsetof(struct block_entry, entry_count));
+      rcache_free_tmp(tmp);
+      rcache_free_tmp(tmp2);
+#endif
+
 #if (DRC_DEBUG & (8|256|512|1024))
       sr = rcache_get_reg(SHR_SR, RC_GR_RMW, NULL);
-      FLUSH_CYCLES(sr);
       rcache_clean();
       tmp = rcache_used_hreg_mask();
       emith_save_caller_regs(tmp);
@@ -2566,7 +2604,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       rcache_get_reg_arg(2, SHR_SR, NULL);
       tmp2 = rcache_get_tmp_arg(0);
       tmp3 = rcache_get_tmp_arg(1);
-      emith_move_r_imm(tmp2, (u32)tcache_ptr);
+      emith_move_r_ptr_imm(tmp2, tcache_ptr);
       emith_move_r_r_ptr(tmp3,CONTEXT_REG);
       emith_call(sh2_drc_log_entry);
       emith_restore_caller_regs(tmp);
@@ -2776,7 +2814,6 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       if ((opd->imm && opd->imm >= base_pc && opd->imm < end_literals) ||
           dr_is_rom(opd->imm))
       {
-        ADD_TO_ARRAY(literal_addr, literal_addr_count, opd->imm,);
         if (opd->size == 2)
           u = FETCH32(opd->imm);
         else
@@ -2862,8 +2899,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       case 0x06: // MOV.L Rm,@(R0,Rn)   0000nnnnmmmm0110
         emit_indirect_indexed_write(sh2, GET_Rm(), SHR_R0, GET_Rn(), op & 3);
         goto end_op;
-      case 0x07:
-        // MUL.L     Rm,Rn      0000nnnnmmmm0111
+      case 0x07: // MUL.L     Rm,Rn      0000nnnnmmmm0111
         tmp  = rcache_get_reg(GET_Rn(), RC_GR_READ, NULL);
         tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ, NULL);
         tmp3 = rcache_get_reg(SHR_MACL, RC_GR_WRITE, NULL);
@@ -2941,8 +2977,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       goto default_;
 
     /////////////////////////////////////////////
-    case 0x01:
-      // MOV.L Rm,@(disp,Rn) 0001nnnnmmmmdddd
+    case 0x01: // MOV.L Rm,@(disp,Rn) 0001nnnnmmmmdddd
       emit_memhandler_write_rr(sh2, GET_Rm(), GET_Rn(), (op & 0x0f) * 4, 2);
       goto end_op;
 
@@ -3346,19 +3381,16 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       case 0x09:
         switch (GET_Fx())
         {
-        case 0:
-          // SHLL2 Rn        0100nnnn00001000
-          // SHLR2 Rn        0100nnnn00001001
+        case 0: // SHLL2 Rn        0100nnnn00001000
+                // SHLR2 Rn        0100nnnn00001001
           tmp = 2;
           break;
-        case 1:
-          // SHLL8 Rn        0100nnnn00011000
-          // SHLR8 Rn        0100nnnn00011001
+        case 1: // SHLL8 Rn        0100nnnn00011000
+                // SHLR8 Rn        0100nnnn00011001
           tmp = 8;
           break;
-        case 2:
-          // SHLL16 Rn       0100nnnn00101000
-          // SHLR16 Rn       0100nnnn00101001
+        case 2: // SHLL16 Rn       0100nnnn00101000
+                // SHLR16 Rn       0100nnnn00101001
           tmp = 16;
           break;
         default:
@@ -3432,8 +3464,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
         } else
           emit_move_r_r(tmp2, GET_Rn());
         goto end_op;
-      case 0x0f:
-        // MAC.W @Rm+,@Rn+  0100nnnnmmmm1111
+      case 0x0f: // MAC.W @Rm+,@Rn+  0100nnnnmmmm1111
         emit_indirect_read_double(sh2, &tmp, &tmp2, GET_Rn(), GET_Rm(), 1);
         sr = rcache_get_reg(SHR_SR, RC_GR_READ, NULL);
         tmp3 = rcache_get_reg(SHR_MACL, RC_GR_RMW, NULL);
@@ -3446,8 +3477,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       goto default_;
 
     /////////////////////////////////////////////
-    case 0x05:
-      // MOV.L @(disp,Rm),Rn 0101nnnnmmmmdddd
+    case 0x05: // MOV.L @(disp,Rm),Rn 0101nnnnmmmmdddd
       emit_memhandler_read_rr(sh2, GET_Rn(), GET_Rm(), (op & 0x0f) * 4, 2);
       goto end_op;
 
@@ -3519,8 +3549,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       goto default_;
 
     /////////////////////////////////////////////
-    case 0x07:
-      // ADD #imm,Rn  0111nnnniiiiiiii
+    case 0x07: // ADD #imm,Rn  0111nnnniiiiiiii
       tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW, &tmp2);
       if (op & 0x80) { // adding negative
         emith_sub_r_r_imm(tmp, tmp2, -op & 0xff);
@@ -3621,8 +3650,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       goto default_;
 
     /////////////////////////////////////////////
-    case 0x0e:
-      // MOV #imm,Rn   1110nnnniiiiiiii
+    case 0x0e: // MOV #imm,Rn   1110nnnniiiiiiii
       emit_move_r_imm32(GET_Rn(), (s8)op);
       goto end_op;
 
@@ -3886,9 +3914,7 @@ static void sh2_generate_utils(void)
 #if BRANCH_CACHE
   // check if PC is in branch target cache
   emith_and_r_r_imm(arg1, arg0, (ARRAY_SIZE(sh2s->branch_cache)-1)*4);
-  // TODO implement emith_add_r_r_r_lsl_ptr, saves one insn on 32bit ARM
-  emith_lsl(arg1, arg1, sizeof(void *) == 8 ? 2 : 1);
-  emith_add_r_r_ptr(arg1, CONTEXT_REG);
+  emith_add_r_r_r_lsl_ptr(arg1, CONTEXT_REG, arg1, sizeof(void *) == 8 ? 2 : 1);
   emith_read_r_r_offs(arg2, arg1, offsetof(SH2, branch_cache));
   emith_cmp_r_r(arg2, arg0);
   EMITH_SJMP_START(DCOND_NE);
@@ -3905,8 +3931,7 @@ static void sh2_generate_utils(void)
   EMITH_SJMP_START(DCOND_EQ);
   emith_ctx_read_c(DCOND_NE, arg2, SHR_PC * 4);
   emith_and_r_r_imm(arg1, arg2, (ARRAY_SIZE(sh2s->branch_cache)-1)*4);
-  emith_lsl(arg1, arg1, sizeof(void *) == 8 ? 2 : 1);
-  emith_add_r_r_ptr(arg1, CONTEXT_REG);
+  emith_add_r_r_r_lsl_ptr(arg1, CONTEXT_REG, arg1, sizeof(void *) == 8 ? 2 : 1);
   emith_write_r_r_offs_c(DCOND_NE, arg2, arg1, offsetof(SH2, branch_cache));
   emith_write_r_r_offs_ptr_c(DCOND_NE, RET_REG, arg1, offsetof(SH2, branch_cache) + sizeof(void *));
   EMITH_SJMP_END(DCOND_EQ);
@@ -4174,7 +4199,8 @@ int sh2_execute_drc(SH2 *sh2c, int cycles)
 static void block_stats(void)
 {
 #if (DRC_DEBUG & 2)
-  int c, b, i, total = 0;
+  int c, b, i;
+  long total = 0;
 
   printf("block stats:\n");
   for (b = 0; b < ARRAY_SIZE(block_tables); b++) {
@@ -4185,8 +4211,9 @@ static void block_stats(void)
       if (block_tables[b][i].addr != 0)
         total += block_tables[b][i].refcount;
   }
+  printf("total: %ld\n",total);
 
-  for (c = 0; c < 10; c++) {
+  for (c = 0; c < 20; c++) {
     struct block_desc *blk, *maxb = NULL;
     int max = 0;
     for (b = 0; b < ARRAY_SIZE(block_tables); b++) {
@@ -4221,6 +4248,63 @@ static void block_stats(void)
 #endif
 }
 
+void entry_stats(void)
+{
+#if (DRC_DEBUG & 32)
+  int c, b, i, j;
+  long total = 0;
+
+  printf("block entry stats:\n");
+  for (b = 0; b < ARRAY_SIZE(block_tables); b++) {
+    for (i = 0; i < block_counts[b]; i++)
+      for (j = 0; j < block_tables[b][i].entry_count; j++)
+        total += block_tables[b][i].entryp[j].entry_count;
+    for (i = block_limit[b]; i < block_max_counts[b]; i++)
+      for (j = 0; j < block_tables[b][i].entry_count; j++)
+        total += block_tables[b][i].entryp[j].entry_count;
+  }
+  printf("total: %ld\n",total);
+
+  for (c = 0; c < 20; c++) {
+    struct block_desc *blk;
+    struct block_entry *maxb = NULL;
+    int max = 0;
+    for (b = 0; b < ARRAY_SIZE(block_tables); b++) {
+      for (i = 0; i < block_counts[b]; i++) {
+        blk = &block_tables[b][i];
+        for (j = 0; j < blk->entry_count; j++)
+          if (blk->entryp[j].entry_count > max) {
+            max = blk->entryp[j].entry_count;
+            maxb = &blk->entryp[j];
+          }
+      }
+      for (i = block_limit[b]; i < block_max_counts[b]; i++) {
+        blk = &block_tables[b][i];
+        for (j = 0; j < blk->entry_count; j++)
+          if (blk->entryp[j].entry_count > max) {
+            max = blk->entryp[j].entry_count;
+            maxb = &blk->entryp[j];
+          }
+      }
+    }
+    if (maxb == NULL)
+      break;
+    printf("%08x %p %9d %2.3f%%\n", maxb->pc, maxb->tcache_ptr, maxb->entry_count,
+      (double)100 * maxb->entry_count / total);
+    maxb->entry_count = 0;
+  }
+
+  for (b = 0; b < ARRAY_SIZE(block_tables); b++) {
+    for (i = 0; i < block_counts[b]; i++)
+      for (j = 0; j < block_tables[b][i].entry_count; j++)
+        block_tables[b][i].entryp[j].entry_count = 0;
+    for (i = block_limit[b]; i < block_max_counts[b]; i++)
+      for (j = 0; j < block_tables[b][i].entry_count; j++)
+        block_tables[b][i].entryp[j].entry_count = 0;
+  }
+#endif
+}
+
 static void backtrace(void)
 {
 #if (DRC_DEBUG & 1024)
@@ -4279,6 +4363,7 @@ void sh2_drc_flush_all(void)
   backtrace();
   state_dump();
   block_stats();
+  entry_stats();
   flush_tcache(0);
   flush_tcache(1);
   flush_tcache(2);
@@ -4364,6 +4449,7 @@ int sh2_drc_init(SH2 *sh2)
     hash_collisions = 0;
 #endif
   }
+  memset(sh2->branch_cache, -1, sizeof(sh2->branch_cache));
 
   return 0;
 

+ 1 - 1
cpu/sh2/mame/sh2pico.c

@@ -214,7 +214,7 @@ int sh2_execute_interpreter(SH2 *sh2, int cycles)
 			if (sh2->pc < *base_pc || sh2->pc >= *end_pc) {
 				*base_pc = sh2->pc;
 				scan_block(*base_pc, sh2->is_slave,
-					op_flags, end_pc, NULL);
+					op_flags, end_pc, NULL, NULL);
 			}
 			if ((op_flags[(sh2->pc - *base_pc) / 2]
 				& OF_BTARGET) || sh2->pc == *base_pc

+ 2 - 2
cpu/sh2/sh2.h

@@ -81,9 +81,9 @@ typedef struct SH2_
 
 #define CYCLE_MULT_SHIFT 10
 #define C_M68K_TO_SH2(xsh2, c) \
-	((int)((long long)(c) * (xsh2)->mult_m68k_to_sh2) >> CYCLE_MULT_SHIFT)
+	(int)(((unsigned long long)(c) * (xsh2)->mult_m68k_to_sh2) >> CYCLE_MULT_SHIFT)
 #define C_SH2_TO_M68K(xsh2, c) \
-	((int)((long long)(c+3) * (xsh2)->mult_sh2_to_m68k) >> CYCLE_MULT_SHIFT)
+	(int)(((unsigned long long)(c+3U) * (xsh2)->mult_sh2_to_m68k) >> CYCLE_MULT_SHIFT)
 
 int  sh2_init(SH2 *sh2, int is_slave, SH2 *other_sh2);
 void sh2_finish(SH2 *sh2);

+ 14 - 14
pico/32x/32x.c

@@ -30,7 +30,7 @@ static int REGPARM(2) sh2_irq_cb(SH2 *sh2, int level)
 }
 
 // MUST specify active_sh2 when called from sh2 memhandlers
-void p32x_update_irls(SH2 *active_sh2, int m68k_cycles)
+void p32x_update_irls(SH2 *active_sh2, unsigned int m68k_cycles)
 {
   int irqs, mlvl = 0, slvl = 0;
   int mrun, srun;
@@ -50,18 +50,18 @@ void p32x_update_irls(SH2 *active_sh2, int m68k_cycles)
     slvl++;
   slvl *= 2;
 
-  mrun = sh2_irl_irq(&msh2, mlvl, active_sh2 == &msh2);
+  mrun = sh2_irl_irq(&msh2, mlvl, msh2.state & SH2_STATE_RUN);
   if (mrun) {
     p32x_sh2_poll_event(&msh2, SH2_IDLE_STATES, m68k_cycles);
-    if (active_sh2 == &msh2)
-      sh2_end_run(active_sh2, 1);
+    if (msh2.state & SH2_STATE_RUN)
+      sh2_end_run(&msh2, 1);
   }
 
-  srun = sh2_irl_irq(&ssh2, slvl, active_sh2 == &ssh2);
+  srun = sh2_irl_irq(&ssh2, slvl, ssh2.state & SH2_STATE_RUN);
   if (srun) {
     p32x_sh2_poll_event(&ssh2, SH2_IDLE_STATES, m68k_cycles);
-    if (active_sh2 == &ssh2)
-      sh2_end_run(active_sh2, 1);
+    if (ssh2.state & SH2_STATE_RUN)
+      sh2_end_run(&ssh2, 1);
   }
 
   elprintf(EL_32X, "update_irls: m %d/%d, s %d/%d", mlvl, mrun, slvl, srun);
@@ -70,7 +70,7 @@ void p32x_update_irls(SH2 *active_sh2, int m68k_cycles)
 // the mask register is inconsistent, CMD is supposed to be a mask,
 // while others are actually irq trigger enables?
 // TODO: test on hw..
-void p32x_trigger_irq(SH2 *sh2, int m68k_cycles, unsigned int mask)
+void p32x_trigger_irq(SH2 *sh2, unsigned int m68k_cycles, unsigned int mask)
 {
   Pico32x.sh2irqs |= mask & P32XI_VRES;
   Pico32x.sh2irqi[0] |= mask & (Pico32x.sh2irq_mask[0] << 3);
@@ -79,7 +79,7 @@ void p32x_trigger_irq(SH2 *sh2, int m68k_cycles, unsigned int mask)
   p32x_update_irls(sh2, m68k_cycles);
 }
 
-void p32x_update_cmd_irq(SH2 *sh2, int m68k_cycles)
+void p32x_update_cmd_irq(SH2 *sh2, unsigned int m68k_cycles)
 {
   if ((Pico32x.sh2irq_mask[0] & 2) && (Pico32x.regs[2 / 2] & 1))
     Pico32x.sh2irqi[0] |= P32XI_CMD;
@@ -207,8 +207,8 @@ void PicoReset32x(void)
 {
   if (PicoIn.AHW & PAHW_32X) {
     p32x_trigger_irq(NULL, SekCyclesDone(), P32XI_VRES);
-    p32x_sh2_poll_event(&msh2, SH2_IDLE_STATES, 0);
-    p32x_sh2_poll_event(&ssh2, SH2_IDLE_STATES, 0);
+    p32x_sh2_poll_event(&msh2, SH2_IDLE_STATES, SekCyclesDone());
+    p32x_sh2_poll_event(&ssh2, SH2_IDLE_STATES, SekCyclesDone());
     p32x_pwm_ctl_changed();
     p32x_timers_recalc();
   }
@@ -258,7 +258,7 @@ static void p32x_start_blank(void)
   p32x_sh2_poll_event(&ssh2, SH2_STATE_VPOLL, SekCyclesDone());
 }
 
-void p32x_schedule_hint(SH2 *sh2, int m68k_cycles)
+void p32x_schedule_hint(SH2 *sh2, unsigned int m68k_cycles)
 {
   // rather rough, 32x hint is useless in practice
   int after;
@@ -370,9 +370,9 @@ static void p32x_run_events(unsigned int until)
       oldest, event_time_next);
 }
 
-static void run_sh2(SH2 *sh2, int m68k_cycles)
+static void run_sh2(SH2 *sh2, unsigned int m68k_cycles)
 {
-  int cycles, done;
+  unsigned int cycles, done;
 
   pevt_log_sh2_o(sh2, EVT_RUN_START);
   sh2->state |= SH2_STATE_RUN;

+ 11 - 9
pico/32x/draw_arm.S

@@ -6,7 +6,7 @@
 @* See COPYING file in the top-level directory.
 @*
 
-#include "pico/pico_int_o32.h"
+#include "pico/pico_int_offs.h"
 
 .extern Pico32x
 .extern Pico
@@ -74,7 +74,7 @@ Pico32xNativePal:
     ldr     lr,=Pico
     ldr     r10,=Pico32x
     ldr     r11, [lr, #OFS_Pico_est+OFS_EST_Draw2FB]
-    ldr     r10,[r10, #0x40] @ Pico32x.vdp_regs[0]
+    ldrh    r10,[r10, #0x40] @ Pico32x.vdp_regs[0]
     add     r9, lr, #OFS_Pico_est+OFS_EST_HighPal   @ palmd
 
     and     r4, r2, #0xff
@@ -118,6 +118,8 @@ Pico32xNativePal:
     mov     r7, r7, lsl #1
     ldreqh  r12,[r9, r7]
     streqh  r12,[r0], #2     @ *dst++ = palmd[*pmd]
+.else
+    addeq   r0, r0, #2
 .endif
     beq     2b @ loop_inner
 
@@ -182,8 +184,8 @@ Pico32xNativePal:
     ldrneb  r8, [r5, #2]!    @ r7,r8 - pixel 0,1 index
     subs    r6, r6, #1
     blt     0b @ loop_outer
-@    cmp     r7, r8            @ is this really improving things?
-@    beq     5f @ check_fill @ +8
+    cmp     r7, r8            @ is this really improving things?
+    beq     5f @ check_fill @ +8
 
 3: @ no_fill:
     mov     r12,r7, lsl #1
@@ -242,7 +244,7 @@ Pico32xNativePal:
     beq     6b
 
 7: @ count_done
-    sub     r5, r5, #4      @ undo readahead
+    sub     r5, r5, #4       @ undo readahead
 
     @ fix alignment and check type
     sub     r8, r5, lr
@@ -268,14 +270,14 @@ Pico32xNativePal:
     b       2b @ loop_inner
 
 9: @ bg_mode:
-    ldrb    r12,[r11],#1     @ MD pixel
+    ldrb    r12,[r11],#1     @ MD pixel 0,1
     ldrb    lr, [r11],#1
-    cmp     r3, lr, lsl #26  @ MD has bg pixel?
+    cmp     r3, r12,lsl #26  @ MD pixel 0 has bg?
 .if \do_md
     mov     r12,r12,lsl #1
     ldrneh  r12,[r9, r12]    @ t = palmd[*pmd]
     moveq   r12,r7
-    cmp     r3, lr, lsl #26
+    cmp     r3, lr, lsl #26  @ MD pixel 1 has bg?
     mov     lr, lr, lsl #1
     ldrneh  lr, [r9, lr]
     moveq   lr, r7
@@ -283,7 +285,7 @@ Pico32xNativePal:
     strh    lr, [r0], #2
 .else
     streqh  r7, [r0]
-    cmp     r3, lr, lsl #26
+    cmp     r3, lr, lsl #26  @ MD pixel 1 has bg?
     streqh  r7, [r0, #2]
     add     r0, r0, #4
 .endif

+ 4 - 4
pico/32x/memory.c

@@ -398,9 +398,6 @@ static void p32x_reg_write8(u32 a, u32 d)
     p32x_sh2_poll_event(&sh2s[1], SH2_STATE_CPOLL, cycles);
     comreg = 1 << (a & 0x0f) / 2;
     Pico32x.comm_dirty |= comreg;
-
-    if (cycles - (int)msh2.m68krcycles_done > 120)
-      p32x_sync_sh2s(cycles);
     return;
   }
 }
@@ -453,6 +450,9 @@ static void p32x_reg_write16(u32 a, u32 d)
     int cycles = SekCyclesDone();
     int comreg;
     
+    if (r[a / 2] == d)
+       return;
+
     p32x_sync_sh2s(cycles);
 
     r[a / 2] = d;
@@ -685,7 +685,7 @@ static void p32x_sh2reg_write8(u32 a, u32 d, SH2 *sh2)
     case 0x3f:
       return;
     pwm_write:
-      p32x_pwm_write16(a & ~1, d, sh2, 0);
+      p32x_pwm_write16(a & ~1, d, sh2, sh2_cycles_done_m68k(sh2));
       return;
   }
 

+ 38 - 38
pico/32x/memory_arm.S

@@ -6,7 +6,7 @@
  * See COPYING file in the top-level directory.
  */
 
-#include "../pico_int_o32.h"
+#include "../pico_int_offs.h"
 
