summary refs log tree commit diff stats
path: root/tcg/tci.c
diff options
context:
space:
mode:
Diffstat (limited to 'tcg/tci.c')
-rw-r--r--tcg/tci.c1099
1 files changed, 739 insertions, 360 deletions
diff --git a/tcg/tci.c b/tcg/tci.c
index 3ccd30c39c..d68c5a4e55 100644
--- a/tcg/tci.c
+++ b/tcg/tci.c
@@ -24,7 +24,7 @@
 #if defined(CONFIG_DEBUG_TCG)
 # define tci_assert(cond) assert(cond)
 #else
-# define tci_assert(cond) ((void)0)
+# define tci_assert(cond) ((void)(cond))
 #endif
 
 #include "qemu-common.h"
@@ -66,22 +66,32 @@ tci_write_reg(tcg_target_ulong *regs, TCGReg index, tcg_target_ulong value)
     regs[index] = value;
 }
 
-#if TCG_TARGET_REG_BITS == 32
 static void tci_write_reg64(tcg_target_ulong *regs, uint32_t high_index,
                             uint32_t low_index, uint64_t value)
 {
     tci_write_reg(regs, low_index, value);
     tci_write_reg(regs, high_index, value >> 32);
 }
-#endif
 
-#if TCG_TARGET_REG_BITS == 32
 /* Create a 64 bit value from two 32 bit values. */
 static uint64_t tci_uint64(uint32_t high, uint32_t low)
 {
     return ((uint64_t)high << 32) + low;
 }
-#endif
+
+/* Read constant byte from bytecode. */
+static uint8_t tci_read_b(const uint8_t **tb_ptr)
+{
+    return *(tb_ptr[0]++);
+}
+
+/* Read register number from bytecode. */
+static TCGReg tci_read_r(const uint8_t **tb_ptr)
+{
+    uint8_t regno = tci_read_b(tb_ptr);
+    tci_assert(regno < TCG_TARGET_NB_REGS);
+    return regno;
+}
 
 /* Read constant (native size) from bytecode. */
 static tcg_target_ulong tci_read_i(const uint8_t **tb_ptr)
@@ -107,59 +117,239 @@ static int32_t tci_read_s32(const uint8_t **tb_ptr)
     return value;
 }
 
+static tcg_target_ulong tci_read_label(const uint8_t **tb_ptr)
+{
+    return tci_read_i(tb_ptr);
+}
+
+/*
+ * Load sets of arguments all at once.  The naming convention is:
+ *   tci_args_<arguments>
+ * where arguments is a sequence of
+ *
+ *   b = immediate (bit position)
+ *   c = condition (TCGCond)
+ *   i = immediate (uint32_t)
+ *   I = immediate (tcg_target_ulong)
+ *   l = label or pointer
+ *   m = immediate (TCGMemOpIdx)
+ *   r = register
+ *   s = signed ldst offset
+ */
+
+static void check_size(const uint8_t *start, const uint8_t **tb_ptr)
+{
+    const uint8_t *old_code_ptr = start - 2;
+    uint8_t op_size = old_code_ptr[1];
+    tci_assert(*tb_ptr == old_code_ptr + op_size);
+}
+
+static void tci_args_l(const uint8_t **tb_ptr, void **l0)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *l0 = (void *)tci_read_label(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
+static void tci_args_rr(const uint8_t **tb_ptr,
+                        TCGReg *r0, TCGReg *r1)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
+static void tci_args_ri(const uint8_t **tb_ptr,
+                        TCGReg *r0, tcg_target_ulong *i1)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *i1 = tci_read_i32(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
 #if TCG_TARGET_REG_BITS == 64
-/* Read constant (64 bit) from bytecode. */
-static uint64_t tci_read_i64(const uint8_t **tb_ptr)
+static void tci_args_rI(const uint8_t **tb_ptr,
+                        TCGReg *r0, tcg_target_ulong *i1)
 {
-    uint64_t value = *(const uint64_t *)(*tb_ptr);
-    *tb_ptr += sizeof(value);
-    return value;
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *i1 = tci_read_i(tb_ptr);
+
+    check_size(start, tb_ptr);
 }
 #endif
 
-/* Read indexed register (native size) from bytecode. */
-static tcg_target_ulong
-tci_read_r(const tcg_target_ulong *regs, const uint8_t **tb_ptr)
+static void tci_args_rrm(const uint8_t **tb_ptr,
+                         TCGReg *r0, TCGReg *r1, TCGMemOpIdx *m2)
 {
-    tcg_target_ulong value = tci_read_reg(regs, **tb_ptr);
-    *tb_ptr += 1;
-    return value;
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *m2 = tci_read_i32(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
+static void tci_args_rrr(const uint8_t **tb_ptr,
+                         TCGReg *r0, TCGReg *r1, TCGReg *r2)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *r2 = tci_read_r(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
+static void tci_args_rrs(const uint8_t **tb_ptr,
+                         TCGReg *r0, TCGReg *r1, int32_t *i2)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *i2 = tci_read_s32(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
+static void tci_args_rrcl(const uint8_t **tb_ptr,
+                          TCGReg *r0, TCGReg *r1, TCGCond *c2, void **l3)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *c2 = tci_read_b(tb_ptr);
+    *l3 = (void *)tci_read_label(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
+static void tci_args_rrrc(const uint8_t **tb_ptr,
+                          TCGReg *r0, TCGReg *r1, TCGReg *r2, TCGCond *c3)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *r2 = tci_read_r(tb_ptr);
+    *c3 = tci_read_b(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
+static void tci_args_rrrm(const uint8_t **tb_ptr,
+                          TCGReg *r0, TCGReg *r1, TCGReg *r2, TCGMemOpIdx *m3)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *r2 = tci_read_r(tb_ptr);
+    *m3 = tci_read_i32(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
+static void tci_args_rrrbb(const uint8_t **tb_ptr, TCGReg *r0, TCGReg *r1,
+                           TCGReg *r2, uint8_t *i3, uint8_t *i4)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *r2 = tci_read_r(tb_ptr);
+    *i3 = tci_read_b(tb_ptr);
+    *i4 = tci_read_b(tb_ptr);
+
+    check_size(start, tb_ptr);
+}
+
+static void tci_args_rrrrm(const uint8_t **tb_ptr, TCGReg *r0, TCGReg *r1,
+                           TCGReg *r2, TCGReg *r3, TCGMemOpIdx *m4)
+{
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *r2 = tci_read_r(tb_ptr);
+    *r3 = tci_read_r(tb_ptr);
+    *m4 = tci_read_i32(tb_ptr);
+
+    check_size(start, tb_ptr);
 }
 
 #if TCG_TARGET_REG_BITS == 32
-/* Read two indexed registers (2 * 32 bit) from bytecode. */
-static uint64_t tci_read_r64(const tcg_target_ulong *regs,
-                             const uint8_t **tb_ptr)
+static void tci_args_rrrr(const uint8_t **tb_ptr,
+                          TCGReg *r0, TCGReg *r1, TCGReg *r2, TCGReg *r3)
 {
-    uint32_t low = tci_read_r(regs, tb_ptr);
-    return tci_uint64(tci_read_r(regs, tb_ptr), low);
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *r2 = tci_read_r(tb_ptr);
+    *r3 = tci_read_r(tb_ptr);
+
+    check_size(start, tb_ptr);
 }
-#elif TCG_TARGET_REG_BITS == 64
-/* Read indexed register (64 bit) from bytecode. */
-static uint64_t tci_read_r64(const tcg_target_ulong *regs,
-                             const uint8_t **tb_ptr)
+
+static void tci_args_rrrrcl(const uint8_t **tb_ptr, TCGReg *r0, TCGReg *r1,
+                            TCGReg *r2, TCGReg *r3, TCGCond *c4, void **l5)
 {
-    return tci_read_r(regs, tb_ptr);
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *r2 = tci_read_r(tb_ptr);
+    *r3 = tci_read_r(tb_ptr);
+    *c4 = tci_read_b(tb_ptr);
+    *l5 = (void *)tci_read_label(tb_ptr);
+
+    check_size(start, tb_ptr);
 }
-#endif
 
-/* Read indexed register(s) with target address from bytecode. */
-static target_ulong
-tci_read_ulong(const tcg_target_ulong *regs, const uint8_t **tb_ptr)
+static void tci_args_rrrrrc(const uint8_t **tb_ptr, TCGReg *r0, TCGReg *r1,
+                            TCGReg *r2, TCGReg *r3, TCGReg *r4, TCGCond *c5)
 {
-    target_ulong taddr = tci_read_r(regs, tb_ptr);
-#if TARGET_LONG_BITS > TCG_TARGET_REG_BITS
-    taddr += (uint64_t)tci_read_r(regs, tb_ptr) << 32;
-#endif
-    return taddr;
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *r2 = tci_read_r(tb_ptr);
+    *r3 = tci_read_r(tb_ptr);
+    *r4 = tci_read_r(tb_ptr);
+    *c5 = tci_read_b(tb_ptr);
+
+    check_size(start, tb_ptr);
 }
 
-static tcg_target_ulong tci_read_label(const uint8_t **tb_ptr)
+static void tci_args_rrrrrr(const uint8_t **tb_ptr, TCGReg *r0, TCGReg *r1,
+                            TCGReg *r2, TCGReg *r3, TCGReg *r4, TCGReg *r5)
 {
-    tcg_target_ulong label = tci_read_i(tb_ptr);
-    tci_assert(label != 0);
-    return label;
+    const uint8_t *start = *tb_ptr;
+
+    *r0 = tci_read_r(tb_ptr);
+    *r1 = tci_read_r(tb_ptr);
+    *r2 = tci_read_r(tb_ptr);
+    *r3 = tci_read_r(tb_ptr);
+    *r4 = tci_read_r(tb_ptr);
+    *r5 = tci_read_r(tb_ptr);
+
+    check_size(start, tb_ptr);
 }
+#endif
 
 static bool tci_compare32(uint32_t u0, uint32_t u1, TCGCond condition)
 {
@@ -299,7 +489,6 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
     tcg_target_ulong regs[TCG_TARGET_NB_REGS];
     long tcg_temps[CPU_TEMP_BUF_NLONGS];
     uintptr_t sp_value = (uintptr_t)(tcg_temps + CPU_TEMP_BUF_NLONGS);
-    uintptr_t ret = 0;
 
     regs[TCG_AREG0] = (tcg_target_ulong)env;
     regs[TCG_REG_CALL_STACK] = sp_value;
@@ -307,494 +496,378 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
 
     for (;;) {
         TCGOpcode opc = tb_ptr[0];
-#if defined(CONFIG_DEBUG_TCG) && !defined(NDEBUG)
-        uint8_t op_size = tb_ptr[1];
-        const uint8_t *old_code_ptr = tb_ptr;
-#endif
-        tcg_target_ulong t0;
+        TCGReg r0, r1, r2, r3;
         tcg_target_ulong t1;
-        tcg_target_ulong t2;
-        tcg_target_ulong label;
         TCGCond condition;
         target_ulong taddr;
-        uint8_t tmp8;
-        uint16_t tmp16;
+        uint8_t pos, len;
         uint32_t tmp32;
         uint64_t tmp64;
 #if TCG_TARGET_REG_BITS == 32
-        uint64_t v64;
+        TCGReg r4, r5;
+        uint64_t T1, T2;
 #endif
         TCGMemOpIdx oi;
+        int32_t ofs;
+        void *ptr;
 
         /* Skip opcode and size entry. */
         tb_ptr += 2;
 
         switch (opc) {
         case INDEX_op_call:
-            t0 = tci_read_i(&tb_ptr);
+            tci_args_l(&tb_ptr, &ptr);
             tci_tb_ptr = (uintptr_t)tb_ptr;
 #if TCG_TARGET_REG_BITS == 32
-            tmp64 = ((helper_function)t0)(tci_read_reg(regs, TCG_REG_R0),
-                                          tci_read_reg(regs, TCG_REG_R1),
-                                          tci_read_reg(regs, TCG_REG_R2),
-                                          tci_read_reg(regs, TCG_REG_R3),
-                                          tci_read_reg(regs, TCG_REG_R4),
-                                          tci_read_reg(regs, TCG_REG_R5),
-                                          tci_read_reg(regs, TCG_REG_R6),
-                                          tci_read_reg(regs, TCG_REG_R7),
-                                          tci_read_reg(regs, TCG_REG_R8),
-                                          tci_read_reg(regs, TCG_REG_R9),
-                                          tci_read_reg(regs, TCG_REG_R10),
-                                          tci_read_reg(regs, TCG_REG_R11));
+            tmp64 = ((helper_function)ptr)(tci_read_reg(regs, TCG_REG_R0),
+                                           tci_read_reg(regs, TCG_REG_R1),
+                                           tci_read_reg(regs, TCG_REG_R2),
+                                           tci_read_reg(regs, TCG_REG_R3),
+                                           tci_read_reg(regs, TCG_REG_R4),
+                                           tci_read_reg(regs, TCG_REG_R5),
+                                           tci_read_reg(regs, TCG_REG_R6),
+                                           tci_read_reg(regs, TCG_REG_R7),
+                                           tci_read_reg(regs, TCG_REG_R8),
+                                           tci_read_reg(regs, TCG_REG_R9),
+                                           tci_read_reg(regs, TCG_REG_R10),
+                                           tci_read_reg(regs, TCG_REG_R11));
             tci_write_reg(regs, TCG_REG_R0, tmp64);
             tci_write_reg(regs, TCG_REG_R1, tmp64 >> 32);
 #else
-            tmp64 = ((helper_function)t0)(tci_read_reg(regs, TCG_REG_R0),
-                                          tci_read_reg(regs, TCG_REG_R1),
-                                          tci_read_reg(regs, TCG_REG_R2),
-                                          tci_read_reg(regs, TCG_REG_R3),
-                                          tci_read_reg(regs, TCG_REG_R4),
-                                          tci_read_reg(regs, TCG_REG_R5));
+            tmp64 = ((helper_function)ptr)(tci_read_reg(regs, TCG_REG_R0),
+                                           tci_read_reg(regs, TCG_REG_R1),
+                                           tci_read_reg(regs, TCG_REG_R2),
+                                           tci_read_reg(regs, TCG_REG_R3),
+                                           tci_read_reg(regs, TCG_REG_R4),
+                                           tci_read_reg(regs, TCG_REG_R5));
             tci_write_reg(regs, TCG_REG_R0, tmp64);
 #endif
             break;
         case INDEX_op_br:
-            label = tci_read_label(&tb_ptr);
-            tci_assert(tb_ptr == old_code_ptr + op_size);
-            tb_ptr = (uint8_t *)label;
+            tci_args_l(&tb_ptr, &ptr);
+            tb_ptr = ptr;
             continue;
         case INDEX_op_setcond_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            condition = *tb_ptr++;
-            tci_write_reg(regs, t0, tci_compare32(t1, t2, condition));
+            tci_args_rrrc(&tb_ptr, &r0, &r1, &r2, &condition);
+            regs[r0] = tci_compare32(regs[r1], regs[r2], condition);
             break;
 #if TCG_TARGET_REG_BITS == 32
         case INDEX_op_setcond2_i32:
-            t0 = *tb_ptr++;
-            tmp64 = tci_read_r64(regs, &tb_ptr);
-            v64 = tci_read_r64(regs, &tb_ptr);
-            condition = *tb_ptr++;
-            tci_write_reg(regs, t0, tci_compare64(tmp64, v64, condition));
+            tci_args_rrrrrc(&tb_ptr, &r0, &r1, &r2, &r3, &r4, &condition);
+            T1 = tci_uint64(regs[r2], regs[r1]);
+            T2 = tci_uint64(regs[r4], regs[r3]);
+            regs[r0] = tci_compare64(T1, T2, condition);
             break;
 #elif TCG_TARGET_REG_BITS == 64
         case INDEX_op_setcond_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            condition = *tb_ptr++;
-            tci_write_reg(regs, t0, tci_compare64(t1, t2, condition));
+            tci_args_rrrc(&tb_ptr, &r0, &r1, &r2, &condition);
+            regs[r0] = tci_compare64(regs[r1], regs[r2], condition);
             break;
 #endif
         CASE_32_64(mov)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, t1);
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = regs[r1];
             break;
         case INDEX_op_tci_movi_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_i32(&tb_ptr);
-            tci_write_reg(regs, t0, t1);
+            tci_args_ri(&tb_ptr, &r0, &t1);
+            regs[r0] = t1;
             break;
 
             /* Load/store operations (32 bit). */
 
         CASE_32_64(ld8u)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            tci_write_reg(regs, t0, *(uint8_t *)(t1 + t2));
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            regs[r0] = *(uint8_t *)ptr;
             break;
         CASE_32_64(ld8s)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            tci_write_reg(regs, t0, *(int8_t *)(t1 + t2));
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            regs[r0] = *(int8_t *)ptr;
             break;
         CASE_32_64(ld16u)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            tci_write_reg(regs, t0, *(uint16_t *)(t1 + t2));
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            regs[r0] = *(uint16_t *)ptr;
             break;
         CASE_32_64(ld16s)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            tci_write_reg(regs, t0, *(int16_t *)(t1 + t2));
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            regs[r0] = *(int16_t *)ptr;
             break;
         case INDEX_op_ld_i32:
         CASE_64(ld32u)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            tci_write_reg(regs, t0, *(uint32_t *)(t1 + t2));
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            regs[r0] = *(uint32_t *)ptr;
             break;
         CASE_32_64(st8)
-            t0 = tci_read_r(regs, &tb_ptr);
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            *(uint8_t *)(t1 + t2) = t0;
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            *(uint8_t *)ptr = regs[r0];
             break;
         CASE_32_64(st16)
-            t0 = tci_read_r(regs, &tb_ptr);
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            *(uint16_t *)(t1 + t2) = t0;
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            *(uint16_t *)ptr = regs[r0];
             break;
         case INDEX_op_st_i32:
         CASE_64(st32)
-            t0 = tci_read_r(regs, &tb_ptr);
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            *(uint32_t *)(t1 + t2) = t0;
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            *(uint32_t *)ptr = regs[r0];
             break;
 
             /* Arithmetic operations (mixed 32/64 bit). */
 
         CASE_32_64(add)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, t1 + t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = regs[r1] + regs[r2];
             break;
         CASE_32_64(sub)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, t1 - t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = regs[r1] - regs[r2];
             break;
         CASE_32_64(mul)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, t1 * t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = regs[r1] * regs[r2];
             break;
         CASE_32_64(and)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, t1 & t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = regs[r1] & regs[r2];
             break;
         CASE_32_64(or)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, t1 | t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = regs[r1] | regs[r2];
             break;
         CASE_32_64(xor)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, t1 ^ t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = regs[r1] ^ regs[r2];
             break;
 
             /* Arithmetic operations (32 bit). */
 
         case INDEX_op_div_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (int32_t)t1 / (int32_t)t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (int32_t)regs[r1] / (int32_t)regs[r2];
             break;
         case INDEX_op_divu_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (uint32_t)t1 / (uint32_t)t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (uint32_t)regs[r1] / (uint32_t)regs[r2];
             break;
         case INDEX_op_rem_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (int32_t)t1 % (int32_t)t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (int32_t)regs[r1] % (int32_t)regs[r2];
             break;
         case INDEX_op_remu_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (uint32_t)t1 % (uint32_t)t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (uint32_t)regs[r1] % (uint32_t)regs[r2];
             break;
 
             /* Shift/rotate operations (32 bit). */
 
         case INDEX_op_shl_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (uint32_t)t1 << (t2 & 31));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (uint32_t)regs[r1] << (regs[r2] & 31);
             break;
         case INDEX_op_shr_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (uint32_t)t1 >> (t2 & 31));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (uint32_t)regs[r1] >> (regs[r2] & 31);
             break;
         case INDEX_op_sar_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (int32_t)t1 >> (t2 & 31));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (int32_t)regs[r1] >> (regs[r2] & 31);
             break;
 #if TCG_TARGET_HAS_rot_i32
         case INDEX_op_rotl_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, rol32(t1, t2 & 31));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = rol32(regs[r1], regs[r2] & 31);
             break;
         case INDEX_op_rotr_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, ror32(t1, t2 & 31));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = ror32(regs[r1], regs[r2] & 31);
             break;
 #endif
 #if TCG_TARGET_HAS_deposit_i32
         case INDEX_op_deposit_i32:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tmp16 = *tb_ptr++;
-            tmp8 = *tb_ptr++;
-            tmp32 = (((1 << tmp8) - 1) << tmp16);
-            tci_write_reg(regs, t0, (t1 & ~tmp32) | ((t2 << tmp16) & tmp32));
+            tci_args_rrrbb(&tb_ptr, &r0, &r1, &r2, &pos, &len);
+            regs[r0] = deposit32(regs[r1], pos, len, regs[r2]);
             break;
 #endif
         case INDEX_op_brcond_i32:
-            t0 = tci_read_r(regs, &tb_ptr);
-            t1 = tci_read_r(regs, &tb_ptr);
-            condition = *tb_ptr++;
-            label = tci_read_label(&tb_ptr);
-            if (tci_compare32(t0, t1, condition)) {
-                tci_assert(tb_ptr == old_code_ptr + op_size);
-                tb_ptr = (uint8_t *)label;
-                continue;
+            tci_args_rrcl(&tb_ptr, &r0, &r1, &condition, &ptr);
+            if (tci_compare32(regs[r0], regs[r1], condition)) {
+                tb_ptr = ptr;
             }
             break;
 #if TCG_TARGET_REG_BITS == 32
         case INDEX_op_add2_i32:
-            t0 = *tb_ptr++;
-            t1 = *tb_ptr++;
-            tmp64 = tci_read_r64(regs, &tb_ptr);
-            tmp64 += tci_read_r64(regs, &tb_ptr);
-            tci_write_reg64(regs, t1, t0, tmp64);
+            tci_args_rrrrrr(&tb_ptr, &r0, &r1, &r2, &r3, &r4, &r5);
+            T1 = tci_uint64(regs[r3], regs[r2]);
+            T2 = tci_uint64(regs[r5], regs[r4]);
+            tci_write_reg64(regs, r1, r0, T1 + T2);
             break;
         case INDEX_op_sub2_i32:
-            t0 = *tb_ptr++;
-            t1 = *tb_ptr++;
-            tmp64 = tci_read_r64(regs, &tb_ptr);
-            tmp64 -= tci_read_r64(regs, &tb_ptr);
-            tci_write_reg64(regs, t1, t0, tmp64);
+            tci_args_rrrrrr(&tb_ptr, &r0, &r1, &r2, &r3, &r4, &r5);
+            T1 = tci_uint64(regs[r3], regs[r2]);
+            T2 = tci_uint64(regs[r5], regs[r4]);
+            tci_write_reg64(regs, r1, r0, T1 - T2);
             break;
         case INDEX_op_brcond2_i32:
-            tmp64 = tci_read_r64(regs, &tb_ptr);
-            v64 = tci_read_r64(regs, &tb_ptr);
-            condition = *tb_ptr++;
-            label = tci_read_label(&tb_ptr);
-            if (tci_compare64(tmp64, v64, condition)) {
-                tci_assert(tb_ptr == old_code_ptr + op_size);
-                tb_ptr = (uint8_t *)label;
+            tci_args_rrrrcl(&tb_ptr, &r0, &r1, &r2, &r3, &condition, &ptr);
+            T1 = tci_uint64(regs[r1], regs[r0]);
+            T2 = tci_uint64(regs[r3], regs[r2]);
+            if (tci_compare64(T1, T2, condition)) {
+                tb_ptr = ptr;
                 continue;
             }
             break;
         case INDEX_op_mulu2_i32:
-            t0 = *tb_ptr++;
-            t1 = *tb_ptr++;
-            t2 = tci_read_r(regs, &tb_ptr);
-            tmp64 = (uint32_t)tci_read_r(regs, &tb_ptr);
-            tci_write_reg64(regs, t1, t0, (uint32_t)t2 * tmp64);
+            tci_args_rrrr(&tb_ptr, &r0, &r1, &r2, &r3);
+            tci_write_reg64(regs, r1, r0, (uint64_t)regs[r2] * regs[r3]);
             break;
 #endif /* TCG_TARGET_REG_BITS == 32 */
 #if TCG_TARGET_HAS_ext8s_i32 || TCG_TARGET_HAS_ext8s_i64
         CASE_32_64(ext8s)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (int8_t)t1);
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = (int8_t)regs[r1];
             break;
 #endif
 #if TCG_TARGET_HAS_ext16s_i32 || TCG_TARGET_HAS_ext16s_i64
         CASE_32_64(ext16s)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (int16_t)t1);
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = (int16_t)regs[r1];
             break;
 #endif
 #if TCG_TARGET_HAS_ext8u_i32 || TCG_TARGET_HAS_ext8u_i64
         CASE_32_64(ext8u)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (uint8_t)t1);
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = (uint8_t)regs[r1];
             break;
 #endif
 #if TCG_TARGET_HAS_ext16u_i32 || TCG_TARGET_HAS_ext16u_i64
         CASE_32_64(ext16u)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (uint16_t)t1);
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = (uint16_t)regs[r1];
             break;
 #endif
 #if TCG_TARGET_HAS_bswap16_i32 || TCG_TARGET_HAS_bswap16_i64
         CASE_32_64(bswap16)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, bswap16(t1));
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = bswap16(regs[r1]);
             break;
 #endif
 #if TCG_TARGET_HAS_bswap32_i32 || TCG_TARGET_HAS_bswap32_i64
         CASE_32_64(bswap32)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, bswap32(t1));
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = bswap32(regs[r1]);
             break;
 #endif
 #if TCG_TARGET_HAS_not_i32 || TCG_TARGET_HAS_not_i64
         CASE_32_64(not)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, ~t1);
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = ~regs[r1];
             break;
 #endif
 #if TCG_TARGET_HAS_neg_i32 || TCG_TARGET_HAS_neg_i64
         CASE_32_64(neg)
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, -t1);
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = -regs[r1];
             break;
 #endif
 #if TCG_TARGET_REG_BITS == 64
         case INDEX_op_tci_movi_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_i64(&tb_ptr);
-            tci_write_reg(regs, t0, t1);
+            tci_args_rI(&tb_ptr, &r0, &t1);
+            regs[r0] = t1;
             break;
 
