From patchwork Mon Jan 11 16:48:10 2010
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: mulu2 not implemented for tcg target sparc
From: Richard Henderson <rth@twiddle.net>
X-Patchwork-Id: 42632
Message-Id: <4B4B564A.1000207@twiddle.net>
To: Palle Lyckegaard <palle@lyckegaard.dk>
Cc: Blue Swirl <blauwirbel@gmail.com>,
 Laurent Desnogues <laurent.desnogues@gmail.com>, qemu-devel@nongnu.org
Date: Mon, 11 Jan 2010 08:48:10 -0800

On 01/10/2010 01:17 PM, Palle Lyckegaard wrote:
> On Sun, 10 Jan 2010, Blue Swirl wrote:
>
>>>>
>>>> Is it needed somewhere?
>>>
>
> I was trying to run qemu-system-mips with a NetBSD malta kernel that
> generates a MIPS mult operation. Tracing the code through tcg points at
> a missing mulu2 opreration for sparc.
>
>  >> It will be generated for instance for some muls when the target >>
> is i386 (not even x86_64) on a 32-bit host.
>>
>> In that case mulu2 support would be useful for Sparc32 host too.
>> Palle, are you planning to dig into this?
>
> Yes I will try even if my tcg and SPARC assembly skills and very limited
> (so far...)

Try this.  You're also missing double-word addition and subtraction.


r~


diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c
index 8f094e5..003f084 100644
--- a/tcg/sparc/tcg-target.c
+++ b/tcg/sparc/tcg-target.c
@@ -215,6 +215,7 @@ static inline int tcg_target_const_match(tcg_target_long val,
 #define BA         (INSN_OP(0) | INSN_COND(COND_A, 0) | INSN_OP2(0x2))
 
 #define ARITH_ADD  (INSN_OP(2) | INSN_OP3(0x00))
+#define ARITH_ADDCC (INSN_OP(2) | INSN_OP3(0x10))
 #define ARITH_AND  (INSN_OP(2) | INSN_OP3(0x01))
 #define ARITH_OR   (INSN_OP(2) | INSN_OP3(0x02))
 #define ARITH_ORCC (INSN_OP(2) | INSN_OP3(0x12))
@@ -238,6 +239,7 @@ static inline int tcg_target_const_match(tcg_target_long val,
 #define SHIFT_SRLX (INSN_OP(2) | INSN_OP3(0x26) | (1 << 12))
 #define SHIFT_SRAX (INSN_OP(2) | INSN_OP3(0x27) | (1 << 12))
 
+#define RDY        (INSN_OP(2) | INSN_OP3(0x28) | INSN_RS1(0))
 #define WRY        (INSN_OP(2) | INSN_OP3(0x30))
 #define JMPL       (INSN_OP(2) | INSN_OP3(0x38))
 #define SAVE       (INSN_OP(2) | INSN_OP3(0x3c))
@@ -403,6 +405,11 @@ static inline void tcg_out_sety(TCGContext *s, tcg_target_long val)
         fprintf(stderr, "unimplemented sety %ld\n", (long)val);
 }
 
+static inline void tcg_out_rdy(TCGContext *s, int rd)
+{
+    tcg_out32(s, RDY | INSN_RD(rd));
+}
+
 static inline void tcg_out_addi(TCGContext *s, int reg, tcg_target_long val)
 {
     if (val != 0) {
@@ -1128,6 +1135,38 @@ static inline void tcg_out_op(TCGContext *s, int opc, const TCGArg *args,
                             args[2], const_args[2],
                             args[3], const_args[3], args[5]);
         break;
+    case INDEX_op_add2_i32:
+        if (const_args[4]) {
+            tcg_out_arithi(s, args[0], args[2], args[4], ARITH_ADDCC);
+        } else {
+            tcg_out_arith(s, args[0], args[2], args[4], ARITH_ADDCC);
+        }
+        if (const_args[5]) {
+            tcg_out_arithi(s, args[1], args[3], args[5], ARITH_ADDX);
+        } else {
+            tcg_out_arithi(s, args[1], args[3], args[5], ARITH_ADDX);
+        }
+        break;
+    case INDEX_op_sub2_i32:
+        if (const_args[4]) {
+            tcg_out_arithi(s, args[0], args[2], args[4], ARITH_SUBCC);
+        } else {
+            tcg_out_arith(s, args[0], args[2], args[4], ARITH_SUBCC);
+        }
+        if (const_args[5]) {
+            tcg_out_arithi(s, args[1], args[3], args[5], ARITH_SUBX);
+        } else {
+            tcg_out_arithi(s, args[1], args[3], args[5], ARITH_SUBX);
+        }
+        break;
+    case INDEX_op_mulu2_i32:
+        if (const_args[3]) {
+            tcg_out_arithi(s, args[0], args[2], args[3], ARITH_UMUL);
+        } else {
+            tcg_out_arith(s, args[0], args[2], args[3], ARITH_UMUL);
+        }
+        tcg_out_rdy(s, args[1]);
+        break;
 #endif
 
     case INDEX_op_qemu_ld8u: