[DYNAREC] Added a few more opcodes, some arm64_printer opcodes, and made sure it works

17 files changed, 540 insertions, 112 deletions

diff --git a/src/dynarec/arm64_emitter.h b/src/dynarec/arm64_emitter.h
index 84d9cdaf..52b58dfc 100755
--- a/src/dynarec/arm64_emitter.h
+++ b/src/dynarec/arm64_emitter.h
@@ -124,12 +124,12 @@
 #define ADDSw_REG(Rd, Rn, Rm)               EMIT(ADDSUB_REG_gen(0, 0, 1, 0b00, Rm, 0, Rn, Rd))
 #define ADDxw_REG(Rd, Rn, Rm)               EMIT(ADDSUB_REG_gen(rex.w, 0, 0, 0b00, Rm, 0, Rn, Rd))
 
-#define ADDSUB_IMM_gen(sf, op, S, shift, imm12, Rn, Rd)    ((sf)<<31 | (op)<<30 | (S)<<29 | 0b10001<<24 | 1<<23 | (shift)<<22 | (imm12)<<10 | (Rn)<<5 | (Rd))
-#define ADDx_U12(Rd, Rn, imm12)     EMIT(ADDSUB_IMM_gen(1, 0, 0, 0, (imm12)&0xfff, Rn, Rd))
-#define ADDSx_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(1, 0, 0, 0, (imm12)&0xfff, Rn, Rd))
-#define ADDw_U12(Rd, Rn, imm12)     EMIT(ADDSUB_IMM_gen(0, 0, 0, 0, (imm12)&0xfff, Rn, Rd))
-#define ADDSw_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(0, 0, 0, 0, (imm12)&0xfff, Rn, Rd))
-#define ADDxw_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(rex.w, 0, 0, 0, (imm12)&0xfff, Rn, Rd))
+#define ADDSUB_IMM_gen(sf, op, S, shift, imm12, Rn, Rd)    ((sf)<<31 | (op)<<30 | (S)<<29 | 0b10001<<24 | (shift)<<22 | (imm12)<<10 | (Rn)<<5 | (Rd))
+#define ADDx_U12(Rd, Rn, imm12)     EMIT(ADDSUB_IMM_gen(1, 0, 0, 0b00, (imm12)&0xfff, Rn, Rd))
+#define ADDSx_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(1, 0, 1, 0b00, (imm12)&0xfff, Rn, Rd))
+#define ADDw_U12(Rd, Rn, imm12)     EMIT(ADDSUB_IMM_gen(0, 0, 0, 0b00, (imm12)&0xfff, Rn, Rd))
+#define ADDSw_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(0, 0, 1, 0b00, (imm12)&0xfff, Rn, Rd))
+#define ADDxw_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(rex.w, 0, 0, 0b00, (imm12)&0xfff, Rn, Rd))
 
 #define SUBx_REG(Rd, Rn, Rm)                EMIT(ADDSUB_REG_gen(1, 1, 0, 0b00, Rm, 0, Rn, Rd))
 #define SUBSx_REG(Rd, Rn, Rm)               EMIT(ADDSUB_REG_gen(1, 1, 1, 0b00, Rm, 0, Rn, Rd))
@@ -138,27 +138,33 @@
 #define SUBSw_REG(Rd, Rn, Rm)               EMIT(ADDSUB_REG_gen(0, 1, 1, 0b00, Rm, 0, Rn, Rd))
 #define SUBxw_REG(Rd, Rn, Rm)               EMIT(ADDSUB_REG_gen(rex.w, 1, 0, 0b00, Rm, 0, Rn, Rd))
 #define SUBSxw_REG(Rd, Rn, Rm)              EMIT(ADDSUB_REG_gen(rex.w, 1, 1, 0b00, Rm, 0, Rn, Rd))
-
-#define SUBx_U12(Rd, Rn, imm12)     EMIT(ADDSUB_IMM_gen(1, 1, 0, 0, (imm12)&0xfff, Rn, Rd))
-#define SUBSx_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(1, 1, 0, 0, (imm12)&0xfff, Rn, Rd))
-#define SUBw_U12(Rd, Rn, imm12)     EMIT(ADDSUB_IMM_gen(0, 1, 0, 0, (imm12)&0xfff, Rn, Rd))
-#define SUBSw_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(0, 1, 0, 0, (imm12)&0xfff, Rn, Rd))
-#define SUBxw_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(rex.w, 1, 0, 0, (imm12)&0xfff, Rn, Rd))
-#define SUBSxw_U12(Rd, Rn, imm12)   EMIT(ADDSUB_IMM_gen(rex.w, 1, 0, 0, (imm12)&0xfff, Rn, Rd))
+#define CMPSx_REG(Rn, Rm)                   SUBSx_REG(xZR, Rn, Rm)
+#define CMPSw_REG(Rn, Rm)                   SUBSw_REG(wZR, Rn, Rm)
+#define CMPSxw_REG(Rn, Rm)                  SUBSxw_REG(xZR, Rn, Rm)
+
+#define SUBx_U12(Rd, Rn, imm12)     EMIT(ADDSUB_IMM_gen(1, 1, 0, 0b00, (imm12)&0xfff, Rn, Rd))
+#define SUBSx_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(1, 1, 1, 0b00, (imm12)&0xfff, Rn, Rd))
+#define SUBw_U12(Rd, Rn, imm12)     EMIT(ADDSUB_IMM_gen(0, 1, 0, 0b00, (imm12)&0xfff, Rn, Rd))
+#define SUBSw_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(0, 1, 1, 0b00, (imm12)&0xfff, Rn, Rd))
+#define SUBxw_U12(Rd, Rn, imm12)    EMIT(ADDSUB_IMM_gen(rex.w, 1, 0, 0b00, (imm12)&0xfff, Rn, Rd))
+#define SUBSxw_U12(Rd, Rn, imm12)   EMIT(ADDSUB_IMM_gen(rex.w, 1, 1, 0b00, (imm12)&0xfff, Rn, Rd))
+#define CMPSx_U12(Rn, imm12)        SUBSx_U12(xZR, Rn, imm12)
+#define CMPSw_U12(Rn, imm12)        SUBSw_U12(wZR, Rn, imm12)
+#define CMPSxw_U12(Rn, imm12)       SUBSxw_U12(xZR, Rn, imm12)
 
 // LDR
-#define LDR_gen(size, op1, imm9, op2, Rn, Rt)    ((size)<<30 | 0b111<<27 | (op1)<<24 | 0b01<<22 | (imm9)<<12 | (op2)<<10 | 0b01<<10 | (Rn)<<5 | (Rt))
+#define LDR_gen(size, op1, imm9, op2, Rn, Rt)    ((size)<<30 | 0b111<<27 | (op1)<<24 | 0b01<<22 | (imm9)<<12 | (op2)<<10 | (Rn)<<5 | (Rt))
 #define LDRx_S9_postindex(Rt, Rn, imm9)   EMIT(LDR_gen(0b11, 0b00, (imm9)&0x1ff, 0b01, Rn, Rt))
 #define LDRx_S9_preindex(Rt, Rn, imm9)    EMIT(LDR_gen(0b11, 0b00, (imm9)&0x1ff, 0b11, Rn, Rt))
 #define LDRw_S9_postindex(Rt, Rn, imm9)   EMIT(LDR_gen(0b10, 0b00, (imm9)&0x1ff, 0b01, Rn, Rt))
 #define LDRw_S9_preindex(Rt, Rn, imm9)    EMIT(LDR_gen(0b10, 0b00, (imm9)&0x1ff, 0b11, Rn, Rt))
 
-#define LD_gen(size, op1, imm12, Rn, Rt)        ((size)<<30 | 0b111<<27 | (op1)<<24 | 0b01<<22 | (imm12)<<10 | 0b01<<10 | (Rn)<<5 | (Rt))
-#define LDRx_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b11, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
-#define LDRw_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b10, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
+#define LD_gen(size, op1, imm12, Rn, Rt)        ((size)<<30 | 0b111<<27 | (op1)<<24 | 0b01<<22 | (imm12)<<10 | (Rn)<<5 | (Rt))
+#define LDRx_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b11, 0b01, ((uint32_t)(imm12>>3))&0xfff, Rn, Rt))
+#define LDRw_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b10, 0b01, ((uint32_t)(imm12>>2))&0xfff, Rn, Rt))
 #define LDRB_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b00, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
-#define LDRH_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b01, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
-#define LDRxw_U12(Rt, Rn, imm12)          EMIT(LD_gen((rex.w)?0b11:0b10, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
+#define LDRH_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b01, 0b01, ((uint32_t)(imm12>>1))&0xfff, Rn, Rt))
+#define LDRxw_U12(Rt, Rn, imm12)          EMIT(LD_gen((rex.w)?0b11:0b10, 0b01, ((uint32_t)(imm12>>(2+rex.w)))&0xfff, Rn, Rt))
 
 #define LDR_REG_gen(size, Rm, option, S, Rn, Rt)    ((size)<<30 | 0b111<<27 | 0b01<<22 | 1<<21 | (Rm)<<16 | (option)<<13 | (S)<<12 | (0b10)<<10 | (Rn)<<5 | (Rt))
 #define LDRx_REG(Rt, Rn, Rm)            EMIT(LDR_REG_gen(0b11, Rm, 0b011, 0, Rn, Rt))
@@ -169,23 +175,26 @@
 #define LDRB_REG(Rt, Rn, Rm)            EMIT(LDR_REG_gen(0b00, Rm, 0b011, 0, Rn, Rt))
 #define LDRH_REG(Rt, Rn, Rm)            EMIT(LDR_REG_gen(0b01, Rm, 0b011, 0, Rn, Rt))
 
-#define LDRSH_gen(size, op1, opc, imm9, op2, Rn, Rt)    ((size)<<30 | 0b111<<27 | (op1)<<24 | (opc)<<22 | (imm9)<<12 | (op2)<<10 | 0b01<<10 | (Rn)<<5 | (Rt))
-#define LDRSHx_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b01, 0b00, 0b10, ((uint32_t)(imm12))&0xfff, Rn, Rt))
-#define LDRSHw_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b01, 0b00, 0b11, ((uint32_t)(imm12))&0xfff, Rn, Rt))
+#define LDRSH_gen(size, op1, opc, imm9, op2, Rn, Rt)    ((size)<<30 | 0b111<<27 | (op1)<<24 | (opc)<<22 | (imm9)<<12 | (op2)<<10 | (Rn)<<5 | (Rt))
+#define LDRSHx_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b01, 0b00, 0b10, ((uint32_t)(imm12>>2))&0xfff, Rn, Rt))
+#define LDRSHw_U12(Rt, Rn, imm12)           EMIT(LD_gen(0b01, 0b00, 0b11, ((uint32_t)(imm12>2))&0xfff, Rn, Rt))
+
+#define LDR_PC_gen(opc, imm19, Rt)      ((opc)<<30 | 0b011<<27 | (imm19)<<5 | (Rt))
+#define LDRx_literal(Rt, imm19)         EMIT(LDR_PC_gen(0b01, ((imm19)>>2)&0x7FFFF, Rt))
 
 // STR
-#define STR_gen(size, op1, imm9, op2, Rn, Rt)    ((size)<<30 | 0b111<<27 | (op1)<<24 | 0b00<<22 | (imm9)<<12 | (op2)<<10 | 0b01<<10 | (Rn)<<5 | (Rt))
+#define STR_gen(size, op1, imm9, op2, Rn, Rt)    ((size)<<30 | 0b111<<27 | (op1)<<24 | 0b00<<22 | (imm9)<<12 | (op2)<<10 | (Rn)<<5 | (Rt))
 #define STRx_S9_postindex(Rt, Rn, imm9)   EMIT(STR_gen(0b11, 0b00, (imm9)&0x1ff, 0b01, Rn, Rt))
 #define STRx_S9_preindex(Rt, Rn, imm9)    EMIT(STR_gen(0b11, 0b00, (imm9)&0x1ff, 0b11, Rn, Rt))
 #define STRw_S9_postindex(Rt, Rn, imm9)   EMIT(STR_gen(0b10, 0b00, (imm9)&0x1ff, 0b01, Rn, Rt))
 #define STRw_S9_preindex(Rt, Rn, imm9)    EMIT(STR_gen(0b10, 0b00, (imm9)&0x1ff, 0b11, Rn, Rt))
 
-#define ST_gen(size, op1, imm12, Rn, Rt)        ((size)<<30 | 0b111<<27 | (op1)<<24 | 0b00<<22 | (imm12)<<10 | 0b01<<10 | (Rn)<<5 | (Rt))
-#define STRx_U12(Rt, Rn, imm12)           EMIT(ST_gen(0b11, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
-#define STRw_U12(Rt, Rn, imm12)           EMIT(ST_gen(0b10, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
+#define ST_gen(size, op1, imm12, Rn, Rt)        ((size)<<30 | 0b111<<27 | (op1)<<24 | 0b00<<22 | (imm12)<<10 | (Rn)<<5 | (Rt))
+#define STRx_U12(Rt, Rn, imm12)           EMIT(ST_gen(0b11, 0b01, ((uint32_t)(imm12>>3))&0xfff, Rn, Rt))
+#define STRw_U12(Rt, Rn, imm12)           EMIT(ST_gen(0b10, 0b01, ((uint32_t)(imm12>>2))&0xfff, Rn, Rt))
 #define STRB_U12(Rt, Rn, imm12)           EMIT(ST_gen(0b00, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
-#define STRH_U12(Rt, Rn, imm12)           EMIT(ST_gen(0b01, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
-#define STRxw_U12(Rt, Rn, imm12)      EMIT(ST_gen((rex.w)?0b11:0b10, 0b01, ((uint32_t)(imm12))&0xfff, Rn, Rt))
+#define STRH_U12(Rt, Rn, imm12)           EMIT(ST_gen(0b01, 0b01, ((uint32_t)(imm12>>1))&0xfff, Rn, Rt))
+#define STRxw_U12(Rt, Rn, imm12)          EMIT(ST_gen((rex.w)?0b11:0b10, 0b01, ((uint32_t)((imm12)>>(2+rex.w)))&0xfff, Rn, Rt))
 
 #define STR_REG_gen(size, Rm, option, S, Rn, Rt)    ((size)<<30 | 0b111<<27 | 0b00<<22 | 1<<21 | (Rm)<<16 | (option)<<13 | (S)<<12 | (0b10)<<10 | (Rn)<<5 | (Rt))
 #define STRx_REG(Rt, Rn, Rm)            EMIT(STR_REG_gen(0b11, Rm, 0b011, 0, Rn, Rt))
@@ -206,13 +215,13 @@
 #define BLR(Rn)                           EMIT(BR_gen(0, 0b01, 0, 0, Rn, 0))
 
 #define CB_gen(sf, op, imm19, Rt)       ((sf)<<31 | 0b011010<<25 | (op)<<24 | (imm19)<<5 | (Rt))
-#define CBNZx(Rt, imm19)                EMIT(CB_gen(1, 1, ((imm19)>>2)&0x80000, Rt))
-#define CBNZw(Rt, imm19)                EMIT(CB_gen(0, 1, ((imm19)>>2)&0x80000, Rt))
-#define CBZx(Rt, imm19)                 EMIT(CB_gen(1, 0, ((imm19)>>2)&0x80000, Rt))
-#define CBZw(Rt, imm19)                 EMIT(CB_gen(0, 0, ((imm19)>>2)&0x80000, Rt))
+#define CBNZx(Rt, imm19)                EMIT(CB_gen(1, 1, ((imm19)>>2)&0x7FFFF, Rt))
+#define CBNZw(Rt, imm19)                EMIT(CB_gen(0, 1, ((imm19)>>2)&0x7FFFF, Rt))
+#define CBZx(Rt, imm19)                 EMIT(CB_gen(1, 0, ((imm19)>>2)&0x7FFFF, Rt))
+#define CBZw(Rt, imm19)                 EMIT(CB_gen(0, 0, ((imm19)>>2)&0x7FFFF, Rt))
 
 #define Bcond_gen(imm19, cond)          (0b0101010<<25 | (imm19)<<5 | (cond))
-#define Bcond(cond, imm19)              EMIT(Bcond_gen(((imm19)>>2)&0x80000, cond))
+#define Bcond(cond, imm19)              EMIT(Bcond_gen(((imm19)>>2)&0x7FFFF, cond))
 
