@@ -20,7 +20,6 @@
#define TCG_TARGET_HAS_add2_i32 1
#define TCG_TARGET_HAS_sub2_i32 1
#define TCG_TARGET_HAS_mulu2_i32 0
-#define TCG_TARGET_HAS_muls2_i32 0
#define TCG_TARGET_HAS_extr_i64_i32 0
#define TCG_TARGET_HAS_qemu_st8_i32 0
@@ -32,7 +31,6 @@
#define TCG_TARGET_HAS_add2_i64 1
#define TCG_TARGET_HAS_sub2_i64 1
#define TCG_TARGET_HAS_mulu2_i64 0
-#define TCG_TARGET_HAS_muls2_i64 0
/*
* Without FEAT_LSE2, we must use LDXP+STXP to implement atomic 128-bit load,
@@ -29,7 +29,6 @@ extern bool use_neon_instructions;
#define TCG_TARGET_HAS_extract2_i32 1
#define TCG_TARGET_HAS_negsetcond_i32 1
#define TCG_TARGET_HAS_mulu2_i32 1
-#define TCG_TARGET_HAS_muls2_i32 1
#define TCG_TARGET_HAS_qemu_st8_i32 0
#define TCG_TARGET_HAS_qemu_ldst_i128 0
@@ -33,7 +33,6 @@
#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_muls2_i32 1
#if TCG_TARGET_REG_BITS == 64
/* Keep 32-bit values zero-extended in a register. */
@@ -46,7 +45,6 @@
#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_muls2_i64 1
#define TCG_TARGET_HAS_qemu_st8_i32 0
#else
#define TCG_TARGET_HAS_qemu_st8_i32 1
@@ -15,7 +15,6 @@
#define TCG_TARGET_HAS_add2_i32 0
#define TCG_TARGET_HAS_sub2_i32 0
#define TCG_TARGET_HAS_mulu2_i32 0
-#define TCG_TARGET_HAS_muls2_i32 0
#define TCG_TARGET_HAS_bswap16_i32 1
#define TCG_TARGET_HAS_bswap32_i32 1
#define TCG_TARGET_HAS_qemu_st8_i32 0
@@ -30,7 +29,6 @@
#define TCG_TARGET_HAS_add2_i64 0
#define TCG_TARGET_HAS_sub2_i64 0
#define TCG_TARGET_HAS_mulu2_i64 0
-#define TCG_TARGET_HAS_muls2_i64 0
#define TCG_TARGET_HAS_qemu_ldst_i128 (cpuinfo & CPUINFO_LSX)
@@ -40,7 +40,6 @@ extern bool use_mips32r2_instructions;
/* optional instructions */
#define TCG_TARGET_HAS_mulu2_i32 (!use_mips32r6_instructions)
-#define TCG_TARGET_HAS_muls2_i32 (!use_mips32r6_instructions)
#define TCG_TARGET_HAS_bswap16_i32 1
#define TCG_TARGET_HAS_bswap32_i32 1
#define TCG_TARGET_HAS_negsetcond_i32 0
@@ -52,7 +51,6 @@ extern bool use_mips32r2_instructions;
#define TCG_TARGET_HAS_add2_i64 0
#define TCG_TARGET_HAS_sub2_i64 0
#define TCG_TARGET_HAS_mulu2_i64 (!use_mips32r6_instructions)
-#define TCG_TARGET_HAS_muls2_i64 (!use_mips32r6_instructions)
#define TCG_TARGET_HAS_ext32s_i64 1
#define TCG_TARGET_HAS_ext32u_i64 1
#define TCG_TARGET_HAS_negsetcond_i64 0
@@ -22,7 +22,6 @@
#define TCG_TARGET_HAS_extract2_i32 0
#define TCG_TARGET_HAS_negsetcond_i32 1
#define TCG_TARGET_HAS_mulu2_i32 0
-#define TCG_TARGET_HAS_muls2_i32 0
#define TCG_TARGET_HAS_qemu_st8_i32 0
#if TCG_TARGET_REG_BITS == 64
@@ -37,7 +36,6 @@
#define TCG_TARGET_HAS_add2_i64 1
#define TCG_TARGET_HAS_sub2_i64 1
#define TCG_TARGET_HAS_mulu2_i64 0
-#define TCG_TARGET_HAS_muls2_i64 0
#endif
#define TCG_TARGET_HAS_qemu_ldst_i128 \
@@ -15,7 +15,6 @@
#define TCG_TARGET_HAS_add2_i32 1
#define TCG_TARGET_HAS_sub2_i32 1
#define TCG_TARGET_HAS_mulu2_i32 0
-#define TCG_TARGET_HAS_muls2_i32 0
#define TCG_TARGET_HAS_bswap16_i32 (cpuinfo & CPUINFO_ZBB)
#define TCG_TARGET_HAS_bswap32_i32 (cpuinfo & CPUINFO_ZBB)
#define TCG_TARGET_HAS_qemu_st8_i32 0
@@ -29,7 +28,6 @@
#define TCG_TARGET_HAS_add2_i64 1
#define TCG_TARGET_HAS_sub2_i64 1
#define TCG_TARGET_HAS_mulu2_i64 0
-#define TCG_TARGET_HAS_muls2_i64 0
#define TCG_TARGET_HAS_qemu_ldst_i128 0
@@ -36,7 +36,6 @@ extern uint64_t s390_facilities[3];
#define TCG_TARGET_HAS_add2_i32 1
#define TCG_TARGET_HAS_sub2_i32 1
#define TCG_TARGET_HAS_mulu2_i32 0
-#define TCG_TARGET_HAS_muls2_i32 0
#define TCG_TARGET_HAS_extr_i64_i32 0
#define TCG_TARGET_HAS_qemu_st8_i32 0
@@ -48,7 +47,6 @@ extern uint64_t s390_facilities[3];
#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_muls2_i64 HAVE_FACILITY(MISC_INSN_EXT2)
#define TCG_TARGET_HAS_qemu_ldst_i128 1
@@ -17,5 +17,6 @@ C_O1_I2(r, r, r)
C_O1_I2(r, r, rJ)
C_O1_I2(r, rz, rJ)
C_O1_I4(r, rz, rJ, rI, 0)
+C_O2_I2(r, r, r, r)
C_O2_I2(r, r, rz, rJ)
C_O2_I4(r, r, rz, rz, rJ, rJ)
@@ -21,7 +21,6 @@ extern bool use_vis3_instructions;
#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_muls2_i32 1
#define TCG_TARGET_HAS_qemu_st8_i32 0
#define TCG_TARGET_HAS_extr_i64_i32 0
@@ -33,7 +32,6 @@ extern bool use_vis3_instructions;
#define TCG_TARGET_HAS_add2_i64 1
#define TCG_TARGET_HAS_sub2_i64 1
#define TCG_TARGET_HAS_mulu2_i64 0
-#define TCG_TARGET_HAS_muls2_i64 0
#define TCG_TARGET_HAS_qemu_ldst_i128 0
@@ -20,7 +20,6 @@
#define TCG_TARGET_HAS_add2_i64 0
#define TCG_TARGET_HAS_sub2_i64 0
#define TCG_TARGET_HAS_mulu2_i64 0
-#define TCG_TARGET_HAS_muls2_i64 0
/* Turn some undef macros into true macros. */
#define TCG_TARGET_HAS_add2_i32 1
#define TCG_TARGET_HAS_sub2_i32 1
@@ -11,7 +11,6 @@
#define TCG_TARGET_HAS_bswap32_i32 1
#define TCG_TARGET_HAS_extract2_i32 0
#define TCG_TARGET_HAS_negsetcond_i32 0
-#define TCG_TARGET_HAS_muls2_i32 1
#define TCG_TARGET_HAS_qemu_st8_i32 0
#if TCG_TARGET_REG_BITS == 64
@@ -21,7 +20,6 @@
#define TCG_TARGET_HAS_bswap64_i64 1
#define TCG_TARGET_HAS_extract2_i64 0
#define TCG_TARGET_HAS_negsetcond_i64 0
-#define TCG_TARGET_HAS_muls2_i64 1
#define TCG_TARGET_HAS_add2_i32 1
#define TCG_TARGET_HAS_sub2_i32 1
#define TCG_TARGET_HAS_mulu2_i32 1
@@ -1162,7 +1162,7 @@ void tcg_gen_mulu2_i32(TCGv_i32 rl, TCGv_i32 rh, TCGv_i32 arg1, TCGv_i32 arg2)
void tcg_gen_muls2_i32(TCGv_i32 rl, TCGv_i32 rh, TCGv_i32 arg1, TCGv_i32 arg2)
{
- if (TCG_TARGET_HAS_muls2_i32) {
+ if (tcg_op_supported(INDEX_op_muls2_i32, TCG_TYPE_I32, 0)) {