 @ 32X bank sizes... TODO this should somehow come from an include file
 .equ SH2_ROM_SHIFT, 10      @ 0x003fffff
@@ -46,92 +46,92 @@
 sh2_read8_rom:
     ldr     ip, [r1, #OFS_SH2_p_rom]
     eor     r0, r0, #1
-    lsl     r0, #SH2_ROM_SHIFT
+    mov     r0, r0, lsl #SH2_ROM_SHIFT
     ldrb    r0, [ip, r0, lsr #SH2_ROM_SHIFT]
     bx      lr
 
 sh2_read8_sdram:
     ldr     ip, [r1, #OFS_SH2_p_sdram]
     eor     r0, r0, #1
-    lsl     r0, #SH2_RAM_SHIFT
+    mov     r0, r0, lsl #SH2_RAM_SHIFT
     ldrb    r0, [ip, r0, lsr #SH2_RAM_SHIFT]
     bx      lr
 
 sh2_read8_da:
     ldr     ip, [r1, #OFS_SH2_p_da]
     eor     r0, r0, #1
-    lsl     r0, #SH2_DA_SHIFT
+    mov     r0, r0, lsl #SH2_DA_SHIFT
     ldrb    r0, [ip, r0, lsr #SH2_DA_SHIFT]
     bx      lr
 
 sh2_read8_dram:
     ldr     ip, [r1, #OFS_SH2_p_dram]
     eor     r0, r0, #1
-    lsl     r0, #SH2_DRAM_SHIFT
+    mov     r0, r0, lsl #SH2_DRAM_SHIFT
     ldrb    r0, [ip, r0, lsr #SH2_DRAM_SHIFT]
     bx      lr
 
 sh2_read16_rom:
     ldr     ip, [r1, #OFS_SH2_p_rom]
-    lsl     r0, #SH2_ROM_SHIFT
-    lsr     r0, #SH2_ROM_SHIFT
+    mov     r0, r0, lsl #SH2_ROM_SHIFT
+    mov     r0, r0, lsr #SH2_ROM_SHIFT
     ldrh    r0, [ip, r0]
     bx      lr
 
 sh2_read16_sdram:
     ldr     ip, [r1, #OFS_SH2_p_sdram]
-    lsl     r0, #SH2_RAM_SHIFT
-    lsr     r0, #SH2_RAM_SHIFT
+    mov     r0, r0, lsl #SH2_RAM_SHIFT
+    mov     r0, r0, lsr #SH2_RAM_SHIFT
     ldrh    r0, [ip, r0]
     bx      lr
 
 sh2_read16_da:
     ldr     ip, [r1, #OFS_SH2_p_da]
-    lsl     r0, #SH2_DA_SHIFT
-    lsr     r0, #SH2_DA_SHIFT
+    mov     r0, r0, lsl #SH2_DA_SHIFT
+    mov     r0, r0, lsr #SH2_DA_SHIFT
     ldrh    r0, [ip, r0]
     bx      lr
 
 sh2_read16_dram:
     ldr     ip, [r1, #OFS_SH2_p_dram]
-    lsl     r0, #SH2_DRAM_SHIFT
-    lsr     r0, #SH2_DRAM_SHIFT
+    mov     r0, r0, lsl #SH2_DRAM_SHIFT
+    mov     r0, r0, lsr #SH2_DRAM_SHIFT
     ldrh    r0, [ip, r0]
     bx      lr
 
 sh2_read32_rom:
     ldr     ip, [r1, #OFS_SH2_p_rom]
-    lsl     r0, #SH2_ROM_SHIFT
+    mov     r0, r0, lsl #SH2_ROM_SHIFT
     ldr     r0, [ip, r0, lsr #SH2_ROM_SHIFT]
-    ror     r0, r0, #16
+    mov     r0, r0, ror #16
     bx      lr
 
 sh2_read32_sdram:
     ldr     ip, [r1, #OFS_SH2_p_sdram]
-    lsl     r0, #SH2_RAM_SHIFT
+    mov     r0, r0, lsl #SH2_RAM_SHIFT
     ldr     r0, [ip, r0, lsr #SH2_RAM_SHIFT]
-    ror     r0, r0, #16
+    mov     r0, r0, ror #16
     bx      lr
 
 sh2_read32_da:
     ldr     ip, [r1, #OFS_SH2_p_da]
-    lsl     r0, #SH2_DA_SHIFT
+    mov     r0, r0, lsl #SH2_DA_SHIFT
     ldr     r0, [ip, r0, lsr #SH2_DA_SHIFT]
-    ror     r0, r0, #16
+    mov     r0, r0, ror #16
     bx      lr
 
 sh2_read32_dram:
     ldr     ip, [r1, #OFS_SH2_p_dram]
-    lsl     r0, #SH2_DRAM_SHIFT
+    mov     r0, r0, lsl #SH2_DRAM_SHIFT
     ldr     r0, [ip, r0, lsr #SH2_DRAM_SHIFT]
-    ror     r0, r0, #16
+    mov     r0, r0, ror #16
     bx      lr
 
 sh2_write8_sdram:
     @ preserve r0 and r2 for tail call
     ldr     ip, [r2, #OFS_SH2_p_sdram]
     eor     r3, r0, #1
-    lsl     r3, #SH2_RAM_SHIFT
+    mov     r3, r3, lsl #SH2_RAM_SHIFT
     strb    r1, [ip, r3, lsr #SH2_RAM_SHIFT]
 #ifdef DRC_SH2
     ldr     ip, [r2, #OFS_SH2_p_drcblk_ram]
@@ -148,7 +148,7 @@ sh2_write8_da:
     @ preserve r0 and r2 for tail call
     ldr     ip, [r2, #OFS_SH2_p_da]
     eor     r3, r0, #1
-    lsl     r3, #SH2_DA_SHIFT
+    mov     r3, r3, lsl #SH2_DA_SHIFT
     strb    r1, [ip, r3, lsr #SH2_DA_SHIFT]
 #ifdef DRC_SH2
     ldr     ip, [r2, #OFS_SH2_p_drcblk_da]
@@ -165,15 +165,15 @@ sh2_write8_dram:
     tst     r1, #0xff
     ldrne   ip, [r2, #OFS_SH2_p_dram]
     eorne   r3, r0, #1
-    lslne   r3, #SH2_DRAM_SHIFT
+    movne   r3, r3, lsl #SH2_DRAM_SHIFT
     strneb  r1, [ip, r3, lsr #SH2_DRAM_SHIFT]
     bx      lr
 
