tcg/tci: Implement goto_ptr

This operation is critical to staying within the interpretation
loop longer, which avoids the overhead of setup and teardown for
many TBs.

The check in tcg_prologue_init is disabled because TCI does
want to use NULL to indicate exit, as opposed to branching to
a real epilogue.

Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

1 files changed, 19 insertions, 0 deletions

diff --git a/tcg/tci.c b/tcg/tci.c
index c82d5b2f51..4696ca161c 100644
--- a/tcg/tci.c
+++ b/tcg/tci.c
@@ -73,6 +73,11 @@ static void tci_args_l(uint32_t insn, const void *tb_ptr, void **l0)
     *l0 = diff ? (void *)tb_ptr + diff : NULL;
 }
 
+static void tci_args_r(uint32_t insn, TCGReg *r0)
+{
+    *r0 = extract32(insn, 8, 4);
+}
+
 static void tci_args_nl(uint32_t insn, const void *tb_ptr,
                         uint8_t *n0, void **l1)
 {
@@ -739,6 +744,15 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
             tb_ptr = *(void **)ptr;
             break;
 
+        case INDEX_op_goto_ptr:
+            tci_args_r(insn, &r0);
+            ptr = (void *)regs[r0];
+            if (!ptr) {
+                return 0;
+            }
+            tb_ptr = ptr;
+            break;
+
         case INDEX_op_qemu_ld_i32:
             if (TARGET_LONG_BITS <= TCG_TARGET_REG_BITS) {
                 tci_args_rrm(insn, &r0, &r1, &oi);
@@ -996,6 +1010,11 @@ int print_insn_tci(bfd_vma addr, disassemble_info *info)
         info->fprintf_func(info->stream, "%-12s  %p", op_name, ptr);
         break;
 
+    case INDEX_op_goto_ptr:
+        tci_args_r(insn, &r0);
+        info->fprintf_func(info->stream, "%-12s  %s", op_name, str_r(r0));
+        break;
+
     case INDEX_op_call:
         tci_args_nl(insn, tb_ptr, &len, &ptr);
         info->fprintf_func(info->stream, "%-12s  %d, %p", op_name, len, ptr);