 // AND / ORR
 #define LOGIC_gen(sf, opc, N, immr, imms, Rn, Rd)  ((sf)<<31 | (opc)<<29 | 0b100100<<23 | (N)<<22 | (immr)<<16 | (imms)<<10 | (Rn) | Rd)
@@ -238,14 +247,14 @@
 #define ORNx_REG(Rd, Rn, Rm)            EMIT(LOGIC_REG_gen(1, 0b01, 0b00, 1, Rm, 0, Rn, Rd))
 #define ORNw_REG(Rd, Rn, Rm)            EMIT(LOGIC_REG_gen(0, 0b01, 0b00, 1, Rm, 0, Rn, Rd))
 #define ORNxw_REG(Rd, Rn, Rm)           EMIT(LOGIC_REG_gen(rex.w, 0b01, 0b00, 1, Rm, 0, Rn, Rd))
-#define MOVx(Rm, Rd)                    ORRx_REG(Rd, xZR, Rm)
+#define MOVx(Rd, Rm)                    ORRx_REG(Rd, xZR, Rm)
 #define MOVx_LSL(Rm, Rd, lsl)           ORRx_REG_LSL(Rd, xZR, Rm, lsl)
-#define MOVw(Rm, Rd)                    ORRw_REG(Rd, xZR, Rm)
-#define MOVxw(Rm, Rd)                   ORRxw_REG(Rd, xZR, Rm)
-#define MVNx(Rm, Rd)                    ORNx_REG(Rd, xZR, Rm)
-#define MVNx_LSL(Rm, Rd, lsl)           ORNx_REG_LSL(Rd, xZR, Rm, lsl)
-#define MVNw(Rm, Rd)                    ORNw_REG(Rd, xZR, Rm)
-#define MVNxw(Rm, Rd)                   ORNxw_REG(Rd, xZR, Rm)
+#define MOVw(Rd, Rm)                    ORRw_REG(Rd, xZR, Rm)
+#define MOVxw(Rd, Rm)                   ORRxw_REG(Rd, xZR, Rm)
+#define MVNx(Rd, Rm)                    ORNx_REG(Rd, xZR, Rm)
+#define MVNx_LSL(Rd, Rm, lsl)           ORNx_REG_LSL(Rd, xZR, Rm, lsl)
+#define MVNw(Rd, Rm)                    ORNw_REG(Rd, xZR, Rm)
+#define MVNxw(Rd, Rm)                   ORNxw_REG(Rd, xZR, Rm)
 #define MOV_frmSP(Rd)                   ADDx_U12(Rd, xSP, 0)
 #define MOV_toSP(Rm)                    ADDx_U12(xSP, Rm, 0)
 #define BICx(Rd, Rn, Rm)                EMIT(LOGIC_REG_gen(1, 0b00, 0b00, 1, Rm, 0, Rn, Rd))
diff --git a/src/dynarec/arm64_epilog.S b/src/dynarec/arm64_epilog.S
index af39c1ba..99723873 100755
--- a/src/dynarec/arm64_epilog.S
+++ b/src/dynarec/arm64_epilog.S
@@ -9,8 +9,8 @@
 .global arm64_epilog
 arm64_epilog:
     //update register -> emu
-    str     x10,  [x0, (8 *  0)]
-    str     x11,  [x0, (8 *  1)]
+    str     x10, [x0, (8 *  0)]
+    str     x11, [x0, (8 *  1)]
     str     x12, [x0, (8 *  2)]
     str     x13, [x0, (8 *  3)]
     str     x14, [x0, (8 *  4)]
@@ -26,7 +26,7 @@ arm64_epilog:
     str     x24, [x0, (8 * 14)]
     str     x25, [x0, (8 * 15)]
     str     x26, [x0, (8 * 16)]
-    str     x27, [x0, (8 * 17)] // put back reg value in emu, including EIP (so x25 must be EIP now)
+    str     x27, [x0, (8 * 17)] // put back reg value in emu, including EIP (so x27 must be EIP now)
     //restore all used register
     //vpop     {d8-d15}
     ldr     x10, [sp, (8 *  0)]
@@ -48,7 +48,7 @@ arm64_epilog:
     ldr     x26, [sp, (8 * 16)]
     ldr     x27, [sp, (8 * 17)]
     add     sp,  sp, (8 * 18)
-    ldp     lr, fp, [sp, 16]!  // saved lr
+    ldp     lr, fp, [sp], 16  // saved lr
     //end, return...
     ret
 
@@ -57,24 +57,24 @@ arm64_epilog:
 arm64_epilog_fast:
     //restore all used register
     //vpop     {d8-d15}
-    ldr     x8,  [sp, (8 *  0)]
-    ldr     x9,  [sp, (8 *  1)]
-    ldr     x10, [sp, (8 *  2)]
-    ldr     x11, [sp, (8 *  3)]
-    ldr     x12, [sp, (8 *  4)]
-    ldr     x13, [sp, (8 *  5)]
-    ldr     x14, [sp, (8 *  6)]
-    ldr     x15, [sp, (8 *  7)]
-    ldr     x16, [sp, (8 *  8)]
-    ldr     x17, [sp, (8 *  9)]
-    ldr     x18, [sp, (8 * 10)]
-    ldr     x19, [sp, (8 * 11)]
-    ldr     x20, [sp, (8 * 12)]
-    ldr     x21, [sp, (8 * 13)]
-    ldr     x22, [sp, (8 * 14)]
-    ldr     x23, [sp, (8 * 15)]
-    ldr     x24, [sp, (8 * 16)]
-    ldr     x25, [sp, (8 * 17)]
+    ldr     x10, [sp, (8 *  0)]
+    ldr     x11, [sp, (8 *  1)]
+    ldr     x12, [sp, (8 *  2)]
+    ldr     x13, [sp, (8 *  3)]
+    ldr     x14, [sp, (8 *  4)]
+    ldr     x15, [sp, (8 *  5)]
+    ldr     x16, [sp, (8 *  6)]
+    ldr     x17, [sp, (8 *  7)]
+    ldr     x18, [sp, (8 *  8)]
+    ldr     x19, [sp, (8 *  9)]
+    ldr     x20, [sp, (8 * 10)]
+    ldr     x21, [sp, (8 * 11)]
+    ldr     x22, [sp, (8 * 12)]
+    ldr     x23, [sp, (8 * 13)]
+    ldr     x24, [sp, (8 * 14)]
+    ldr     x25, [sp, (8 * 15)]
+    ldr     x26, [sp, (8 * 16)]
+    ldr     x27, [sp, (8 * 17)]
     add     sp,  sp, (8 * 18)
     ldp     lr, fp, [sp, 16]!  // saved lr
     //end, return...
diff --git a/src/dynarec/arm64_next.S b/src/dynarec/arm64_next.S
index 2410750c..47dfe737 100755
--- a/src/dynarec/arm64_next.S
+++ b/src/dynarec/arm64_next.S
@@ -11,9 +11,8 @@
 .global arm64_next
 arm64_next:
     // emu is r0
-    // don't put put back reg value in emu, faster but more tricky to debug
     // IP address is r1
-    sub     sp,  sp, (8 * 11)
+    sub     sp,  sp, (8 * 12)
     str     x0,  [sp, (8 *  0)]
     str     x1,  [sp, (8 *  1)]
     str     x10, [sp, (8 *  2)]
@@ -41,7 +40,7 @@ arm64_next:
     ldr     x16, [sp, (8 *  8)]
     ldr     x17, [sp, (8 *  9)]
     ldr     x18, [sp, (8 * 10)]
-    add     sp,  sp, (8 * 11)
+    add     sp,  sp, (8 * 12)
     // return offset is jump address
     br      x3
 
diff --git a/src/dynarec/arm64_printer.c b/src/dynarec/arm64_printer.c
index 49539a1a..dcfa3032 100755
--- a/src/dynarec/arm64_printer.c
+++ b/src/dynarec/arm64_printer.c
@@ -4,11 +4,279 @@
 #include <stdio.h>

 

 #include "arm64_printer.h"

+#include "debug.h"

 

-const char* arm64_print(uint32_t opcode)

+static const char* Xt[] = {"xEmu", "x1", "x2", "x3", "x4", "x5", "x6", "x7", "x8", "x9", "xRAX", "xRCX", "xRDX", "xRBX", "xRSP", "xRBP", "xRSI", "xRDI", "xR8", "xR9", "xR10", "xR11", "xR12", "xR13", "xR14", "xR15", "xFlags", "xRIP", "x28", "FP", "LR", "xZR"};

+static const char* XtSp[] = {"xEmu", "x1", "x2", "x3", "x4", "x5", "x6", "x7", "x8", "x9", "xRAX", "xRCX", "xRDX", "xRBX", "xRSP", "xRBP", "xRSI", "xRDI", "xR8", "xR9", "xR10", "xR11", "xR12", "xR13", "xR14", "xR15", "xFlags", "xRIP", "x28", "FP", "LR", "SP"};

+static const char* Wt[] = {"w0", "w1", "w2", "w3", "w4", "w5", "w6", "w7", "w8", "w9", "wEAX", "wECX", "wEDX", "wEBX", "wESP", "wEBP", "wESI", "wEDI", "wR8", "wR9", "wR10", "wR11", "wR12", "wR13", "wR14", "wR15", "wFlags", "w27", "w28", "w29", "w30", "wZR"};

+static const char* WtSp[] = {"w0", "w1", "w2", "w3", "w4", "w5", "w6", "w7", "w8", "w9", "wEAX", "wECX", "wEDX", "wEBX", "wESP", "wEBP", "wESI", "wEDI", "wR8", "wR9", "wR10", "wR11", "wR12", "wR13", "wR14", "wR15", "wFlags", "w27", "w28", "w29", "w30", "wSP"};

+

+static const char* conds[] = {"cEQ", "cNE", "cCS", "cCC", "cMI", "cPL", "cVS", "cVC", "cHI", "cLS", "cGE", "cLT", "cGT", "cLE", "c__", "inv"};

+

+typedef struct arm64_print_s {

+    int N, S;

+    int t, n, m, d;

+    int f, c, o, h;

+    int i, r, s;

+    int x, w;

+} arm64_print_t;

+

+int isMask(uint32_t opcode, const char* mask, arm64_print_t *a)

+{

+    if(strlen(mask)!=32) {

+        printf_log(LOG_NONE, "Error: printer mask \"%s\" in not len 32 but %ld\n", mask, strlen(mask));

+        return 0;

+    }

+    memset(a, 0, sizeof(*a));

+    int i = 31;

+    while(*mask) {

+        uint8_t v = (opcode>>i)&1;

+        switch(*mask) {

+            case '0': if(v!=0) return 0; break;

+            case '1': if(v!=1) return 0; break;

+            case 'N': a->N = (a->N<<1) | v; break;

+            case 'S': a->S = (a->S<<1) | v; break;

+            case 't': a->t = (a->t<<1) | v; break;

+            case 'n': a->n = (a->n<<1) | v; break;

+            case 'm': a->m = (a->m<<1) | v; break;

+            case 'd': a->d = (a->d<<1) | v; break;

+            case 'f': a->f = (a->f<<1) | v; break;

+            case 'c': a->c = (a->c<<1) | v; break;

+            case 'i': a->i = (a->i<<1) | v; break;

+            case 'r': a->r = (a->r<<1) | v; break;

+            case 's': a->s = (a->s<<1) | v; break;

+            case 'o': a->o = (a->o<<1) | v; break;

+            case 'h': a->h = (a->h<<1) | v; break;

+            case 'w': a->w = (a->w<<1) | v; break;

+            case 'x': a->x = (a->x<<1) | v; break;

+            default:

+                printf_log(LOG_NONE, "Warning, printer mask use unhandled '%c'\n", *mask);

+        }

+        mask++;

+        --i;

+    }

+    

+    return 1;

+}

+

+int64_t signExtend(uint32_t val, int sz)

+{

+    int64_t ret = val;

+    if((val>>(sz-1))&1 == 1)

+        val |= (0xffffffffffffffffll<<sz);

+    return val;

+}

+

+const char* arm64_print(uint32_t opcode, uintptr_t addr)

 {

     static char buff[200];

+    arm64_print_t a;

+    #define Rn a.n

+    #define Rt a.t

+    #define Rm a.m

+    #define Rd a.d

+    #define sf a.f

+    #define imm a.i

+    #define option a.o

+    #define shift a.h

+    #define hw a.w

+    #define cond a.c

+    // --- LDR / STR

+    if(isMask(opcode, "1x111000010iiiiiiiii01nnnnnttttt", &a)) {

+        int size = (opcode>>30)&3;

+        int offset = signExtend(imm, 9);

+        snprintf(buff, sizeof(buff), "LDR %s, [%s], %d", (size==0b10)?Wt[Rt]:Xt[Rt], XtSp[Rn], offset);

+        return buff;

+    }

+    if(isMask(opcode, "1x111000010iiiiiiiii11nnnnnttttt", &a)) {

+        int size = (opcode>>30)&3;

+        int offset = signExtend(imm, 9);

+        snprintf(buff, sizeof(buff), "LDR %s, [%s, %d]!", (size==0b10)?Wt[Rt]:Xt[Rt], XtSp[Rn], offset);

+        return buff;

+    }

+    if(isMask(opcode, "1x11100101iiiiiiiiiiiinnnnnttttt", &a)) {

+        int size = (opcode>>30)&3;

+        int offset = (imm)<<size;

+        if(offset)

+            snprintf(buff, sizeof(buff), "LDR %s, [%s, %d]", (size==0b10)?Wt[Rt]:Xt[Rt], XtSp[Rn], offset);

+        else

+            snprintf(buff, sizeof(buff), "LDR %s, [%s]", (size==0b10)?Wt[Rt]:Xt[Rt], XtSp[Rn]);

+        return buff;

+    }

+    if(isMask(opcode, "0x011000iiiiiiiiiiiiiiiiiiittttt", &a)) {

+        int size = ((opcode>>30)&1)?3:2;

+        int offset = signExtend(imm, 9)<<2;

+        snprintf(buff, sizeof(buff), "LDR %s, [#%+d]\t;%p", (size==2)?Wt[Rt]:Xt[Rt], offset, (void*)(addr+offset));

+        return buff;

+    }

+    if(isMask(opcode, "1x111000011mmmmmoooS10nnnnnttttt", &a)) {

+        int size = (opcode>>30)&3;

+        const char* extend[] = {"?0", "?1", "UXTW", "LSL", "?4", "?5", "SXTW", "SXTX"};

+        int amount = size*a.S;

+        if(option==3 && !amount)

+            snprintf(buff, sizeof(buff), "LDR %s, [%s, %s]", (size==2)?Wt[Rt]:Xt[Rt], XtSp[Rn], ((option&1)==0)?Wt[Rm]:Xt[Rm]);

+        else if(!amount)

+        if(option==3 && !amount)

+            snprintf(buff, sizeof(buff), "LDR %s, [%s, %s, %s]", (size==2)?Wt[Rt]:Xt[Rt], XtSp[Rn], ((option&1)==0)?Wt[Rm]:Xt[Rm], extend[option]);

+        else

+            snprintf(buff, sizeof(buff), "LDR %s, [%s, %s, %s %d]", (size==2)?Wt[Rt]:Xt[Rt], XtSp[Rn], ((option&1)==0)?Wt[Rm]:Xt[Rm], extend[option], amount);

+        return buff;

+    }

+    if(isMask(opcode, "1x111000000iiiiiiiii01nnnnnttttt", &a)) {

+        int size = (opcode>>30)&3;

+        int offset = signExtend(imm, 9);

+        snprintf(buff, sizeof(buff), "STR %s, [%s], %d", (size==0b10)?Wt[Rt]:Xt[Rt], XtSp[Rn], offset);

+        return buff;

+    }

+    if(isMask(opcode, "1x111000000iiiiiiiii11nnnnnttttt", &a)) {

+        int size = (opcode>>30)&3;

+        int offset = signExtend(imm, 9);

+        snprintf(buff, sizeof(buff), "STR %s, [%s, %d]!", (size==0b10)?Wt[Rt]:Xt[Rt], XtSp[Rn], offset);

+        return buff;

+    }

+    if(isMask(opcode, "1x11100100iiiiiiiiiiiinnnnnttttt", &a)) {

+        int size = (opcode>>30)&3;

+        int offset = (imm)<<size;

+        if(offset)

+            snprintf(buff, sizeof(buff), "STR %s, [%s, %d]", (size==0b10)?Wt[Rt]:Xt[Rt], XtSp[Rn], offset);

+        else

+            snprintf(buff, sizeof(buff), "STR %s, [%s]", (size==0b10)?Wt[Rt]:Xt[Rt], XtSp[Rn]);

+        return buff;

+    }

+    if(isMask(opcode, "1x111000001mmmmmoooS10nnnnnttttt", &a)) {

+        int size = (opcode>>30)&3;

+        const char* extend[] = {"?0", "?1", "UXTW", "LSL", "?4", "?5", "SXTW", "SXTX"};

+        int amount = size*a.S;

+        if(option==3 && !amount)

+            snprintf(buff, sizeof(buff), "STR %s, [%s, %s]", (size==2)?Wt[Rt]:Xt[Rt], XtSp[Rn], ((option&1)==0)?Wt[Rm]:Xt[Rm]);

+        else if(!amount)

+        if(option==3 && !amount)

+            snprintf(buff, sizeof(buff), "STR %s, [%s, %s, %s]", (size==2)?Wt[Rt]:Xt[Rt], XtSp[Rn], ((option&1)==0)?Wt[Rm]:Xt[Rm], extend[option]);

+        else

+            snprintf(buff, sizeof(buff), "STR %s, [%s, %s, %s %d]", (size==2)?Wt[Rt]:Xt[Rt], XtSp[Rn], ((option&1)==0)?Wt[Rm]:Xt[Rm], extend[option], amount);

+        return buff;

+    }

+    // --- MOV

+    if(isMask(opcode, "f0101010000mmmmm00000011111ddddd", &a)) {

+        snprintf(buff, sizeof(buff), "MOV %s, %s", sf?Xt[Rd]:Wt[Rd], sf?Xt[Rm]:Wt[Rm]);

+        return buff;

+    }

+    if(isMask(opcode, "f10100101wwiiiiiiiiiiiiiiiiddddd", &a)) {

+        if(!hw)

+            snprintf(buff, sizeof(buff), "MOVZ %s, 0x%x", sf?Xt[Rd]:Wt[Rd], imm);

+        else

+            snprintf(buff, sizeof(buff), "MOVZ %s, 0x%x LSL %d", sf?Xt[Rd]:Wt[Rd], imm, 16*hw);

+        return buff;

+    }

+    if(isMask(opcode, "f11100101wwiiiiiiiiiiiiiiiiddddd", &a)) {

+        if(!hw)

+            snprintf(buff, sizeof(buff), "MOVK %s, 0x%x", sf?Xt[Rd]:Wt[Rd], imm);

+        else

+            snprintf(buff, sizeof(buff), "MOVK %s, 0x%x LSL %d", sf?Xt[Rd]:Wt[Rd], imm, 16*hw);

+        return buff;

+    }

+

+    // --- MATH

+    if(isMask(opcode, "f1101011001mmmmmoooiiinnnnn11111", &a)) {

+        int R = 0;

+        if(option==0b011 || option==0b111)

+            R = 1;

+        const char* extends[] ={"UXTB", "UXTH", "UXTW", "UXTX", "SXTB", "SXTH", "SXTW", "SXTX"};

+        if(imm==0 && option==(sf?3:2))

+            snprintf(buff, sizeof(buff), "CMP %s, %s", sf?XtSp[Rn]:WtSp[Rn], (sf&&R)?Xt[Rm]:Wt[Rm]);

+        else

+            snprintf(buff, sizeof(buff), "CMP %s, %s, %s 0x%x", sf?XtSp[Rn]:WtSp[Rn], (sf&&R)?Xt[Rm]:Wt[Rm], extends[option], imm);

+        return buff;

+    }

+    if(isMask(opcode, "f11100010hiiiiiiiiiiiinnnnn11111", &a)) {

+        if(shift==0)

+            snprintf(buff, sizeof(buff), "CMP %s, 0x%x", sf?XtSp[Rn]:WtSp[Rn], imm);

+        else

+            snprintf(buff, sizeof(buff), "ADD %s, 0x%x", sf?XtSp[Rn]:WtSp[Rn], imm<<12);

+        return buff;

+    }

+    if(isMask(opcode, "f1101011hh0mmmmmiiiiiinnnnn11111", &a)) {

+        const char* shifts[] = { "LSL", "LSR", "ASR", "???"};

+        if(shift==0 && imm==0)

+            snprintf(buff, sizeof(buff), "CMP %s, %s", sf?Xt[Rn]:Wt[Rn], sf?Xt[Rm]:Wt[Rm]);

+        else 

+            snprintf(buff, sizeof(buff), "CMP %s, %s %s %d", sf?Xt[Rn]:Wt[Rn], sf?Xt[Rm]:Wt[Rm], shifts[shift], imm);

+        return buff;

+    }

+    if(isMask(opcode, "f0010001hhiiiiiiiiiiiinnnnnddddd", &a)) {

+        if(shift==0)

+            snprintf(buff, sizeof(buff), "ADD %s, %s, 0x%x", sf?XtSp[Rd]:WtSp[Rd], sf?XtSp[Rn]:WtSp[Rn], imm);

+        else if (shift==1)

+            snprintf(buff, sizeof(buff), "ADD %s, %s, 0x%x", sf?XtSp[Rd]:WtSp[Rd], sf?XtSp[Rn]:WtSp[Rn], imm<<12);

+        else 

+            snprintf(buff, sizeof(buff), "ADD with unhandled shift %d", shift);

+        return buff;

+    }

+    if(isMask(opcode, "f0110001hhiiiiiiiiiiiinnnnnddddd", &a)) {

+        if(shift==0)

+            snprintf(buff, sizeof(buff), "ADDS %s, %s, 0x%x", sf?Xt[Rd]:Wt[Rd], sf?XtSp[Rn]:WtSp[Rn], imm);

+        else if (shift==1)

+            snprintf(buff, sizeof(buff), "ADDS %s, %s, 0x%x", sf?Xt[Rd]:Wt[Rd], sf?XtSp[Rn]:WtSp[Rn], imm<<12);

+        else 

+            snprintf(buff, sizeof(buff), "ADDS with unhandled shift %d", shift);

+        return buff;

+    }

+    if(isMask(opcode, "f0001011hh0mmmmmiiiiiinnnnnddddd", &a)) {

+        const char* shifts[] = { "LSL", "LSR", "ASR", "???"};

+        if(shift==0 && imm==0)

+            snprintf(buff, sizeof(buff), "ADD %s, %s, %s", sf?Xt[Rd]:Wt[Rd], sf?Xt[Rn]:Wt[Rn], sf?Xt[Rm]:Wt[Rm]);

+        else 

+            snprintf(buff, sizeof(buff), "ADD %s, %s, %s %s %d", sf?Xt[Rd]:Wt[Rd], sf?Xt[Rn]:Wt[Rn], sf?Xt[Rm]:Wt[Rm], shifts[shift], imm);

+        return buff;

+    }

+    if(isMask(opcode, "f1010001hhiiiiiiiiiiiinnnnnddddd", &a)) {

+        if(shift==0)

+            snprintf(buff, sizeof(buff), "SUB %s, %s, 0x%x", sf?XtSp[Rd]:WtSp[Rd], sf?XtSp[Rn]:WtSp[Rn], imm);

+        else if (shift==1)

+            snprintf(buff, sizeof(buff), "SUB %s, %s, 0x%x", sf?XtSp[Rd]:WtSp[Rd], sf?XtSp[Rn]:WtSp[Rn], imm<<12);

+        else 

+            snprintf(buff, sizeof(buff), "SUB with unhandled shift %d", shift);

+        return buff;

+    }

+    if(isMask(opcode, "f1110001hhiiiiiiiiiiiinnnnnddddd", &a)) {

+        if(shift==0)

+            snprintf(buff, sizeof(buff), "SUBS %s, %s, 0x%x", sf?Xt[Rd]:Wt[Rd], sf?XtSp[Rn]:WtSp[Rn], imm);

+        else if (shift==1)

+            snprintf(buff, sizeof(buff), "SUBS %s, %s, 0x%x", sf?Xt[Rd]:Wt[Rd], sf?XtSp[Rn]:WtSp[Rn], imm<<12);

+        else 

+            snprintf(buff, sizeof(buff), "SUBS with unhandled shift %d", shift);

+        return buff;

+    }

+    if(isMask(opcode, "f1001011hh0mmmmmiiiiiinnnnnddddd", &a)) {

+        const char* shifts[] = { "LSL", "LSR", "ASR", "???"};

+        if(shift==0 && imm==0)

+            snprintf(buff, sizeof(buff), "SUB %s, %s, %s", sf?Xt[Rd]:Wt[Rd], sf?Xt[Rn]:Wt[Rn], sf?Xt[Rm]:Wt[Rm]);

+        else 

+            snprintf(buff, sizeof(buff), "SUB %s, %s, %s %s %d", sf?Xt[Rd]:Wt[Rd], sf?Xt[Rn]:Wt[Rn], sf?Xt[Rm]:Wt[Rm], shifts[shift], imm);

+        return buff;

+    }

+    // ---- LOGIC

+

+    // ---- BRANCH / TEST

+    if(isMask(opcode, "1101011000011111000000nnnnn00000", &a)) {

+        snprintf(buff, sizeof(buff), "BR %s", Xt[Rn]);

+        return buff;

+    }

+    if(isMask(opcode, "1101011000111111000000nnnnn00000", &a)) {

+        snprintf(buff, sizeof(buff), "BLR %s", Xt[Rn]);

+        return buff;

+    }

+    if(isMask(opcode, "01010100iiiiiiiiiiiiiiiiiii0cccc", &a)) {

+        int offset = signExtend(imm, 19)<<2;

+        snprintf(buff, sizeof(buff), "B.%s #+%d\t; %p", conds[cond], offset, (void*)(addr + offset));

+        return buff;

+    }

+    if(isMask(opcode, "f0110100iiiiiiiiiiiiiiiiiiittttt", &a)) {

+        int offset = signExtend(imm, 19)<<2;

+        snprintf(buff, sizeof(buff), "CBZ %s, #%+d\t; %p", Xt[Rt], offset, (void*)(addr + offset));

+        return buff;

+    }

 

-    snprintf(buff, sizeof(buff), "0x%8X ???", opcode);

+    snprintf(buff, sizeof(buff), "%08X ???", __builtin_bswap32(opcode));

     return buff;

 }
\ No newline at end of file
diff --git a/src/dynarec/arm64_printer.h b/src/dynarec/arm64_printer.h
index 6fe21c33..2475f631 100644
--- a/src/dynarec/arm64_printer.h
+++ b/src/dynarec/arm64_printer.h
@@ -1,6 +1,6 @@
 #ifndef _ARM_PRINTER_H_
 #define _ARM_PRINTER_H_
 
-const char* arm64_print(uint32_t opcode);
+const char* arm64_print(uint32_t opcode, uint64_t addr);
 
 #endif //_ARM_PRINTER_H_
diff --git a/src/dynarec/arm64_prolog.S b/src/dynarec/arm64_prolog.S
index f480f2ea..c1dc3b48 100755
--- a/src/dynarec/arm64_prolog.S
+++ b/src/dynarec/arm64_prolog.S
@@ -9,7 +9,7 @@
 .global arm64_prolog
 arm64_prolog:
     //save all 18 used register
-    stp     lr, fp, [sp, 16]!  // save lr
+    stp     lr, fp, [sp, -16]!  // save lr
     sub     sp,  sp, (8 * 18)
     str     x10, [sp, (8 *  0)]
     str     x11, [sp, (8 *  1)]
diff --git a/src/dynarec/dynablock.c b/src/dynarec/dynablock.c
index 337e1751..06694e06 100755
--- a/src/dynarec/dynablock.c
+++ b/src/dynarec/dynablock.c
@@ -327,7 +327,7 @@ static dynablock_t* internalDBGetBlock(x64emu_t* emu, uintptr_t addr, uintptr_t
         pthread_mutex_lock(&my_context->mutex_dyndump);
     // fill the block
     block->x64_addr = (void*)addr;
-    if(FillBlock64(block, filladdr)) {
+    if(!FillBlock64(block, filladdr)) {
         void* old = (void*)arm64_lock_xchg(&dynablocks->direct[addr-dynablocks->text], 0);
         if(old!=block && old) {// put it back in place, strange things are happening here!
             dynarec_log(LOG_INFO, "Warning, a wild block appeared at %p: %p\n", (void*)addr, old);
diff --git a/src/dynarec/dynarec_arm64.c b/src/dynarec/dynarec_arm64.c
index 138fdb3a..9918f4a5 100755
--- a/src/dynarec/dynarec_arm64.c
+++ b/src/dynarec/dynarec_arm64.c
@@ -305,6 +305,26 @@ instsize_t* addInst(instsize_t* insts, size_t* size, size_t* cap, int x64_size,
     return insts;
 }
 