 sh2_write16_sdram:
     @ preserve r0 and r2 for tail call
     ldr     ip, [r2, #OFS_SH2_p_sdram]
-    lsl     r3, r0, #SH2_RAM_SHIFT
-    lsr     r3, r3, #SH2_RAM_SHIFT
+    mov     r3, r0, lsl #SH2_RAM_SHIFT
+    mov     r3, r3, lsr #SH2_RAM_SHIFT
     strh    r1, [ip, r3]
 #ifdef DRC_SH2
     ldr     ip, [r2, #OFS_SH2_p_drcblk_ram]
@@ -188,8 +188,8 @@ sh2_write16_sdram:
 sh2_write16_da:
     @ preserve r0 and r2 for tail call
     ldr     ip, [r2, #OFS_SH2_p_da]
-    lsl     r3, r0, #SH2_DA_SHIFT
-    lsr     r3, r3, #SH2_DA_SHIFT
+    mov     r3, r0, lsl #SH2_DA_SHIFT
+    mov     r3, r3, lsr #SH2_DA_SHIFT
     strh    r1, [ip, r3]
 #ifdef DRC_SH2
     ldr     ip, [r2, #OFS_SH2_p_drcblk_da]
@@ -204,23 +204,23 @@ sh2_write16_da:
 sh2_write16_dram:
     ldr     ip, [r2, #OFS_SH2_p_dram]
     tst     r0, #SH2_DRAM_OW
-    lsl     r3, r0, #SH2_DRAM_SHIFT
-    lsr     r3, r3, #SH2_DRAM_SHIFT
+    mov     r3, r0, lsl #SH2_DRAM_SHIFT
+    mov     r3, r3, lsr #SH2_DRAM_SHIFT
     streqh  r1, [ip, r3]
     bxeq    lr
     add     ip, ip, r3
     tst     r1, #0xff
     strneb  r1, [ip, #0]
     tst     r1, #0xff00
-    lsrne   r1, r1, #8
+    movne   r1, r1, lsr #8
     strneb  r1, [ip, #1]
     bx      lr
 
 sh2_write32_sdram:
     @ preserve r0 and r2 for tail call
     ldr     ip, [r2, #OFS_SH2_p_sdram]
-    ror     r1, r1, #16
-    lsl     r3, r0, #SH2_RAM_SHIFT
+    mov     r1, r1, ror #16
+    mov     r3, r0, lsl #SH2_RAM_SHIFT
     str     r1, [ip, r3, lsr #SH2_RAM_SHIFT]
 #ifdef DRC_SH2
     ldr     ip, [r2, #OFS_SH2_p_drcblk_ram]
@@ -242,8 +242,8 @@ sh2_write32_sdram:
 sh2_write32_da:
     @ preserve r0 and r2 for tail call
     ldr     ip, [r2, #OFS_SH2_p_da]
-    ror     r1, r1, #16
-    lsl     r3, r0, #SH2_DA_SHIFT
+    mov     r1, r1, ror #16
+    mov     r3, r0, lsl #SH2_DA_SHIFT
     str     r1, [ip, r3, lsr #SH2_DA_SHIFT]
 #ifdef DRC_SH2
     ldr     ip, [r2, #OFS_SH2_p_drcblk_da]
@@ -265,13 +265,13 @@ sh2_write32_da:
 sh2_write32_dram:
     ldr     ip, [r2, #OFS_SH2_p_dram]
     tst     r0, #SH2_DRAM_OW
-    lsl     r3, r0, #SH2_DRAM_SHIFT
-    roreq   r1, r1, #16
+    mov     r3, r0, lsl #SH2_DRAM_SHIFT
+    moveq   r1, r1, ror #16
     streq   r1, [ip, r3, lsr #SH2_DRAM_SHIFT]
     bxeq    lr
 #if 1
     ldr     r0, [ip, r3, lsr #SH2_DRAM_SHIFT]
-    ror     r1, r1, #16
+    mov     r1, r1, ror #16
     mov     r2, #0
     tst     r1, #0x00ff0000
     orrne   r2, r2, #0x00ff0000

+ 0 - 4
pico/cd/gfx_dma.c

@@ -10,10 +10,6 @@
 
 #include "cell_map.c"
 
-#ifndef UTYPES_DEFINED
-typedef unsigned short u16;
-#endif
-
 // check: Heart of the alien, jaguar xj 220
 PICO_INTERNAL void DmaSlowCell(unsigned int source, unsigned int a, int len, unsigned char inc)
 {

+ 1 - 1
pico/cd/memory_arm.S

@@ -6,7 +6,7 @@
 @* See COPYING file in the top-level directory.
 @*
 
-#include "../pico_int_o32.h"
+#include "../pico_int_offs.h"
 
 .equiv PCM_STEP_SHIFT, 11
 

+ 1 - 1
pico/draw2_arm.S

@@ -8,7 +8,7 @@
  * this is highly specialized, be careful if changing related C code!
  */
 
-#include "pico_int_o32.h"
+#include "pico_int_offs.h"
 
 @ define these constants in your include file:
 @ .equiv START_ROW, 		1

+ 1 - 1
pico/draw_arm.S

@@ -8,7 +8,7 @@
  * this is highly specialized, be careful if changing related C code!
  */
 
-#include "pico_int_o32.h"
+#include "pico_int_offs.h"
 
 .extern DrawStripInterlace
 

+ 0 - 5
pico/memory.h

@@ -2,11 +2,6 @@
 
 #include "pico_port.h"
 
-typedef unsigned char  u8;
-typedef unsigned short u16;
-typedef unsigned int   u32;
-typedef uintptr_t      uptr; // unsigned pointer-sized int
-
 #define M68K_MEM_SHIFT 16
 // minimum size we can map
 #define M68K_BANK_SIZE (1 << M68K_MEM_SHIFT)

