tcg/tci: Add TCG_TARGET_CALL_{RET,ARG}_I128

Fill in the parameters for libffi for Int128.
Adjust the interpreter to allow for 16-byte return values.
Adjust tcg_out_call to record the return value length.

Call parameters are no longer all the same size, so we
cannot reuse the same call_slots array for every function.
Compute it each time now, but only fill in slots required
for the call we're about to make.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
diff --git a/tcg/tci.c b/tcg/tci.c
index eeccdde..022fe9d 100644
--- a/tcg/tci.c
+++ b/tcg/tci.c
@@ -470,12 +470,9 @@
     tcg_target_ulong regs[TCG_TARGET_NB_REGS];
     uint64_t stack[(TCG_STATIC_CALL_ARGS_SIZE + TCG_STATIC_FRAME_SIZE)
                    / sizeof(uint64_t)];
-    void *call_slots[TCG_STATIC_CALL_ARGS_SIZE / sizeof(uint64_t)];
 
     regs[TCG_AREG0] = (tcg_target_ulong)env;
     regs[TCG_REG_CALL_STACK] = (uintptr_t)stack;
-    /* Other call_slots entries initialized at first use (see below). */
-    call_slots[0] = NULL;
     tci_assert(tb_ptr);
 
     for (;;) {
@@ -498,26 +495,26 @@
 
         switch (opc) {
         case INDEX_op_call:
-            /*
-             * Set up the ffi_avalue array once, delayed until now
-             * because many TB's do not make any calls. In tcg_gen_callN,
-             * we arranged for every real argument to be "left-aligned"
-             * in each 64-bit slot.
-             */
-            if (unlikely(call_slots[0] == NULL)) {
-                for (int i = 0; i < ARRAY_SIZE(call_slots); ++i) {
-                    call_slots[i] = &stack[i];
-                }
-            }
-
-            tci_args_nl(insn, tb_ptr, &len, &ptr);
-
-            /* Helper functions may need to access the "return address" */
-            tci_tb_ptr = (uintptr_t)tb_ptr;
-
             {
-                void **pptr = ptr;
-                ffi_call(pptr[1], pptr[0], stack, call_slots);
+                void *call_slots[MAX_CALL_IARGS];
+                ffi_cif *cif;
+                void *func;
+                unsigned i, s, n;
+
+                tci_args_nl(insn, tb_ptr, &len, &ptr);
+                func = ((void **)ptr)[0];
+                cif = ((void **)ptr)[1];
+
+                n = cif->nargs;
+                for (i = s = 0; i < n; ++i) {
+                    ffi_type *t = cif->arg_types[i];
+                    call_slots[i] = &stack[s];
+                    s += DIV_ROUND_UP(t->size, 8);
+                }
+
+                /* Helper functions may need to access the "return address" */
+                tci_tb_ptr = (uintptr_t)tb_ptr;
+                ffi_call(cif, func, stack, call_slots);
             }
 
             switch (len) {
@@ -542,6 +539,9 @@
                  */
                 memcpy(&regs[TCG_REG_R0], stack, 8);
                 break;
+            case 3: /* Int128 */
+                memcpy(&regs[TCG_REG_R0], stack, 16);
+                break;
             default:
                 g_assert_not_reached();
             }