+// add a value to etable64 (if needed) and gives back the imm19 to use in LDR_literal
+int Table64(dynarec_arm_t *dyn, uint64_t val)
+{
+    // find the value if already present
+    int idx = -1;
+    for(int i=0; i<dyn->table64size && (idx==-1); ++i)
+        if(dyn->table64[i] == val)
+            idx = i;
+    // not found, add it
+    if(idx==-1) {
+        idx = dyn->table64size++;
+        dyn->table64 = (uint64_t*)realloc(dyn->table64, dyn->table64size * sizeof(uint64_t));
+        dyn->table64[idx] = val;
+    }
+    // calculate offset
+    int delta = dyn->tablestart + idx*sizeof(uint64_t) - (uintptr_t)dyn->block;
+    return delta;
+}
+
+
 void arm_pass0(dynarec_arm_t* dyn, uintptr_t addr);
 void arm_pass1(dynarec_arm_t* dyn, uintptr_t addr);
 void arm_pass2(dynarec_arm_t* dyn, uintptr_t addr);
@@ -356,16 +376,18 @@ void* FillBlock64(dynablock_t* block, uintptr_t addr) {
     // pass 2, instruction size
     arm_pass2(&helper, addr);
     // ok, now allocate mapped memory, with executable flag on
-    int sz = helper.arm_size;
+    int sz = helper.arm_size + helper.table64size*sizeof(uint64_t);
     void* p = (void*)AllocDynarecMap(block, sz);
     if(p==NULL) {
         dynarec_log(LOG_DEBUG, "AllocDynarecMap(%p, %d) failed, cancelling block\n", block, sz);
         free(helper.insts);
         free(helper.next);
+        free(helper.table64);
         return NULL;
     }
     helper.block = p;
     helper.arm_start = (uintptr_t)p;
+    helper.tablestart = helper.arm_start + helper.arm_size;
     if(helper.sons_size) {
         helper.sons_x64 = (uintptr_t*)calloc(helper.sons_size, sizeof(uintptr_t));
         helper.sons_arm = (void**)calloc(helper.sons_size, sizeof(void*));
@@ -378,8 +400,8 @@ void* FillBlock64(dynablock_t* block, uintptr_t addr) {
     }
     helper.arm_size = 0;
     arm_pass3(&helper, addr);
-    if(sz!=helper.arm_size) {
-        printf_log(LOG_NONE, "BOX64: Warning, size difference in block between pass2 (%d) & pass3 (%d)!\n", sz, helper.arm_size);
+    if(sz!=(helper.arm_size + helper.table64size*8)) {
+        printf_log(LOG_NONE, "BOX64: Warning, size difference in block between pass2 (%d) & pass3 (%d)!\n", sz, helper.arm_size+helper.table64size*8);
         uint8_t *dump = (uint8_t*)helper.start;
         printf_log(LOG_NONE, "Dump of %d x64 opcodes:\n", helper.size);
         for(int i=0; i<helper.size; ++i) {
@@ -390,6 +412,10 @@ void* FillBlock64(dynablock_t* block, uintptr_t addr) {
         }
         printf_log(LOG_NONE, " ------------\n");
     }
+    // add table64 if needed
+    if(helper.table64size) {
+        memcpy(p+helper.arm_size, helper.table64, helper.table64size*8);
+    }
     // all done...
     __clear_cache(p, p+sz);   // need to clear the cache before execution...
     // keep size of instructions for signal handling
@@ -406,6 +432,7 @@ void* FillBlock64(dynablock_t* block, uintptr_t addr) {
     // ok, free the helper now
     free(helper.insts);
     free(helper.next);
+    free(helper.table64);
     block->size = sz;
     block->isize = helper.size;
     block->block = p;
diff --git a/src/dynarec/dynarec_arm64_00.c b/src/dynarec/dynarec_arm64_00.c
index 17d0a8a9..a5d903af 100755
--- a/src/dynarec/dynarec_arm64_00.c
+++ b/src/dynarec/dynarec_arm64_00.c
@@ -92,7 +92,11 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
         case 0x56:
         case 0x57:
             INST_NAME("PUSH reg");
-            gd = xRAX+(opcode&0x07)+(rex.r<<3);
+            gd = xRAX+(opcode&0x07)+(rex.b<<3);
+            if(gd==xRSP) {
+                MOVx(x1, gd);
+                gd = x1;
+            }
             PUSH1(gd);
             break;
         case 0x58:
@@ -104,8 +108,13 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
         case 0x5E:
         case 0x5F:
             INST_NAME("POP reg");
-            gd = xRAX+(opcode&0x07)+(rex.r<<3);
-            POP1(gd);
+            gd = xRAX+(opcode&0x07)+(rex.b<<3);
+            if(gd == xRSP) {
+                POP1(x1);
+                MOVx(gd, x1);
+            } else {
+                POP1(gd);
+            }
             break;
 
         case 0x81:
@@ -132,9 +141,9 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
             nextop=F8;
             GETGD;
             if(MODREG) {   // reg <= reg
-                MOVxw(xRAX+(nextop&7), gd);
+                MOVxw(xRAX+(nextop&7)+(rex.b<<3), gd);
             } else {                    // mem <= reg
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 4095, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, 0, 0);
                 STRxw_U12(gd, ed, fixedaddress);
             }
             break;
@@ -153,6 +162,81 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
             }
             break;
 
+        case 0xE8:
+            INST_NAME("CALL Id");
+            i32 = F32S;
+            if(addr+i32==0) {
+                #if STEP == 3
+                printf_log(LOG_INFO, "Warning, CALL to 0x0 at %p (%p)\n", (void*)addr, (void*)(addr-1));
+                #endif
+            }
+            #if STEP == 0
+            if(isNativeCall(dyn, addr+i32, NULL, NULL))
+                tmp = 3;
+            else 
+                tmp = 0;
+            #elif STEP < 2
+            if(isNativeCall(dyn, addr+i32, &dyn->insts[ninst].natcall, &dyn->insts[ninst].retn))
+                tmp = dyn->insts[ninst].pass2choice = 3;
+            else 
+                tmp = dyn->insts[ninst].pass2choice = 0;
+            #else
+                tmp = dyn->insts[ninst].pass2choice;
+            #endif
+            switch(tmp) {
+                case 3:
+                    SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
+                    BARRIER(1);
+                    BARRIER_NEXT(1);
+                    TABLE64(x2, addr);
+                    PUSH1(x2);
+                    MESSAGE(LOG_DUMP, "Native Call to %s (retn=%d)\n", GetNativeName(GetNativeFnc(dyn->insts[ninst].natcall-1)), dyn->insts[ninst].retn);
+                    // calling a native function
+                    x87_forget(dyn, ninst, x3, x4, 0);
+                    TABLE64(xRIP, dyn->insts[ninst].natcall); // read the 0xCC already
+                    STORE_XEMU_REGS(xRIP);
+                    CALL_S(x64Int3, -1);
+                    LOAD_XEMU_REGS(xRIP);
+                    TABLE64(x3, dyn->insts[ninst].natcall);
+                    ADDx_U12(x3, x3, 2+8+8);
+                    CMPSx_REG(xRIP, x3);
+                    B_MARK(cNE);    // Not the expected address, exit dynarec block
+                    POP1(xRIP);   // pop the return address
+                    if(dyn->insts[ninst].retn) {
+                        ADDx_U12(xRSP, xRSP, dyn->insts[ninst].retn);
+                    }
+                    TABLE64(x3, addr);
+                    CMPSx_REG(xRIP, x3);
+                    B_MARK(cNE);    // Not the expected address again
+                    LDRw_U12(w1, xEmu, offsetof(x64emu_t, quit));
+                    CBZw_NEXT(w1);  // not quitting, so lets continue
+                    MARK;
+                    jump_to_epilog(dyn, 0, xRIP, ninst);
+                    break;
+                default:
+                    if(ninst && dyn->insts && dyn->insts[ninst-1].x64.set_flags) {
+                        READFLAGS(X_PEND);  // that's suspicious
+                    } else {
+                        SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
+                    }
+                    // regular call
+                    BARRIER(1);
+                    BARRIER_NEXT(1);
+                    if(!dyn->insts || ninst==dyn->size-1) {
+                        *need_epilog = 0;
+                        *ok = 0;
+                    }
+                    TABLE64(x2, addr);
+                    PUSH1(x2);
+                    if(addr+i32==0) {   // self modifying code maybe? so use indirect address fetching
+                        TABLE64(x4, addr-4);
+                        LDRx_U12(x4, x4, 0);
+                        jump_to_next(dyn, 0, x4, ninst);
+                    } else
+                        jump_to_next(dyn, addr+i32, 0, ninst);
+                    break;
+            }
+            break;
 