+ 1 - 1
pico/memory_amips.S

@@ -8,7 +8,7 @@
 
 # OUT OF DATE
 
-#include "pico_int_o32.h"
+#include "pico_int_offs.h"
 
 .set noreorder
 .set noat

+ 1 - 1
pico/memory_arm.S

@@ -6,7 +6,7 @@
  * See COPYING file in the top-level directory.
  */
 
-#include "pico_int_o32.h"
+#include "pico_int_offs.h"
 
 .equ SRR_MAPPED,    (1 <<  0)
 .equ SRR_READONLY,  (1 <<  1)

+ 13 - 5
pico/pico_int.h

@@ -33,6 +33,14 @@ extern "C" {
 #endif
 
 
+typedef unsigned char  u8;
+typedef signed char    s8;
+typedef unsigned short u16;
+typedef signed short   s16;
+typedef unsigned int   u32;
+typedef signed int     s32;
+typedef uintptr_t      uptr; // unsigned pointer-sized int
+
 // ----------------------- 68000 CPU -----------------------
 #ifdef EMU_C68K
 #include "../cpu/cyclone/Cyclone.h"
@@ -427,7 +435,7 @@ struct PicoSound
   short psg_line;
 };
 
-// run tools/mkoffsets pico/pico_int_o32.h if you change these
+// run tools/mkoffsets pico/pico_int_offs.h if you change these
 // careful with savestate compat
 struct Pico
 {
@@ -905,13 +913,13 @@ void PicoFrame32x(void);
 void Pico32xStateLoaded(int is_early);
 void p32x_sync_sh2s(unsigned int m68k_target);
 void p32x_sync_other_sh2(SH2 *sh2, unsigned int m68k_target);
-void p32x_update_irls(SH2 *active_sh2, int m68k_cycles);
-void p32x_trigger_irq(SH2 *sh2, int m68k_cycles, unsigned int mask);
-void p32x_update_cmd_irq(SH2 *sh2, int m68k_cycles);
+void p32x_update_irls(SH2 *active_sh2, unsigned int m68k_cycles);
+void p32x_trigger_irq(SH2 *sh2, unsigned int m68k_cycles, unsigned int mask);
+void p32x_update_cmd_irq(SH2 *sh2, unsigned int m68k_cycles);
 void p32x_reset_sh2s(void);
 void p32x_event_schedule(unsigned int now, enum p32x_event event, int after);
 void p32x_event_schedule_sh2(SH2 *sh2, enum p32x_event event, int after);
-void p32x_schedule_hint(SH2 *sh2, int m68k_cycles);
+void p32x_schedule_hint(SH2 *sh2, unsigned int m68k_cycles);
 
 // 32x/memory.c
 extern struct Pico32xMem *Pico32xMem;

+ 1 - 0
platform/common/common.mak

@@ -9,6 +9,7 @@ asm_render = 0
 asm_ym2612 = 0
 asm_misc = 0
 asm_cdmemory = 0
+asm_32xdraw = 0
 asm_mix = 0
 endif
 

+ 23 - 14
platform/common/memcpy.c

@@ -9,7 +9,7 @@
  *   to avoid under/overstepping the src region).
  *
  * ATTN does dirty aliasing tricks with undefined behaviour by standard.
- *	(however, this was needed to improve the generated code).
+ *	(however, this improved the generated code).
  * ATTN uses struct assignment, which only works if the compiler is inlining
  *	this (else it would probably call memcpy :-)).
  */
@@ -33,22 +33,24 @@ void *memcpy(void *dest, const void *src, size_t n)
 	const int lm = sizeof(uint32_t)-1;
 
 	/* align src to word */
-	while (((unsigned)ss.c & lm) && n > 0)
+	while (((uintptr_t)ss.c & lm) && n > 0)
 		*ds.c++ = *ss.c++, n--;
-	if (((unsigned)ds.c & lm) == 0) {
+	if (((uintptr_t)ds.c & lm) == 0) {
 		/* fast copy if pointers have the same aligment */
-		while (n >= sizeof(struct _16))	/* copy 16 bytes blocks */
+		while (n >= sizeof(struct _16))	/* copy 16 byte blocks */
 			*ds.s++ = *ss.s++, n -= sizeof(struct _16);
 		if (n >= sizeof(uint64_t))	/* copy leftover 8 byte block */
 			*ds.l++ = *ss.l++, n -= sizeof(uint64_t);
+//		if (n >= sizeof(uint32_t))	/* copy leftover 4 byte block */
+//			*ds.i++ = *ss.i++, n -= sizeof(uint32_t);
 	} else if (n >= 2*sizeof(uint32_t)) {
 		/* unaligned data big enough to avoid overstepping src */
 		uint32_t v1, v2, b, s;
 		/* align dest to word */
-		while (((unsigned)ds.c & lm) && n > 0)
+		while (((uintptr_t)ds.c & lm) && n > 0)
 			*ds.c++ = *ss.c++, n--;
 		/* copy loop: load aligned words and store shifted words */
-		b = (unsigned)ss.c & lm, s = b*8; ss.c -= b;
+		b = (uintptr_t)ss.c & lm, s = b*8; ss.c -= b;
 		v1 = *ss.i++, v2 = *ss.i++;
 		while (n >= 3*sizeof(uint32_t)) {
 			*ds.i++ = (v1 _L_ s) | (v2 _U_ (32-s)); v1 = *ss.i++;
@@ -78,28 +80,35 @@ void *memmove (void *dest, const void *src, size_t n)
 	struct _16 { uint32_t a[4]; };
 	union { const void *v; uint8_t *c; uint32_t *i; uint64_t *l; struct _16 *s; }
 		ss = { src+n }, ds = { dest+n };
+	size_t pd = dest > src ? dest - src : src - dest;
 	const int lm = sizeof(uint32_t)-1;
 
 	if (dest <= src || dest >= src+n)
 		return memcpy(dest, src, n);
 
