izzy/xemu
0
0
Fork 0
mirror of https://github.com/izzy2lost/xemu.git synced 2026-07-06 00:20:22 -07:00
Code Issues Packages Projects Releases Wiki Activity

tcg: Convert muls2 to TCGOutOpMul2

Browse Source 
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This commit is contained in:
Richard Henderson
2025-04-28 13:40:16 -07:00
parent 97218ae918
commit 5641afdf9b
25 changed files with 158 additions and 79 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tcg/aarch64/tcg-target-has.h
-2
View File
@@ -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,
tcg/aarch64/tcg-target.c.inc
+4
View File
@@ -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;
tcg/arm/tcg-target-has.h
-1
View File
@@ -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
tcg/arm/tcg-target.c.inc
+13 -12
View File
@@ -929,14 +929,6 @@ 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 */
@@ -1973,6 +1965,19 @@ 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,
};
@@ -2246,9 +2251,6 @@ 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]);
@@ -2373,7 +2375,6 @@ 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:
tcg/i386/tcg-target-has.h
-2
View File
@@ -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
tcg/i386/tcg-target.c.inc
+12 -5
View File
@@ -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:
tcg/loongarch64/tcg-target-has.h
-2
View File
@@ -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)
tcg/loongarch64/tcg-target.c.inc
+4
View File
@@ -1421,6 +1421,10 @@ 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)
{
tcg/mips/tcg-target-has.h
-2
View File
@@ -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
tcg/mips/tcg-target.c.inc
+20 -8
View File
@@ -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)
{
@@ -2161,15 +2181,9 @@ 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:
@@ -2347,9 +2361,7 @@ 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);
tcg/ppc/tcg-target-has.h
-2
View File
@@ -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 \
tcg/ppc/tcg-target.c.inc
+4
View File
@@ -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)
{
tcg/riscv/tcg-target-has.h
-2
View File
@@ -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
tcg/riscv/tcg-target.c.inc
+4
View File
@@ -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);
tcg/s390x/tcg-target-has.h
-2
View File
@@ -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
tcg/s390x/tcg-target.c.inc
+20 -7
View File
@@ -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:
tcg/sparc64/tcg-target-con-set.h
+1
View File
@@ -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)
tcg/sparc64/tcg-target-has.h
-2
View File
@@ -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
tcg/sparc64/tcg-target.c.inc
+24 -5
View File
@@ -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:
tcg/tcg-has.h
-1
View File
@@ -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

Some files were not shown because too many files have changed in this diff Show More