[PATCH v4 68/71] tcg/tci: Implement add2, sub2

[Richard Henderson <richard.henderson@linaro.org>]

We already had the 32-bit versions for a 32-bit host; expand this
to 64-bit hosts as well.  The 64-bit opcodes are new.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 tcg/tci/tcg-target.h     |  8 ++++----
 tcg/tci.c                | 40 ++++++++++++++++++++++++++--------------
 tcg/tci/tcg-target.c.inc | 15 ++++++++-------
 3 files changed, 38 insertions(+), 25 deletions(-)

diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h
index 71a44bbfb0..515b3c7a56 100644
--- a/tcg/tci/tcg-target.h
+++ b/tcg/tci/tcg-target.h
@@ -121,11 +121,11 @@
 #define TCG_TARGET_HAS_rot_i64          1
 #define TCG_TARGET_HAS_movcond_i64      1
 #define TCG_TARGET_HAS_muls2_i64        1
-#define TCG_TARGET_HAS_add2_i32         0
-#define TCG_TARGET_HAS_sub2_i32         0
+#define TCG_TARGET_HAS_add2_i32         1
+#define TCG_TARGET_HAS_sub2_i32         1
 #define TCG_TARGET_HAS_mulu2_i32        1
-#define TCG_TARGET_HAS_add2_i64         0
-#define TCG_TARGET_HAS_sub2_i64         0
+#define TCG_TARGET_HAS_add2_i64         1
+#define TCG_TARGET_HAS_sub2_i64         1
 #define TCG_TARGET_HAS_mulu2_i64        1
 #define TCG_TARGET_HAS_muluh_i64        0
 #define TCG_TARGET_HAS_mulsh_i64        0
diff --git a/tcg/tci.c b/tcg/tci.c
index d76b9f5798..0240d850cf 100644
--- a/tcg/tci.c
+++ b/tcg/tci.c
@@ -189,7 +189,6 @@ static void tci_args_rrrrrc(uint32_t insn, TCGReg *r0, TCGReg *r1,
     *c5 = extract32(insn, 28, 4);
 }
 
-#if TCG_TARGET_REG_BITS == 32
 static void tci_args_rrrrrr(uint32_t insn, TCGReg *r0, TCGReg *r1,
                             TCGReg *r2, TCGReg *r3, TCGReg *r4, TCGReg *r5)
 {
@@ -200,7 +199,6 @@ static void tci_args_rrrrrr(uint32_t insn, TCGReg *r0, TCGReg *r1,
     *r4 = extract32(insn, 24, 4);
     *r5 = extract32(insn, 28, 4);
 }
-#endif
 
 static bool tci_compare32(uint32_t u0, uint32_t u1, TCGCond condition)
 {
@@ -351,17 +349,14 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
     for (;;) {
         uint32_t insn;
         TCGOpcode opc;
-        TCGReg r0, r1, r2, r3, r4;
+        TCGReg r0, r1, r2, r3, r4, r5;
         tcg_target_ulong t1;
         TCGCond condition;
         target_ulong taddr;
         uint8_t pos, len;
         uint32_t tmp32;
         uint64_t tmp64;
-#if TCG_TARGET_REG_BITS == 32
-        TCGReg r5;
         uint64_t T1, T2;
-#endif
         TCGMemOpIdx oi;
         int32_t ofs;
         void *ptr;
@@ -655,20 +650,22 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
                 tb_ptr = ptr;
             }
             break;
-#if TCG_TARGET_REG_BITS == 32
+#if TCG_TARGET_REG_BITS == 32 || TCG_TARGET_HAS_add2_i32
         case INDEX_op_add2_i32:
             tci_args_rrrrrr(insn, &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;
+#endif
+#if TCG_TARGET_REG_BITS == 32 || TCG_TARGET_HAS_sub2_i32
         case INDEX_op_sub2_i32:
             tci_args_rrrrrr(insn, &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;
-#endif /* TCG_TARGET_REG_BITS == 32 */
+#endif
 #if TCG_TARGET_HAS_mulu2_i32
         case INDEX_op_mulu2_i32:
             tci_args_rrrr(insn, &r0, &r1, &r2, &r3);
@@ -798,6 +795,24 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
             muls64(&regs[r0], &regs[r1], regs[r2], regs[r3]);
             break;
 #endif
+#if TCG_TARGET_HAS_add2_i64
+        case INDEX_op_add2_i64:
+            tci_args_rrrrrr(insn, &r0, &r1, &r2, &r3, &r4, &r5);
+            T1 = regs[r2] + regs[r4];
+            T2 = regs[r3] + regs[r5] + (T1 < regs[r2]);
+            regs[r0] = T1;
+            regs[r1] = T2;
+            break;
+#endif
+#if TCG_TARGET_HAS_add2_i64
+        case INDEX_op_sub2_i64:
+            tci_args_rrrrrr(insn, &r0, &r1, &r2, &r3, &r4, &r5);
+            T1 = regs[r2] - regs[r4];
+            T2 = regs[r3] - regs[r5] - (regs[r2] < regs[r4]);
+            regs[r0] = T1;
+            regs[r1] = T2;
+            break;
+#endif
 