         default:
             DEFAULT;
diff --git a/src/dynarec/dynarec_arm64_emit_math.c b/src/dynarec/dynarec_arm64_emit_math.c
index 64a9a00d..1c14991e 100755
--- a/src/dynarec/dynarec_arm64_emit_math.c
+++ b/src/dynarec/dynarec_arm64_emit_math.c
@@ -159,8 +159,8 @@
 void emit_sub32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3, int s4, int s5)
 {
     IFX(X_PEND) {
-        STRw_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRw_U12(s2, xEmu, offsetof(x64emu_t, op2));
+        STRxw_U12(s1, xEmu, offsetof(x64emu_t, op1));
+        STRxw_U12(s2, xEmu, offsetof(x64emu_t, op2));
         SET_DF(s3, d_sub32);
     } else IFX(X_ALL) {
         SET_DFNONE(s3);
@@ -176,7 +176,7 @@ void emit_sub32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3
         SUBxw_REG(s1, s1, s2);
     }
     IFX(X_PEND) {
-        STRx_U12(s1, xEmu, offsetof(x64emu_t, res));
+        STRxw_U12(s1, xEmu, offsetof(x64emu_t, res));
     }
     IFX(X_AF) {
         ANDxw_REG(s3, s3, s1);   // s3 = (~op1 | op2) & res
@@ -216,7 +216,12 @@ void emit_sub32c(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int64_t c, in
     if(s1==xRSP && (!dyn->insts || dyn->insts[ninst].x64.need_flags==X_PEND))
     {
         // special case when doing math on RSP and only PEND is needed: ignoring it!
-        SUBxw_U12(s1, s1, c);
+        if(c>=0 && c<0x1000) {
+            SUBxw_U12(s1, s1, c);
+        } else {
+            MOV64x(s5, c);
+            SUBxw_REG(s1, s1, s5);
+        }
         return;
     }
     IFX(X_PEND) {
diff --git a/src/dynarec/dynarec_arm64_functions.c b/src/dynarec/dynarec_arm64_functions.c
index d3a1e2bb..9ad74131 100755
--- a/src/dynarec/dynarec_arm64_functions.c
+++ b/src/dynarec/dynarec_arm64_functions.c
@@ -319,12 +319,12 @@ uintptr_t fakeed(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop)
 int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int* retn)
 {
 #define PK(a)       *(uint8_t*)(addr+a)
-#define PK64(a)     *(uint64_t*)(addr+a)
+#define PK32(a)     *(int32_t*)(addr+a)
 
     if(!addr)
         return 0;
-    if(PK(0)==0xff && PK(1)==0x25) {  // absolute jump, maybe the GOT
-        uintptr_t a1 = (PK64(2));   // need to add a check to see if the address is from the GOT !
+    if(PK(0)==0xff && PK(1)==0x25) {            // "absolute" jump, maybe the GOT (well, RIP relative in fact)
+        uintptr_t a1 = addr + 6 + (PK32(2));    // need to add a check to see if the address is from the GOT !
         addr = *(uintptr_t*)a1; 
     }
     if(addr<0x10000)    // too low, that is suspicious
diff --git a/src/dynarec/dynarec_arm64_helper.c b/src/dynarec/dynarec_arm64_helper.c
index 9e7239bf..c0a084db 100755
--- a/src/dynarec/dynarec_arm64_helper.c
+++ b/src/dynarec/dynarec_arm64_helper.c
@@ -63,7 +63,7 @@ uintptr_t geted(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop, u
         } else if((nextop&7)==5) {
             uint64_t tmp = F32S64;
             MOV64x(ret, tmp);
-            MOV64x(xRIP, addr+delta);
+            TABLE64(xRIP, addr+delta);
             ADDx_REG(ret, ret, xRIP);
         } else {
             ret = xRAX+(nextop&7)+(rex.b<<3);
@@ -86,22 +86,22 @@ uintptr_t geted(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop, u
                 if (sib_reg!=4) {
                     ADDx_REG_LSL(ret, xRAX+(sib&0x07)+(rex.b<<3), xRAX+sib_reg+(rex.x<<3), (sib>>6));
                 } else {
-                    ret = xRAX+(sib&0x07)+rex.b<<3;
+                    ret = xRAX+(sib&0x07)+(rex.b<<3);
                 }
             } else
-                ret = xRAX+(nextop&0x07)+rex.b<<3;
+                ret = xRAX+(nextop&0x07)+(rex.b<<3);
         } else {
             int64_t sub = (i64<0)?1:0;
             if(sub) i64 = -i64;
             if(i64<0x1000) {
                 if((nextop&7)==4) {
                     if (sib_reg!=4) {
-                        ADDx_REG_LSL(scratch, xRAX+(sib&0x07), xRAX+sib_reg, (sib>>6));
+                        ADDx_REG_LSL(scratch, xRAX+(sib&0x07)+(rex.b<<3), xRAX+sib_reg+(rex.x<<3), (sib>>6));
                     } else {
-                        scratch = xRAX+(sib&0x07);
+                        scratch = xRAX+(sib&0x07)+(rex.b<<3);
                     }
                 } else
-                    scratch = xRAX+(nextop&0x07);
+                    scratch = xRAX+(nextop&0x07)+(rex.b<<3);
                 if(sub) {
                     SUBx_U12(ret, scratch, i64);
                 } else {
@@ -225,9 +225,9 @@ void jump_to_epilog(dynarec_arm_t* dyn, uintptr_t ip, int reg, int ninst)
             MOVx(xRIP, reg);
         }
     } else {
-        MOV64x(xRIP, ip);
+        TABLE64(xRIP, ip);
     }
-    MOV64x(x2, (uintptr_t)arm64_epilog);
+    TABLE64(x2, (uintptr_t)arm64_epilog);
     BR(x2);
 }
 
@@ -240,7 +240,7 @@ void jump_to_next(dynarec_arm_t* dyn, uintptr_t ip, int reg, int ninst)
             MOVx(xRIP, reg);
         }
         uintptr_t tbl = getJumpTable64();
-        MOV64x(x2, tbl);
+        TABLE64(x2, tbl);
         UBFXx(x3, xRIP, 48, JMPTABL_SHIFT);
         LDRx_REG_LSL3(x2, x2, x3);
         UBFXx(x3, xRIP, 32, JMPTABL_SHIFT);
@@ -250,8 +250,8 @@ void jump_to_next(dynarec_arm_t* dyn, uintptr_t ip, int reg, int ninst)
         LDRx_REG_UXTW(x3, x2, xRIP);
     } else {
         uintptr_t p = getJumpTableAddress64(ip); 
-        MOV64x(x2, p);
-        MOV64x(xRIP, ip);
+        TABLE64(x2, p);
+        TABLE64(xRIP, ip);
         LDRx_U12(x3, x2, 0);
     }
     MOVx(x1, xRIP);
@@ -323,7 +323,7 @@ void iret_to_epilog(dynarec_arm_t* dyn, int ninst)
 void call_c(dynarec_arm_t* dyn, int ninst, void* fnc, int reg, int ret, int saveflags)
 {
     if(ret!=-2) {
-        STRx_S9_preindex(xSP, xEmu, -16);   // ARM64 stack needs to be 16byte aligned
+        STRx_S9_preindex(xEmu, xSP, -16);   // ARM64 stack needs to be 16byte aligned
     }
     fpu_pushcache(dyn, ninst, reg);
     if(saveflags) {
@@ -336,7 +336,7 @@ void call_c(dynarec_arm_t* dyn, int ninst, void* fnc, int reg, int ret, int save
         MOVx(ret, xEmu);
     }
     if(ret!=-2) {
-        LDRx_S9_postindex(xSP, xEmu, 16);
+        LDRx_S9_postindex(xEmu, xSP, 16);
     }
     if(saveflags) {
         LDRx_U12(xFlags, xEmu, offsetof(x64emu_t, eflags));
diff --git a/src/dynarec/dynarec_arm64_helper.h b/src/dynarec/dynarec_arm64_helper.h
index d4e5a543..c7f645b0 100755
--- a/src/dynarec/dynarec_arm64_helper.h
+++ b/src/dynarec/dynarec_arm64_helper.h
@@ -35,7 +35,7 @@
                     ed = xRAX+(nextop&7)+(rex.b<<3);    \
                     wback = 0;                          \
                 } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff, 0, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, 0, D); \
                     LDRxw_U12(x1, wback, fixedaddress); \
                     ed = x1;                            \
                 }
@@ -43,7 +43,7 @@
                     ed = xRAX+(nextop&7)+(rex.b<<3);    \
                     wback = 0;                          \
                 } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff, 0, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<3, 3, rex, 0, D); \
                     LDRx_U12(x1, wback, fixedaddress);  \
                     ed = x1;                            \
                 }
@@ -51,7 +51,7 @@
                     ed = xEAX+(nextop&7)+(rex.b<<3);    \
                     wback = 0;                          \
                 } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff, 0, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 2, rex, 0, D); \
                     LDRw_U12(x1, wback, fixedaddress);  \
                     ed = x1;                            \
                 }
@@ -84,7 +84,7 @@
 #define SBACK(wb)   if(wback) {STR_IMM9(wb, wback, fixedaddress);} else {MOV_REG(ed, wb);}
 //GETEDO can use r1 for ed, and r2 for wback. wback is 0 if ed is xEAX..xEDI
 #define GETEDO(O)   if((nextop&0xC0)==0xC0) {   \
-                    ed = xEAX+(nextop&7);   \
+                    ed = xEAX+(nextop&7)+(rex.b<<3);   \
                     wback = 0;              \
                 } else {                    \
                     addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0); \
@@ -246,6 +246,10 @@
 #define B_NEXT(cond)     \
     j32 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size)):0; \
     Bcond(cond, j32)
+// Branch to next instruction if reg is 0 (use j32)
+#define CBZw_NEXT(reg)    \
+    j32 =  (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size)):0; \
+    CBZw(reg, j32)
 // Branch to MARKSEG if cond (use j32)
 #define B_MARKSEG(cond)    \
     j32 = GETMARKSEG-(dyn->arm_size);   \
@@ -322,6 +326,27 @@
     STRx_U12(xFlags, xEmu, offsetof(x64emu_t, eflags)); \
     if(A) {STRx_U12(A, xEmu, offsetof(x64emu_t, ip));}
 