 	/* align src to word */
-	while (((unsigned)ss.c & lm) && n > 0)
+	while (((uintptr_t)ss.c & lm) && n > 0)
 		*--ds.c = *--ss.c, n--;
-	if (((unsigned)ds.c & lm) == 0) {
+	/* take care not to copy multi-byte data if it overlaps */
+	if (((uintptr_t)ds.c & lm) == 0) {
 		/* fast copy if pointers have the same aligment */
-		while (n >= sizeof(struct _16))	/* copy 16 byte blocks */
+		while (n >= sizeof(struct _16) && pd >= sizeof(struct _16))
+			/* copy 16 bytes blocks if no overlap */
 			*--ds.s = *--ss.s, n -= sizeof(struct _16);
-		if (n >= sizeof(uint64_t))	/* copy leftover 8 byte block */
+		while (n >= sizeof(uint64_t) && pd >= sizeof(uint64_t))
+			/* copy leftover 8 byte blocks if no overlap */
 			*--ds.l = *--ss.l, n -= sizeof(uint64_t);
-	} else if (n >= 2*sizeof(uint32_t)) {
+		while (n >= sizeof(uint32_t) && pd >= sizeof(uint32_t))
+			/* copy leftover 4 byte blocks if no overlap */
+			*--ds.i = *--ss.i, n -= sizeof(uint32_t);
+	} else if (n >= 2*sizeof(uint32_t) && pd >= 2*sizeof(uint32_t)) {
 		/* unaligned data big enough to avoid understepping src */
 		uint32_t v1, v2, b, s;
 		/* align dest to word */
-		while (((unsigned)ds.c & lm) && n > 0)
+		while (((uintptr_t)ds.c & lm) && n > 0)
 			*--ds.c = *--ss.c, n--;
 		/* copy loop: load aligned words and store shifted words */
-		b = (unsigned)ss.c & lm, s = b*8; ss.c += b;
+		b = (uintptr_t)ss.c & lm, s = b*8; ss.c += b;
 		v1 = *--ss.i, v2 = *--ss.i;
 		while (n >= 3*sizeof(uint32_t)) {
 			*--ds.i = (v1 _U_ s) | (v2 _L_ (32-s)); v1 = *--ss.i;
@@ -114,7 +123,7 @@ void *memmove (void *dest, const void *src, size_t n)
 		}
 		ss.c -= b - 2*sizeof(uint32_t);
 	}
-	/* copy 0-7 leftover bytes */
+	/* copy 0-7 leftover bytes (or upto everything if ptrs are too close) */
 	while (n >= 4) {
 		*--ds.c = *--ss.c, n--; *--ds.c = *--ss.c, n--;
 		*--ds.c = *--ss.c, n--; *--ds.c = *--ss.c, n--;

+ 4 - 8
platform/gp2x/code940/memcpy.s

@@ -114,14 +114,12 @@ subs r2, r2, #0x14
 blt Lmemcpy_fl32  /* less than 32 bytes (12 from above) */
 stmdb sp!, {r4, r7, r8, r9, r10}  /* borrow r4 */
 
-/* blat 64 bytes at a time */
+/* blat 32 bytes at a time */
 /* XXX for really big copies perhaps we should use more registers */
 Lmemcpy_floop32:
 ldmia r1!, {r3, r4, r7, r8, r9, r10, r12, lr}
 stmia r0!, {r3, r4, r7, r8, r9, r10, r12, lr}
-ldmia r1!, {r3, r4, r7, r8, r9, r10, r12, lr}
-stmia r0!, {r3, r4, r7, r8, r9, r10, r12, lr}
-subs r2, r2, #0x40        
+subs r2, r2, #0x20        
 bge Lmemcpy_floop32
 
 cmn r2, #0x10
@@ -314,14 +312,12 @@ stmdb sp!, {r4, r7, r8, r9, r10, lr}
 subs r2, r2, #0x14  /* less than 32 bytes (12 from above) */
 blt Lmemcpy_bl32
 
-/* blat 64 bytes at a time */
+/* blat 32 bytes at a time */
 /* XXX for really big copies perhaps we should use more registers */
 Lmemcpy_bloop32:
 ldmdb r1!, {r3, r4, r7, r8, r9, r10, r12, lr}
 stmdb r0!, {r3, r4, r7, r8, r9, r10, r12, lr}
-ldmdb r1!, {r3, r4, r7, r8, r9, r10, r12, lr}
-stmdb r0!, {r3, r4, r7, r8, r9, r10, r12, lr}
-subs r2, r2, #0x40        
+subs r2, r2, #0x20        
 bge Lmemcpy_bloop32
 
 Lmemcpy_bl32:

+ 1 - 4
tools/mkoffsets.sh

@@ -49,11 +49,8 @@ get_define () # prefix struct member member...
 echo "const int one = 1;" >/tmp/getoffs.c
 compile_rodata
 ENDIAN=$(if [ "$rodata" -eq 1 ]; then echo be; else echo le; fi)
-# determine output file
-echo "const int vsz = sizeof(void *);" >/tmp/getoffs.c
-compile_rodata
-fn="${1:-.}/pico_int_o$((8*$rodata)).h"
 # output header
+fn="${1:-.}/pico_int_offs.h"
 echo "/* autogenerated by mkoffset.sh, do not edit */" >$fn
 echo "/* target endianess: $ENDIAN, compiled with: $CC $CFLAGS */" >>$fn
 # output offsets