             /* Shift/rotate operations (64 bit). */
 
@@ -1114,10 +1129,7 @@ int print_insn_tci(bfd_vma addr, disassemble_info *info)
     const char *op_name;
     uint32_t insn;
     TCGOpcode op;
-    TCGReg r0, r1, r2, r3, r4;
-#if TCG_TARGET_REG_BITS == 32
-    TCGReg r5;
-#endif
+    TCGReg r0, r1, r2, r3, r4, r5;
     tcg_target_ulong i1;
     int32_t s2;
     TCGCond c;
@@ -1315,15 +1327,15 @@ int print_insn_tci(bfd_vma addr, disassemble_info *info)
                            str_r(r2), str_r(r3));
         break;
 
-#if TCG_TARGET_REG_BITS == 32
     case INDEX_op_add2_i32:
+    case INDEX_op_add2_i64:
     case INDEX_op_sub2_i32:
+    case INDEX_op_sub2_i64:
         tci_args_rrrrrr(insn, &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:
diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc
index eb48633fba..9b2e2c32a1 100644
--- a/tcg/tci/tcg-target.c.inc
+++ b/tcg/tci/tcg-target.c.inc
@@ -134,11 +134,13 @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op)
     case INDEX_op_brcond_i64:
         return C_O0_I2(r, r);
 
-#if TCG_TARGET_REG_BITS == 32
-    /* TODO: Support R, R, R, R, RI, RI? Will it be faster? */
     case INDEX_op_add2_i32:
+    case INDEX_op_add2_i64:
     case INDEX_op_sub2_i32:
+    case INDEX_op_sub2_i64:
         return C_O2_I4(r, r, r, r, r, r);
+
+#if TCG_TARGET_REG_BITS == 32
     case INDEX_op_brcond2_i32:
         return C_O0_I4(r, r, r, r);
 #endif
@@ -467,7 +469,6 @@ static void tcg_out_op_rrrrrc(TCGContext *s, TCGOpcode op,
     tcg_out32(s, insn);
 }
 
-#if TCG_TARGET_REG_BITS == 32
 static void tcg_out_op_rrrrrr(TCGContext *s, TCGOpcode op,
                               TCGReg r0, TCGReg r1, TCGReg r2,
                               TCGReg r3, TCGReg r4, TCGReg r5)
@@ -483,7 +484,6 @@ static void tcg_out_op_rrrrrr(TCGContext *s, TCGOpcode op,
     insn = deposit32(insn, 28, 4, r5);
     tcg_out32(s, insn);
 }
-#endif
 
 static void tcg_out_ldst(TCGContext *s, TCGOpcode op, TCGReg val,
                          TCGReg base, intptr_t offset)
@@ -719,12 +719,13 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args,
         tcg_out_op_rr(s, opc, args[0], args[1]);
         break;
 
-#if TCG_TARGET_REG_BITS == 32
-    case INDEX_op_add2_i32:
-    case INDEX_op_sub2_i32:
+    CASE_32_64(add2)
+    CASE_32_64(sub2)
         tcg_out_op_rrrrrr(s, opc, args[0], args[1], args[2],
                           args[3], args[4], args[5]);
         break;
+
+#if TCG_TARGET_REG_BITS == 32
     case INDEX_op_brcond2_i32:
         tcg_out_op_rrrrrc(s, INDEX_op_setcond2_i32, TCG_REG_TMP,
                           args[0], args[1], args[2], args[3], args[4]);
-- 
2.25.1