+#define LOAD_REG(A)    LDRx_U12(x##A, xEmu, offsetof(x64emu_t, regs[_##A]))
+#define LOAD_XEMU_REGS(A)  \
+    LOAD_REG(RAX);         \
+    LOAD_REG(RCX);         \
+    LOAD_REG(RDX);         \
+    LOAD_REG(RBX);         \
+    LOAD_REG(RSP);         \
+    LOAD_REG(RBP);         \
+    LOAD_REG(RSI);         \
+    LOAD_REG(RDI);         \
+    LOAD_REG(R8);          \
+    LOAD_REG(R9);          \
+    LOAD_REG(R10);         \
+    LOAD_REG(R11);         \
+    LOAD_REG(R12);         \
+    LOAD_REG(R13);         \
+    LOAD_REG(R14);         \
+    LOAD_REG(R15);         \
+    LDRx_U12(xFlags, xEmu, offsetof(x64emu_t, eflags)); \
+    if(A) {LDRx_U12(A, xEmu, offsetof(x64emu_t, ip));}
+
 #define SET_DFNONE(S)    if(!dyn->dfnone) {MOVZw(S, d_none); STRw_U12(S, xEmu, offsetof(x64emu_t, df)); dyn->dfnone=1;}
 #define SET_DF(S, N)     if(N) {MOVZw(S, N); STRw_U12(S, xEmu, offsetof(x64emu_t, df)); dyn->dfnone=0;} else SET_DFNONE(S)
 #define SET_NODF()          dyn->dfnone = 0
@@ -368,6 +393,9 @@
 #ifndef NEW_BARRIER_INST
 #define NEW_BARRIER_INST
 #endif
+#ifndef TABLE64
+#define TABLE64(A, V)
+#endif
 
 #if STEP < 2
 #define PASS2IF(A, B) if(A)
diff --git a/src/dynarec/dynarec_arm64_pass2.h b/src/dynarec/dynarec_arm64_pass2.h
index 095fe4f7..f0bf7fbd 100755
--- a/src/dynarec/dynarec_arm64_pass2.h
+++ b/src/dynarec/dynarec_arm64_pass2.h
@@ -6,4 +6,5 @@
 #define NEW_INST    if(ninst) {dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);}
 #define INST_EPILOG dyn->insts[ninst].epilog = dyn->arm_size; 
 #define INST_NAME(name) 
-#define NEW_BARRIER_INST    if(ninst) ++dyn->sons_size
\ No newline at end of file
+#define NEW_BARRIER_INST    if(ninst) ++dyn->sons_size
+#define TABLE64(A, V)   {Table64(dyn, (V)); EMIT(0);}
\ No newline at end of file
diff --git a/src/dynarec/dynarec_arm64_pass3.h b/src/dynarec/dynarec_arm64_pass3.h
index 2f06a608..a94c61c7 100755
--- a/src/dynarec/dynarec_arm64_pass3.h
+++ b/src/dynarec/dynarec_arm64_pass3.h
@@ -1,7 +1,7 @@
 #define INIT    
 #define FINI
 #define EMIT(A)     \
-    if(box64_dynarec_dump) {dynarec_log(LOG_NONE, "\t%08x\t%s\n", (uint32_t)(A), arm64_print(A));} \
+    if(box64_dynarec_dump) {dynarec_log(LOG_NONE, "\t%08x\t%s\n", (uint32_t)(A), arm64_print(A, (uintptr_t)dyn->block));} \
     *(uint32_t*)(dyn->block) = (uint32_t)(A);       \
     dyn->block += 4; dyn->arm_size += 4;\
     dyn->insts[ninst].size2 += 4
@@ -32,3 +32,4 @@
     ++dyn->sons_size;                               \
     }
 
+#define TABLE64(A, V)   {int val64offset = Table64(dyn, (V)); MESSAGE(LOG_DUMP, "  Table64: 0x%lx\n", (V)); LDRx_literal(A, val64offset);}
\ No newline at end of file
diff --git a/src/dynarec/dynarec_arm64_private.h b/src/dynarec/dynarec_arm64_private.h
index 2d82fd35..c375dd10 100755
--- a/src/dynarec/dynarec_arm64_private.h
+++ b/src/dynarec/dynarec_arm64_private.h
@@ -40,6 +40,9 @@ typedef struct dynarec_arm_s {
     int                 fpu_extra_qscratch; // some opcode need an extra quad scratch register
     int                 fpu_reg;    // x87/sse/mmx reg counter
     int                 dfnone;     // if defered flags is already set to df_none
+    uint64_t            *table64;   // table of 64bits value
+    int                 table64size;// size of table (will be appended at end of executable code)
+    uintptr_t           tablestart;
     uintptr_t*          next;       // variable array of "next" jump address
     int                 next_sz;
     int                 next_cap;
@@ -53,4 +56,6 @@ uintptr_t get_closest_next(dynarec_arm_t *dyn, uintptr_t addr);
 int is_nops(dynarec_arm_t *dyn, uintptr_t addr, int n);
 int is_instructions(dynarec_arm_t *dyn, uintptr_t addr, int n);
 
+int Table64(dynarec_arm_t *dyn, uint64_t val);  // add a value to etable64 (if needed) and gives back the imm19 to use in LDR_literal
+
 #endif //__DYNAREC_ARM_PRIVATE_H_
\ No newline at end of file
diff --git a/src/emu/x64emu_private.h b/src/emu/x64emu_private.h
index 05570965..144a6f33 100755
--- a/src/emu/x64emu_private.h
+++ b/src/emu/x64emu_private.h
@@ -30,14 +30,15 @@ typedef struct x64emu_s {
 	x87flags_t  sw;
 	uint32_t    top;        // top is part of sw, but it's faster to have it separatly
     int         fpu_stack;
+    uint32_t    mxcsr;
 	fpu_round_t round;
     fpu_ld_t    fpu_ld[8]; // for long double emulation / 80bits fld fst
     fpu_ll_t    fpu_ll[8]; // for 64bits fild / fist sequence
 	fpu_p_reg_t p_regs[8];
     // sse
     sse_regs_t  xmm[16];
-    uint32_t    mxcsr;
     // defered flags
+    int         dummy1;     // to align on 64bits with df
     defered_flags_t df;
     uint64_t    op1;
     uint64_t    op2;
@@ -50,6 +51,13 @@ typedef struct x64emu_s {
     uint32_t    segs[6];        // only 32bits value?
     uintptr_t   segs_offs[6];   // computed offset associate with segment
     uint64_t    segs_serial[6];  // are seg offset clean (not 0) or does they need to be re-computed (0)? For GS, serial need to be the same as context->sel_serial
+    // parent context
+    box64context_t *context;
+    // cpu helpers
+    reg64_t     zero;
+    reg64_t     *sbiidx[16];
+    // scratch stack, used for alignement of double and 64bits ints on arm. 200 elements should be enough
+    uint64_t    scratch[200];
     // emu control
     int         quit;
     int         error;
@@ -58,13 +66,6 @@ typedef struct x64emu_s {
     int         exit;
     int         quitonlongjmp;  // quit if longjmp is called
     int         longjmp;        // if quit because of longjmp
-    // parent context
-    box64context_t *context;
-    // cpu helpers
-    reg64_t     zero;
-    reg64_t     *sbiidx[16];
-    // scratch stack, used for alignement of double and 64bits ints on arm. 200 elements should be enough
-    uint64_t    scratch[200];
     // local stack, do be deleted when emu is freed
     void*       stack2free; // this is the stack to free (can be NULL)
     void*       init_stack; // initial stack (owned or not)