             /* Load/store operations (64 bit). */
 
         case INDEX_op_ld32s_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            tci_write_reg(regs, t0, *(int32_t *)(t1 + t2));
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            regs[r0] = *(int32_t *)ptr;
             break;
         case INDEX_op_ld_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            tci_write_reg(regs, t0, *(uint64_t *)(t1 + t2));
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            regs[r0] = *(uint64_t *)ptr;
             break;
         case INDEX_op_st_i64:
-            t0 = tci_read_r(regs, &tb_ptr);
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_s32(&tb_ptr);
-            *(uint64_t *)(t1 + t2) = t0;
+            tci_args_rrs(&tb_ptr, &r0, &r1, &ofs);
+            ptr = (void *)(regs[r1] + ofs);
+            *(uint64_t *)ptr = regs[r0];
             break;
 
             /* Arithmetic operations (64 bit). */
 
         case INDEX_op_div_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (int64_t)t1 / (int64_t)t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (int64_t)regs[r1] / (int64_t)regs[r2];
             break;
         case INDEX_op_divu_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (uint64_t)t1 / (uint64_t)t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (uint64_t)regs[r1] / (uint64_t)regs[r2];
             break;
         case INDEX_op_rem_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (int64_t)t1 % (int64_t)t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (int64_t)regs[r1] % (int64_t)regs[r2];
             break;
         case INDEX_op_remu_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (uint64_t)t1 % (uint64_t)t2);
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (uint64_t)regs[r1] % (uint64_t)regs[r2];
             break;
 
             /* Shift/rotate operations (64 bit). */
 
         case INDEX_op_shl_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, t1 << (t2 & 63));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = regs[r1] << (regs[r2] & 63);
             break;
         case INDEX_op_shr_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, t1 >> (t2 & 63));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = regs[r1] >> (regs[r2] & 63);
             break;
         case INDEX_op_sar_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, ((int64_t)t1 >> (t2 & 63)));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = (int64_t)regs[r1] >> (regs[r2] & 63);
             break;
 #if TCG_TARGET_HAS_rot_i64
         case INDEX_op_rotl_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, rol64(t1, t2 & 63));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = rol64(regs[r1], regs[r2] & 63);
             break;
         case INDEX_op_rotr_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, ror64(t1, t2 & 63));
+            tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+            regs[r0] = ror64(regs[r1], regs[r2] & 63);
             break;
 #endif
 #if TCG_TARGET_HAS_deposit_i64
         case INDEX_op_deposit_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            t2 = tci_read_r(regs, &tb_ptr);
-            tmp16 = *tb_ptr++;
-            tmp8 = *tb_ptr++;
-            tmp64 = (((1ULL << tmp8) - 1) << tmp16);
-            tci_write_reg(regs, t0, (t1 & ~tmp64) | ((t2 << tmp16) & tmp64));
+            tci_args_rrrbb(&tb_ptr, &r0, &r1, &r2, &pos, &len);
+            regs[r0] = deposit64(regs[r1], pos, len, regs[r2]);
             break;
 #endif
         case INDEX_op_brcond_i64:
-            t0 = tci_read_r(regs, &tb_ptr);
-            t1 = tci_read_r(regs, &tb_ptr);
-            condition = *tb_ptr++;
-            label = tci_read_label(&tb_ptr);
-            if (tci_compare64(t0, t1, condition)) {
-                tci_assert(tb_ptr == old_code_ptr + op_size);
-                tb_ptr = (uint8_t *)label;
-                continue;
+            tci_args_rrcl(&tb_ptr, &r0, &r1, &condition, &ptr);
+            if (tci_compare64(regs[r0], regs[r1], condition)) {
+                tb_ptr = ptr;
             }
             break;
-#if TCG_TARGET_HAS_ext32s_i64
         case INDEX_op_ext32s_i64:
-#endif
         case INDEX_op_ext_i32_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (int32_t)t1);
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = (int32_t)regs[r1];
             break;
-#if TCG_TARGET_HAS_ext32u_i64
         case INDEX_op_ext32u_i64:
-#endif
         case INDEX_op_extu_i32_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, (uint32_t)t1);
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = (uint32_t)regs[r1];
             break;
 #if TCG_TARGET_HAS_bswap64_i64
         case INDEX_op_bswap64_i64:
-            t0 = *tb_ptr++;
-            t1 = tci_read_r(regs, &tb_ptr);
-            tci_write_reg(regs, t0, bswap64(t1));
+            tci_args_rr(&tb_ptr, &r0, &r1);
+            regs[r0] = bswap64(regs[r1]);
             break;
 #endif
 #endif /* TCG_TARGET_REG_BITS == 64 */
@@ -802,21 +875,22 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
             /* QEMU specific operations. */
 
         case INDEX_op_exit_tb:
-            ret = *(uint64_t *)tb_ptr;
-            goto exit;
-            break;
+            tci_args_l(&tb_ptr, &ptr);
+            return (uintptr_t)ptr;
+
         case INDEX_op_goto_tb:
-            /* Jump address is aligned */
-            tb_ptr = QEMU_ALIGN_PTR_UP(tb_ptr, 4);
-            t0 = qatomic_read((int32_t *)tb_ptr);
-            tb_ptr += sizeof(int32_t);
-            tci_assert(tb_ptr == old_code_ptr + op_size);
-            tb_ptr += (int32_t)t0;
-            continue;
+            tci_args_l(&tb_ptr, &ptr);
+            tb_ptr = *(void **)ptr;
+            break;
+
         case INDEX_op_qemu_ld_i32:
-            t0 = *tb_ptr++;
-            taddr = tci_read_ulong(regs, &tb_ptr);
-            oi = tci_read_i(&tb_ptr);
+            if (TARGET_LONG_BITS <= TCG_TARGET_REG_BITS) {
+                tci_args_rrm(&tb_ptr, &r0, &r1, &oi);
+                taddr = regs[r1];
+            } else {
+                tci_args_rrrm(&tb_ptr, &r0, &r1, &r2, &oi);
+                taddr = tci_uint64(regs[r2], regs[r1]);
+            }
             switch (get_memop(oi) & (MO_BSWAP | MO_SSIZE)) {
             case MO_UB:
                 tmp32 = qemu_ld_ub;
@@ -845,15 +919,20 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
             default:
                 g_assert_not_reached();
             }
-            tci_write_reg(regs, t0, tmp32);
+            regs[r0] = tmp32;
             break;
+
         case INDEX_op_qemu_ld_i64:
-            t0 = *tb_ptr++;
-            if (TCG_TARGET_REG_BITS == 32) {
-                t1 = *tb_ptr++;
+            if (TCG_TARGET_REG_BITS == 64) {
+                tci_args_rrm(&tb_ptr, &r0, &r1, &oi);
+                taddr = regs[r1];
+            } else if (TARGET_LONG_BITS <= TCG_TARGET_REG_BITS) {
+                tci_args_rrrm(&tb_ptr, &r0, &r1, &r2, &oi);
+                taddr = regs[r2];
+            } else {
+                tci_args_rrrrm(&tb_ptr, &r0, &r1, &r2, &r3, &oi);
+                taddr = tci_uint64(regs[r3], regs[r2]);
             }
-            taddr = tci_read_ulong(regs, &tb_ptr);
-            oi = tci_read_i(&tb_ptr);
             switch (get_memop(oi) & (MO_BSWAP | MO_SSIZE)) {
             case MO_UB:
                 tmp64 = qemu_ld_ub;
@@ -894,39 +973,58 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
             default:
                 g_assert_not_reached();
             }
-            tci_write_reg(regs, t0, tmp64);
             if (TCG_TARGET_REG_BITS == 32) {
-                tci_write_reg(regs, t1, tmp64 >> 32);
+                tci_write_reg64(regs, r1, r0, tmp64);
+            } else {
+                regs[r0] = tmp64;
             }
             break;
+
         case INDEX_op_qemu_st_i32:
-            t0 = tci_read_r(regs, &tb_ptr);
-            taddr = tci_read_ulong(regs, &tb_ptr);
-            oi = tci_read_i(&tb_ptr);
+            if (TARGET_LONG_BITS <= TCG_TARGET_REG_BITS) {
+                tci_args_rrm(&tb_ptr, &r0, &r1, &oi);
+                taddr = regs[r1];
+            } else {
+                tci_args_rrrm(&tb_ptr, &r0, &r1, &r2, &oi);
+                taddr = tci_uint64(regs[r2], regs[r1]);
+            }
+            tmp32 = regs[r0];
             switch (get_memop(oi) & (MO_BSWAP | MO_SIZE)) {
             case MO_UB:
-                qemu_st_b(t0);
+                qemu_st_b(tmp32);
                 break;
             case MO_LEUW:
-                qemu_st_lew(t0);
+                qemu_st_lew(tmp32);
                 break;
             case MO_LEUL:
-                qemu_st_lel(t0);
+                qemu_st_lel(tmp32);
                 break;
             case MO_BEUW:
-                qemu_st_bew(t0);
+                qemu_st_bew(tmp32);
                 break;
             case MO_BEUL:
-                qemu_st_bel(t0);
+                qemu_st_bel(tmp32);
                 break;
             default:
                 g_assert_not_reached();
             }
             break;
+
         case INDEX_op_qemu_st_i64:
-            tmp64 = tci_read_r64(regs, &tb_ptr);
-            taddr = tci_read_ulong(regs, &tb_ptr);
-            oi = tci_read_i(&tb_ptr);
+            if (TCG_TARGET_REG_BITS == 64) {
+                tci_args_rrm(&tb_ptr, &r0, &r1, &oi);
+                taddr = regs[r1];
+                tmp64 = regs[r0];
+            } else {
+                if (TARGET_LONG_BITS <= TCG_TARGET_REG_BITS) {
+                    tci_args_rrrm(&tb_ptr, &r0, &r1, &r2, &oi);
+                    taddr = regs[r2];
+                } else {
+                    tci_args_rrrrm(&tb_ptr, &r0, &r1, &r2, &r3, &oi);
+                    taddr = tci_uint64(regs[r3], regs[r2]);
+                }
+                tmp64 = tci_uint64(regs[r1], regs[r0]);
+            }
             switch (get_memop(oi) & (MO_BSWAP | MO_SIZE)) {
             case MO_UB:
                 qemu_st_b(tmp64);
@@ -953,6 +1051,7 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
                 g_assert_not_reached();
             }
             break;
+
         case INDEX_op_mb:
             /* Ensure ordering for all kinds */
             smp_mb();
@@ -960,8 +1059,288 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
         default:
             g_assert_not_reached();
         }
-        tci_assert(tb_ptr == old_code_ptr + op_size);
     }
-exit:
-    return ret;
+}
+
+/*
+ * Disassembler that matches the interpreter
+ */
+
+static const char *str_r(TCGReg r)
+{
+    static const char regs[TCG_TARGET_NB_REGS][4] = {
+        "r0", "r1", "r2",  "r3",  "r4",  "r5",  "r6",  "r7",
+        "r8", "r9", "r10", "r11", "r12", "r13", "env", "sp"
+    };
+
+    QEMU_BUILD_BUG_ON(TCG_AREG0 != TCG_REG_R14);
+    QEMU_BUILD_BUG_ON(TCG_REG_CALL_STACK != TCG_REG_R15);
+
+    assert((unsigned)r < TCG_TARGET_NB_REGS);
+    return regs[r];
+}
+
+static const char *str_c(TCGCond c)
+{
+    static const char cond[16][8] = {
+        [TCG_COND_NEVER] = "never",
+        [TCG_COND_ALWAYS] = "always",
+        [TCG_COND_EQ] = "eq",
+        [TCG_COND_NE] = "ne",
+        [TCG_COND_LT] = "lt",
+        [TCG_COND_GE] = "ge",
+        [TCG_COND_LE] = "le",
+        [TCG_COND_GT] = "gt",
+        [TCG_COND_LTU] = "ltu",
+        [TCG_COND_GEU] = "geu",
+        [TCG_COND_LEU] = "leu",
+        [TCG_COND_GTU] = "gtu",
+    };
+
+    assert((unsigned)c < ARRAY_SIZE(cond));
+    assert(cond[c][0] != 0);
+    return cond[c];
+}
+
+/* Disassemble TCI bytecode. */
+int print_insn_tci(bfd_vma addr, disassemble_info *info)
+{
+    uint8_t buf[256];
+    int length, status;
+    const TCGOpDef *def;
+    const char *op_name;
+    TCGOpcode op;
+    TCGReg r0, r1, r2, r3;
+#if TCG_TARGET_REG_BITS == 32
+    TCGReg r4, r5;
+#endif
+    tcg_target_ulong i1;
+    int32_t s2;
+    TCGCond c;
+    TCGMemOpIdx oi;
+    uint8_t pos, len;
+    void *ptr;
+    const uint8_t *tb_ptr;
+
+    status = info->read_memory_func(addr, buf, 2, info);
+    if (status != 0) {
+        info->memory_error_func(status, addr, info);
+        return -1;
+    }
+    op = buf[0];
+    length = buf[1];
+
+    if (length < 2) {
+        info->fprintf_func(info->stream, "invalid length %d", length);
+        return 1;
+    }
+
+    status = info->read_memory_func(addr + 2, buf + 2, length - 2, info);
+    if (status != 0) {
+        info->memory_error_func(status, addr + 2, info);
+        return -1;
+    }
+
+    def = &tcg_op_defs[op];
+    op_name = def->name;
+    tb_ptr = buf + 2;
+
+    switch (op) {
+    case INDEX_op_br:
+    case INDEX_op_call:
+    case INDEX_op_exit_tb:
+    case INDEX_op_goto_tb:
+        tci_args_l(&tb_ptr, &ptr);
+        info->fprintf_func(info->stream, "%-12s  %p", op_name, ptr);
+        break;
+
+    case INDEX_op_brcond_i32:
+    case INDEX_op_brcond_i64:
+        tci_args_rrcl(&tb_ptr, &r0, &r1, &c, &ptr);
+        info->fprintf_func(info->stream, "%-12s  %s, %s, %s, %p",
+                           op_name, str_r(r0), str_r(r1), str_c(c), ptr);
+        break;
+
+    case INDEX_op_setcond_i32:
+    case INDEX_op_setcond_i64:
+        tci_args_rrrc(&tb_ptr, &r0, &r1, &r2, &c);
+        info->fprintf_func(info->stream, "%-12s  %s, %s, %s, %s",
+                           op_name, str_r(r0), str_r(r1), str_r(r2), str_c(c));
+        break;
+
+    case INDEX_op_tci_movi_i32:
+        tci_args_ri(&tb_ptr, &r0, &i1);
+        info->fprintf_func(info->stream, "%-12s  %s, 0x%" TCG_PRIlx,
+                           op_name, str_r(r0), i1);
+        break;
+
+#if TCG_TARGET_REG_BITS == 64
+    case INDEX_op_tci_movi_i64:
+        tci_args_rI(&tb_ptr, &r0, &i1);
+        info->fprintf_func(info->stream, "%-12s  %s, 0x%" TCG_PRIlx,
+                           op_name, str_r(r0), i1);
+        break;
+#endif
+
+    case INDEX_op_ld8u_i32:
+    case INDEX_op_ld8u_i64:
+    case INDEX_op_ld8s_i32:
+    case INDEX_op_ld8s_i64:
+    case INDEX_op_ld16u_i32:
+    case INDEX_op_ld16u_i64:
+    case INDEX_op_ld16s_i32:
+    case INDEX_op_ld16s_i64:
+    case INDEX_op_ld32u_i64:
+    case INDEX_op_ld32s_i64:
+    case INDEX_op_ld_i32:
+    case INDEX_op_ld_i64:
+    case INDEX_op_st8_i32:
+    case INDEX_op_st8_i64:
+    case INDEX_op_st16_i32:
+    case INDEX_op_st16_i64:
+    case INDEX_op_st32_i64:
+    case INDEX_op_st_i32:
+    case INDEX_op_st_i64:
+        tci_args_rrs(&tb_ptr, &r0, &r1, &s2);
+        info->fprintf_func(info->stream, "%-12s  %s, %s, %d",
+                           op_name, str_r(r0), str_r(r1), s2);
+        break;
+
+    case INDEX_op_mov_i32:
+    case INDEX_op_mov_i64:
+    case INDEX_op_ext8s_i32:
+    case INDEX_op_ext8s_i64:
+    case INDEX_op_ext8u_i32:
+    case INDEX_op_ext8u_i64:
+    case INDEX_op_ext16s_i32:
+    case INDEX_op_ext16s_i64:
+    case INDEX_op_ext16u_i32:
+    case INDEX_op_ext32s_i64:
+    case INDEX_op_ext32u_i64:
+    case INDEX_op_ext_i32_i64:
+    case INDEX_op_extu_i32_i64:
+    case INDEX_op_bswap16_i32:
+    case INDEX_op_bswap16_i64:
+    case INDEX_op_bswap32_i32:
+    case INDEX_op_bswap32_i64:
+    case INDEX_op_bswap64_i64:
+    case INDEX_op_not_i32:
+    case INDEX_op_not_i64:
+    case INDEX_op_neg_i32:
+    case INDEX_op_neg_i64:
+        tci_args_rr(&tb_ptr, &r0, &r1);
+        info->fprintf_func(info->stream, "%-12s  %s, %s",
+                           op_name, str_r(r0), str_r(r1));
+        break;
+
+    case INDEX_op_add_i32:
+    case INDEX_op_add_i64:
+    case INDEX_op_sub_i32:
+    case INDEX_op_sub_i64:
+    case INDEX_op_mul_i32:
+    case INDEX_op_mul_i64:
+    case INDEX_op_and_i32:
+    case INDEX_op_and_i64:
+    case INDEX_op_or_i32:
+    case INDEX_op_or_i64:
+    case INDEX_op_xor_i32:
+    case INDEX_op_xor_i64:
+    case INDEX_op_div_i32:
+    case INDEX_op_div_i64:
+    case INDEX_op_rem_i32:
+    case INDEX_op_rem_i64:
+    case INDEX_op_divu_i32:
+    case INDEX_op_divu_i64:
+    case INDEX_op_remu_i32:
+    case INDEX_op_remu_i64:
+    case INDEX_op_shl_i32:
+    case INDEX_op_shl_i64:
+    case INDEX_op_shr_i32:
+    case INDEX_op_shr_i64:
+    case INDEX_op_sar_i32:
+    case INDEX_op_sar_i64:
+    case INDEX_op_rotl_i32:
+    case INDEX_op_rotl_i64:
+    case INDEX_op_rotr_i32:
+    case INDEX_op_rotr_i64:
+        tci_args_rrr(&tb_ptr, &r0, &r1, &r2);
+        info->fprintf_func(info->stream, "%-12s  %s, %s, %s",
+                           op_name, str_r(r0), str_r(r1), str_r(r2));
+        break;
+
+    case INDEX_op_deposit_i32:
+    case INDEX_op_deposit_i64:
+        tci_args_rrrbb(&tb_ptr, &r0, &r1, &r2, &pos, &len);
+        info->fprintf_func(info->stream, "%-12s  %s, %s, %s, %d, %d",
+                           op_name, str_r(r0), str_r(r1), str_r(r2), pos, len);
+        break;
+
+#if TCG_TARGET_REG_BITS == 32
+    case INDEX_op_setcond2_i32:
+        tci_args_rrrrrc(&tb_ptr, &r0, &r1, &r2, &r3, &r4, &c);
+        info->fprintf_func(info->stream, "%-12s  %s, %s, %s, %s, %s, %s",
+                           op_name, str_r(r0), str_r(r1), str_r(r2),
+                           str_r(r3), str_r(r4), str_c(c));
+        break;
+
+    case INDEX_op_brcond2_i32:
+        tci_args_rrrrcl(&tb_ptr, &r0, &r1, &r2, &r3, &c, &ptr);
+        info->fprintf_func(info->stream, "%-12s  %s, %s, %s, %s, %s, %p",
+                           op_name, str_r(r0), str_r(r1),
+                           str_r(r2), str_r(r3), str_c(c), ptr);
+        break;
+
+    case INDEX_op_mulu2_i32:
+        tci_args_rrrr(&tb_ptr, &r0, &r1, &r2, &r3);
+        info->fprintf_func(info->stream, "%-12s  %s, %s, %s, %s",
+                           op_name, str_r(r0), str_r(r1),
+                           str_r(r2), str_r(r3));
+        break;
+
+    case INDEX_op_add2_i32:
+    case INDEX_op_sub2_i32:
+        tci_args_rrrrrr(&tb_ptr, &r0, &r1, &r2, &r3, &r4, &r5);
+        info->fprintf_func(info->stream, "%-12s  %s, %s, %s, %s, %s, %s",
+                           op_name, str_r(r0), str_r(r1), str_r(r2),
+                           str_r(r3), str_r(r4), str_r(r5));
+        break;
+#endif
+
+    case INDEX_op_qemu_ld_i64:
+    case INDEX_op_qemu_st_i64:
+        len = DIV_ROUND_UP(64, TCG_TARGET_REG_BITS);
+        goto do_qemu_ldst;
+    case INDEX_op_qemu_ld_i32:
+    case INDEX_op_qemu_st_i32:
+        len = 1;
+    do_qemu_ldst:
+        len += DIV_ROUND_UP(TARGET_LONG_BITS, TCG_TARGET_REG_BITS);
+        switch (len) {
+        case 2:
+            tci_args_rrm(&tb_ptr, &r0, &r1, &oi);
+            info->fprintf_func(info->stream, "%-12s  %s, %s, %x",
+                               op_name, str_r(r0), str_r(r1), oi);
+            break;
+        case 3:
+            tci_args_rrrm(&tb_ptr, &r0, &r1, &r2, &oi);
+            info->fprintf_func(info->stream, "%-12s  %s, %s, %s, %x",
+                               op_name, str_r(r0), str_r(r1), str_r(r2), oi);
+            break;
+        case 4:
+            tci_args_rrrrm(&tb_ptr, &r0, &r1, &r2, &r3, &oi);
+            info->fprintf_func(info->stream, "%-12s  %s, %s, %s, %s, %x",
+                               op_name, str_r(r0), str_r(r1),
+                               str_r(r2), str_r(r3), oi);
+            break;
+        default:
+            g_assert_not_reached();
+        }
+        break;
+
+    default:
+        info->fprintf_func(info->stream, "illegal opcode %d", op);
+        break;
+    }
+
+    return length;
 }