tcg_gen_op4_i32(INDEX_op_muls2_i32, rl, rh, arg1, arg2);
} else if (tcg_op_supported(INDEX_op_mulsh, TCG_TYPE_I32, 0)) {
TCGv_i32 t = tcg_temp_ebb_new_i32();
@@ -2880,7 +2880,7 @@ void tcg_gen_mulu2_i64(TCGv_i64 rl, TCGv_i64 rh, TCGv_i64 arg1, TCGv_i64 arg2)
void tcg_gen_muls2_i64(TCGv_i64 rl, TCGv_i64 rh, TCGv_i64 arg1, TCGv_i64 arg2)
{
- if (TCG_TARGET_HAS_muls2_i64) {
+ if (tcg_op_supported(INDEX_op_muls2_i64, TCG_TYPE_I64, 0)) {
tcg_gen_op4_i64(INDEX_op_muls2_i64, rl, rh, arg1, arg2);
} else if (tcg_op_supported(INDEX_op_mulsh, TCG_TYPE_I64, 0)) {
TCGv_i64 t = tcg_temp_ebb_new_i64();
@@ -992,6 +992,12 @@ typedef struct TCGOutOpDivRem {
TCGReg a0, TCGReg a1, TCGReg a4);
} TCGOutOpDivRem;
+typedef struct TCGOutOpMul2 {
+ TCGOutOp base;
+ void (*out_rrrr)(TCGContext *s, TCGType type,
+ TCGReg a0, TCGReg a1, TCGReg a2, TCGReg a3);
+} TCGOutOpMul2;
+
typedef struct TCGOutOpUnary {
TCGOutOp base;
void (*out_rr)(TCGContext *s, TCGType type, TCGReg a0, TCGReg a1);
@@ -1035,6 +1041,8 @@ static const TCGOutOp * const all_outop[NB_OPS] = {
OUTOP(INDEX_op_divu2, TCGOutOpDivRem, outop_divu2),
OUTOP(INDEX_op_eqv, TCGOutOpBinary, outop_eqv),
OUTOP(INDEX_op_mul, TCGOutOpBinary, outop_mul),
+ OUTOP(INDEX_op_muls2_i32, TCGOutOpMul2, outop_muls2),
+ OUTOP(INDEX_op_muls2_i64, TCGOutOpMul2, outop_muls2),
OUTOP(INDEX_op_mulsh, TCGOutOpBinary, outop_mulsh),
OUTOP(INDEX_op_muluh, TCGOutOpBinary, outop_muluh),
OUTOP(INDEX_op_nand, TCGOutOpBinary, outop_nand),
@@ -2285,8 +2293,6 @@ bool tcg_op_supported(TCGOpcode op, TCGType type, unsigned flags)
return TCG_TARGET_HAS_sub2_i32;
case INDEX_op_mulu2_i32:
return TCG_TARGET_HAS_mulu2_i32;
- case INDEX_op_muls2_i32:
- return TCG_TARGET_HAS_muls2_i32;
case INDEX_op_bswap16_i32:
return TCG_TARGET_HAS_bswap16_i32;
case INDEX_op_bswap32_i32:
@@ -2336,8 +2342,6 @@ bool tcg_op_supported(TCGOpcode op, TCGType type, unsigned flags)
return TCG_TARGET_HAS_sub2_i64;
case INDEX_op_mulu2_i64:
return TCG_TARGET_HAS_mulu2_i64;
- case INDEX_op_muls2_i64:
- return TCG_TARGET_HAS_muls2_i64;
case INDEX_op_mov_vec:
case INDEX_op_dup_vec:
@@ -5461,6 +5465,20 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
}
break;
+ case INDEX_op_muls2_i32:
+ case INDEX_op_muls2_i64:
+ {
+ const TCGOutOpMul2 *out =
+ container_of(all_outop[op->opc], TCGOutOpMul2, base);
+
+ tcg_debug_assert(!const_args[2]);
+ tcg_debug_assert(!const_args[3]);
+ out->out_rrrr(s, type, new_args[0], new_args[1],
+ new_args[2], new_args[3]);
+ }
+ break;
+
+
default:
if (def->flags & TCG_OPF_VECTOR) {
tcg_out_vec_op(s, op->opc, type - TCG_TYPE_V64,
@@ -581,6 +581,16 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
tci_args_rr(insn, &r0, &r1);
regs[r0] = ctpop_tr(regs[r1]);
break;
+ case INDEX_op_muls2_i32:
+ case INDEX_op_muls2_i64:
+ tci_args_rrrr(insn, &r0, &r1, &r2, &r3);
+#if TCG_TARGET_REG_BITS == 32
+ tmp64 = (int64_t)(int32_t)regs[r2] * (int32_t)regs[r3];
+ tci_write_reg64(regs, r1, r0, tmp64);
+#else
+ muls64(®s[r0], ®s[r1], regs[r2], regs[r3]);
+#endif
+ break;
/* Arithmetic operations (32 bit). */
@@ -675,13 +685,6 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
tci_write_reg64(regs, r1, r0, tmp64);
break;
#endif
-#if TCG_TARGET_HAS_muls2_i32
- case INDEX_op_muls2_i32:
- tci_args_rrrr(insn, &r0, &r1, &r2, &r3);
- tmp64 = (int64_t)(int32_t)regs[r2] * (int32_t)regs[r3];
- tci_write_reg64(regs, r1, r0, tmp64);
- break;
-#endif
#if TCG_TARGET_HAS_bswap16_i32 || TCG_TARGET_HAS_bswap16_i64
CASE_32_64(bswap16)
tci_args_rr(insn, &r0, &r1);
@@ -745,12 +748,6 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
mulu64(®s[r0], ®s[r1], regs[r2], regs[r3]);
break;
#endif
-#if TCG_TARGET_HAS_muls2_i64
- case INDEX_op_muls2_i64:
- tci_args_rrrr(insn, &r0, &r1, &r2, &r3);
- muls64(®s[r0], ®s[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);
@@ -2205,6 +2205,10 @@ static const TCGOutOpBinary outop_mul = {
.out_rrr = tgen_mul,
};
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_NotImplemented,
+};
+
static TCGConstraintSetIndex cset_mulh(TCGType type, unsigned flags)
{
return type == TCG_TYPE_I64 ? C_O1_I2(r, r, r) : C_NotImplemented;
@@ -935,14 +935,6 @@ static void tcg_out_umull32(TCGContext *s, ARMCond cond, TCGReg rd0,
(rd1 << 16) | (rd0 << 12) | (rm << 8) | rn);
}
-static void tcg_out_smull32(TCGContext *s, ARMCond cond, TCGReg rd0,
- TCGReg rd1, TCGReg rn, TCGReg rm)
-{
- /* smull */
- tcg_out32(s, (cond << 28) | 0x00c00090 |
- (rd1 << 16) | (rd0 << 12) | (rm << 8) | rn);
-}
-
static void tcg_out_ext8s(TCGContext *s, TCGType t, TCGReg rd, TCGReg rn)
{
/* sxtb */
@@ -1979,6 +1971,19 @@ static const TCGOutOpBinary outop_mul = {
.out_rrr = tgen_mul,
};
+static void tgen_muls2(TCGContext *s, TCGType type,
+ TCGReg rd0, TCGReg rd1, TCGReg rn, TCGReg rm)
+{
+ /* smull */
+ tcg_out32(s, (COND_AL << 28) | 0x00c00090 |
+ (rd1 << 16) | (rd0 << 12) | (rm << 8) | rn);
+}
+
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_O2_I2(r, r, r, r),
+ .out_rrrr = tgen_muls2,
+};
+
static const TCGOutOpBinary outop_mulsh = {
.base.static_constraint = C_NotImplemented,
};
@@ -2252,9 +2257,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type,
case INDEX_op_mulu2_i32:
tcg_out_umull32(s, COND_AL, args[0], args[1], args[2], args[3]);
break;
- case INDEX_op_muls2_i32:
- tcg_out_smull32(s, COND_AL, args[0], args[1], args[2], args[3]);
- break;
case INDEX_op_brcond_i32:
c = tcg_out_cmp(s, args[2], args[0], args[1], const_args[1]);
@@ -2379,7 +2381,6 @@ tcg_target_op_def(TCGOpcode op, TCGType type, unsigned flags)
return C_O1_I2(r, r, rIN);
case INDEX_op_mulu2_i32:
- case INDEX_op_muls2_i32:
return C_O2_I2(r, r, r, r);
case INDEX_op_brcond_i32:
@@ -2743,6 +2743,18 @@ static const TCGOutOpBinary outop_mul = {
.out_rri = tgen_muli,
};
+static void tgen_muls2(TCGContext *s, TCGType type,
+ TCGReg a0, TCGReg a1, TCGReg a2, TCGReg a3)
+{
+ int rexw = type == TCG_TYPE_I32 ? 0 : P_REXW;
+ tcg_out_modrm(s, OPC_GRP3_Ev + rexw, EXT3_IMUL, a3);
+}
+
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_O2_I2(a, d, a, r),
+ .out_rrrr = tgen_muls2,
+};
+
static const TCGOutOpBinary outop_mulsh = {
.base.static_constraint = C_NotImplemented,
};
@@ -3136,9 +3148,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type,
OP_32_64(mulu2):
tcg_out_modrm(s, OPC_GRP3_Ev + rexw, EXT3_MUL, args[3]);
break;
- OP_32_64(muls2):
- tcg_out_modrm(s, OPC_GRP3_Ev + rexw, EXT3_IMUL, args[3]);
- break;
OP_32_64(add2):
if (const_args[4]) {
tgen_arithi(s, ARITH_ADD + rexw, a0, args[4], 1);
@@ -3928,8 +3937,6 @@ tcg_target_op_def(TCGOpcode op, TCGType type, unsigned flags)
case INDEX_op_mulu2_i32:
case INDEX_op_mulu2_i64:
- case INDEX_op_muls2_i32:
- case INDEX_op_muls2_i64:
return C_O2_I2(a, d, a, r);
case INDEX_op_add2_i32:
@@ -1409,6 +1409,10 @@ static const TCGOutOpBinary outop_mul = {
.out_rrr = tgen_mul,
};
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_NotImplemented,
+};
+
static void tgen_mulsh(TCGContext *s, TCGType type,
TCGReg a0, TCGReg a1, TCGReg a2)
{
@@ -1823,6 +1823,26 @@ static const TCGOutOpBinary outop_mul = {
.out_rrr = tgen_mul,
};
+static void tgen_muls2(TCGContext *s, TCGType type,
+ TCGReg a0, TCGReg a1, TCGReg a2, TCGReg a3)
+{
+ MIPSInsn insn = type == TCG_TYPE_I32 ? OPC_MULT : OPC_DMULT;
+ tcg_out_opc_reg(s, insn, 0, a2, a3);
+ tcg_out_opc_reg(s, OPC_MFLO, a0, 0, 0);
+ tcg_out_opc_reg(s, OPC_MFHI, a1, 0, 0);
+}
+
+static TCGConstraintSetIndex cset_mul2(TCGType type, unsigned flags)
+{
+ return use_mips32r6_instructions ? C_NotImplemented : C_O2_I2(r, r, r, r);
+}
+
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_Dynamic,
+ .base.dynamic_constraint = cset_mul2,
+ .out_rrrr = tgen_muls2,
+};
+
static void tgen_mulsh(TCGContext *s, TCGType type,
TCGReg a0, TCGReg a1, TCGReg a2)
{
@@ -2167,15 +2187,9 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type,
tcg_out_ldst(s, i1, a0, a1, a2);
break;
- case INDEX_op_muls2_i32:
- i1 = OPC_MULT;
- goto do_hilo2;
case INDEX_op_mulu2_i32:
i1 = OPC_MULTU;
goto do_hilo2;
- case INDEX_op_muls2_i64:
- i1 = OPC_DMULT;
- goto do_hilo2;
case INDEX_op_mulu2_i64:
i1 = OPC_DMULTU;
do_hilo2:
@@ -2353,9 +2367,7 @@ tcg_target_op_def(TCGOpcode op, TCGType type, unsigned flags)
case INDEX_op_setcond_i32:
case INDEX_op_setcond_i64:
return C_O1_I2(r, rz, rz);
- case INDEX_op_muls2_i32:
case INDEX_op_mulu2_i32:
- case INDEX_op_muls2_i64:
case INDEX_op_mulu2_i64:
return C_O2_I2(r, r, r, r);
@@ -3079,6 +3079,10 @@ static const TCGOutOpBinary outop_mul = {
.out_rri = tgen_muli,
};
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_NotImplemented,
+};
+
static void tgen_mulsh(TCGContext *s, TCGType type,
TCGReg a0, TCGReg a1, TCGReg a2)
{
@@ -2117,6 +2117,10 @@ static const TCGOutOpBinary outop_mul = {
.out_rrr = tgen_mul,
};
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_NotImplemented,
+};
+
static TCGConstraintSetIndex cset_mulh(TCGType type, unsigned flags)
{
return type == TCG_TYPE_I32 ? C_NotImplemented : C_O1_I2(r, r, r);
@@ -2391,6 +2391,26 @@ static const TCGOutOpBinary outop_mul = {
.out_rri = tgen_muli,
};
+static void tgen_muls2(TCGContext *s, TCGType type,
+ TCGReg a0, TCGReg a1, TCGReg a2, TCGReg a3)
+{
+ tcg_debug_assert((a1 & 1) == 0);
+ tcg_debug_assert(a0 == a1 + 1);
+ tcg_out_insn(s, RRFa, MGRK, a1, a2, a3);
+}
+
+static TCGConstraintSetIndex cset_muls2(TCGType type, unsigned flags)
+{
+ return (type == TCG_TYPE_I64 && HAVE_FACILITY(MISC_INSN_EXT2)
+ ? C_O2_I2(o, m, r, r) : C_NotImplemented);
+}
+
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_Dynamic,
+ .base.dynamic_constraint = cset_muls2,
+ .out_rrrr = tgen_muls2,
+};
+
static const TCGOutOpBinary outop_mulsh = {
.base.static_constraint = C_NotImplemented,
};
@@ -2846,11 +2866,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type,
tcg_debug_assert(args[0] == args[1] + 1);
tcg_out_insn(s, RRE, MLGR, args[1], args[3]);
break;
- case INDEX_op_muls2_i64:
- tcg_debug_assert((args[1] & 1) == 0);
- tcg_debug_assert(args[0] == args[1] + 1);
- tcg_out_insn(s, RRFa, MGRK, args[1], args[2], args[3]);
- break;
case INDEX_op_add2_i64:
if (const_args[4]) {
@@ -3451,8 +3466,6 @@ tcg_target_op_def(TCGOpcode op, TCGType type, unsigned flags)
case INDEX_op_mulu2_i64:
return C_O2_I2(o, m, 0, r);
- case INDEX_op_muls2_i64:
- return C_O2_I2(o, m, r, r);
case INDEX_op_add2_i32:
case INDEX_op_sub2_i32:
@@ -1429,6 +1429,30 @@ static const TCGOutOpBinary outop_mul = {
.out_rri = tgen_muli,
};
+/*
+ * The 32-bit multiply insns produce a full 64-bit result.
+ * Supporting 32-bit mul[us]2 opcodes avoids sign/zero-extensions
+ * before the actual multiply; we only need extract the high part
+ * into the separate operand.
+ */
+static TCGConstraintSetIndex cset_mul2(TCGType type, unsigned flags)
+{
+ return type == TCG_TYPE_I32 ? C_O2_I2(r, r, r, r) : C_NotImplemented;
+}
+
+static void tgen_muls2(TCGContext *s, TCGType type,
+ TCGReg a0, TCGReg a1, TCGReg a2, TCGReg a3)
+{
+ tcg_out_arith(s, a0, a2, a3, ARITH_SMUL);
+ tcg_out_arithi(s, a1, a0, 32, SHIFT_SRLX);
+}
+
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_Dynamic,
+ .base.dynamic_constraint = cset_mul2,
+ .out_rrrr = tgen_muls2,
+};
+
static const TCGOutOpBinary outop_mulsh = {
.base.static_constraint = C_NotImplemented,
};
@@ -1696,10 +1720,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type,
break;
case INDEX_op_mulu2_i32:
c = ARITH_UMUL;
- goto do_mul2;
- case INDEX_op_muls2_i32:
- c = ARITH_SMUL;
- do_mul2:
/* The 32-bit multiply insns produce a full 64-bit result. */
tcg_out_arithc(s, a0, a2, args[3], const_args[3], c);
tcg_out_arithi(s, a1, a0, 32, SHIFT_SRLX);
@@ -1828,7 +1848,6 @@ tcg_target_op_def(TCGOpcode op, TCGType type, unsigned flags)
case INDEX_op_sub2_i64:
return C_O2_I4(r, r, rz, rz, rJ, rJ);
case INDEX_op_mulu2_i32:
- case INDEX_op_muls2_i32:
return C_O2_I2(r, r, rz, rJ);
default:
@@ -100,8 +100,6 @@ tcg_target_op_def(TCGOpcode op, TCGType type, unsigned flags)
case INDEX_op_mulu2_i32:
case INDEX_op_mulu2_i64:
- case INDEX_op_muls2_i32:
- case INDEX_op_muls2_i64:
return C_O2_I2(r, r, r, r);
case INDEX_op_movcond_i32:
@@ -710,6 +708,24 @@ static const TCGOutOpBinary outop_mul = {
.out_rrr = tgen_mul,
};
+static TCGConstraintSetIndex cset_mul2(TCGType type, unsigned flags)
+{
+ return type == TCG_TYPE_REG ? C_O2_I2(r, r, r, r) : C_NotImplemented;
+}
+
+static void tgen_muls2(TCGContext *s, TCGType type,
+ TCGReg a0, TCGReg a1, TCGReg a2, TCGReg a3)
+{
+ tcg_out_op_rrrr(s, glue(INDEX_op_muls2_i,TCG_TARGET_REG_BITS),
+ a0, a1, a2, a3);
+}
+
+static const TCGOutOpMul2 outop_muls2 = {
+ .base.static_constraint = C_Dynamic,
+ .base.dynamic_constraint = cset_mul2,
+ .out_rrrr = tgen_muls2,
+};
+
static const TCGOutOpBinary outop_mulsh = {
.base.static_constraint = C_NotImplemented,
};
@@ -1009,7 +1025,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type,
#endif
CASE_32_64(mulu2)
- CASE_32_64(muls2)
tcg_out_op_rrrr(s, opc, args[0], args[1], args[2], args[3]);
break;
Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/aarch64/tcg-target-has.h | 2 -- tcg/arm/tcg-target-has.h | 1 - tcg/i386/tcg-target-has.h | 2 -- tcg/loongarch64/tcg-target-has.h | 2 -- tcg/mips/tcg-target-has.h | 2 -- tcg/ppc/tcg-target-has.h | 2 -- tcg/riscv/tcg-target-has.h | 2 -- tcg/s390x/tcg-target-has.h | 2 -- tcg/sparc64/tcg-target-con-set.h | 1 + tcg/sparc64/tcg-target-has.h | 2 -- tcg/tcg-has.h | 1 - tcg/tci/tcg-target-has.h | 2 -- tcg/tcg-op.c | 4 ++-- tcg/tcg.c | 26 ++++++++++++++++++++++---- tcg/tci.c | 23 ++++++++++------------- tcg/aarch64/tcg-target.c.inc | 4 ++++ tcg/arm/tcg-target.c.inc | 25 +++++++++++++------------ tcg/i386/tcg-target.c.inc | 17 ++++++++++++----- tcg/loongarch64/tcg-target.c.inc | 4 ++++ tcg/mips/tcg-target.c.inc | 28 ++++++++++++++++++++-------- tcg/ppc/tcg-target.c.inc | 4 ++++ tcg/riscv/tcg-target.c.inc | 4 ++++ tcg/s390x/tcg-target.c.inc | 27 ++++++++++++++++++++------- tcg/sparc64/tcg-target.c.inc | 29 ++++++++++++++++++++++++----- tcg/tci/tcg-target.c.inc | 21 ++++++++++++++++++--- 25 files changed, 158 insertions(+), 79 deletions(-)