aboutsummaryrefslogtreecommitdiff
path: root/tcg/sparc/tcg-target.c
diff options
context:
space:
mode:
Diffstat (limited to 'tcg/sparc/tcg-target.c')
-rw-r--r--tcg/sparc/tcg-target.c129
1 files changed, 115 insertions, 14 deletions
diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c
index 40f2ec1027..0c4b028580 100644
--- a/tcg/sparc/tcg-target.c
+++ b/tcg/sparc/tcg-target.c
@@ -197,8 +197,8 @@ static const int tcg_target_call_oarg_regs[] = {
#define ARITH_XOR (INSN_OP(2) | INSN_OP3(0x03))
#define ARITH_SUB (INSN_OP(2) | INSN_OP3(0x04))
#define ARITH_SUBCC (INSN_OP(2) | INSN_OP3(0x14))
-#define ARITH_ADDX (INSN_OP(2) | INSN_OP3(0x08))
-#define ARITH_SUBX (INSN_OP(2) | INSN_OP3(0x0c))
+#define ARITH_ADDC (INSN_OP(2) | INSN_OP3(0x08))
+#define ARITH_SUBC (INSN_OP(2) | INSN_OP3(0x0c))
#define ARITH_UMUL (INSN_OP(2) | INSN_OP3(0x0a))
#define ARITH_SMUL (INSN_OP(2) | INSN_OP3(0x0b))
#define ARITH_UDIV (INSN_OP(2) | INSN_OP3(0x0e))
@@ -209,6 +209,9 @@ static const int tcg_target_call_oarg_regs[] = {
#define ARITH_MOVCC (INSN_OP(2) | INSN_OP3(0x2c))
#define ARITH_MOVR (INSN_OP(2) | INSN_OP3(0x2f))
+#define ARITH_ADDXC (INSN_OP(2) | INSN_OP3(0x36) | INSN_OPF(0x11))
+#define ARITH_UMULXHI (INSN_OP(2) | INSN_OP3(0x36) | INSN_OPF(0x16))
+
#define SHIFT_SLL (INSN_OP(2) | INSN_OP3(0x25))
#define SHIFT_SRL (INSN_OP(2) | INSN_OP3(0x26))
#define SHIFT_SRA (INSN_OP(2) | INSN_OP3(0x27))
@@ -262,6 +265,10 @@ static const int tcg_target_call_oarg_regs[] = {
#define STW_LE (STWA | INSN_ASI(ASI_PRIMARY_LITTLE))
#define STX_LE (STXA | INSN_ASI(ASI_PRIMARY_LITTLE))
+#ifndef use_vis3_instructions
+bool use_vis3_instructions;
+#endif
+
static inline int check_fit_i64(int64_t val, unsigned int bits)
{
return val == sextract64(val, 0, bits);
@@ -657,7 +664,7 @@ static void tcg_out_movcond_i64(TCGContext *s, TCGCond cond, TCGReg ret,
static void tcg_out_setcond_i32(TCGContext *s, TCGCond cond, TCGReg ret,
TCGReg c1, int32_t c2, int c2const)
{
- /* For 32-bit comparisons, we can play games with ADDX/SUBX. */
+ /* For 32-bit comparisons, we can play games with ADDC/SUBC. */
switch (cond) {
case TCG_COND_LTU:
case TCG_COND_GEU:
@@ -668,9 +675,12 @@ static void tcg_out_setcond_i32(TCGContext *s, TCGCond cond, TCGReg ret,
case TCG_COND_NE:
/* For equality, we can transform to inequality vs zero. */
if (c2 != 0) {
- tcg_out_arithc(s, ret, c1, c2, c2const, ARITH_XOR);
+ tcg_out_arithc(s, TCG_REG_T1, c1, c2, c2const, ARITH_XOR);
+ c2 = TCG_REG_T1;
+ } else {
+ c2 = c1;
}
- c1 = TCG_REG_G0, c2 = ret, c2const = 0;
+ c1 = TCG_REG_G0, c2const = 0;
cond = (cond == TCG_COND_EQ ? TCG_COND_GEU : TCG_COND_LTU);
break;
@@ -698,15 +708,32 @@ static void tcg_out_setcond_i32(TCGContext *s, TCGCond cond, TCGReg ret,
tcg_out_cmp(s, c1, c2, c2const);
if (cond == TCG_COND_LTU) {
- tcg_out_arithi(s, ret, TCG_REG_G0, 0, ARITH_ADDX);
+ tcg_out_arithi(s, ret, TCG_REG_G0, 0, ARITH_ADDC);
} else {
- tcg_out_arithi(s, ret, TCG_REG_G0, -1, ARITH_SUBX);
+ tcg_out_arithi(s, ret, TCG_REG_G0, -1, ARITH_SUBC);
}
}
static void tcg_out_setcond_i64(TCGContext *s, TCGCond cond, TCGReg ret,
TCGReg c1, int32_t c2, int c2const)
{
+ if (use_vis3_instructions) {
+ switch (cond) {
+ case TCG_COND_NE:
+ if (c2 != 0) {
+ break;
+ }
+ c2 = c1, c2const = 0, c1 = TCG_REG_G0;
+ /* FALLTHRU */
+ case TCG_COND_LTU:
+ tcg_out_cmp(s, c1, c2, c2const);
+ tcg_out_arith(s, ret, TCG_REG_G0, TCG_REG_G0, ARITH_ADDXC);
+ return;
+ default:
+ break;
+ }
+ }
+
/* For 64-bit signed comparisons vs zero, we can avoid the compare
if the input does not overlap the output. */
if (c2 == 0 && !is_unsigned_cond(cond) && c1 != ret) {
@@ -719,9 +746,9 @@ static void tcg_out_setcond_i64(TCGContext *s, TCGCond cond, TCGReg ret,
}
}
-static void tcg_out_addsub2(TCGContext *s, TCGReg rl, TCGReg rh,
- TCGReg al, TCGReg ah, int32_t bl, int blconst,
- int32_t bh, int bhconst, int opl, int oph)
+static void tcg_out_addsub2_i32(TCGContext *s, TCGReg rl, TCGReg rh,
+ TCGReg al, TCGReg ah, int32_t bl, int blconst,
+ int32_t bh, int bhconst, int opl, int oph)
{
TCGReg tmp = TCG_REG_T1;
@@ -735,6 +762,54 @@ static void tcg_out_addsub2(TCGContext *s, TCGReg rl, TCGReg rh,
tcg_out_mov(s, TCG_TYPE_I32, rl, tmp);
}
+static void tcg_out_addsub2_i64(TCGContext *s, TCGReg rl, TCGReg rh,
+ TCGReg al, TCGReg ah, int32_t bl, int blconst,
+ int32_t bh, int bhconst, bool is_sub)
+{
+ TCGReg tmp = TCG_REG_T1;
+
+ /* Note that the low parts are fully consumed before tmp is set. */
+ if (rl != ah && (bhconst || rl != bh)) {
+ tmp = rl;
+ }
+
+ tcg_out_arithc(s, tmp, al, bl, blconst, is_sub ? ARITH_SUBCC : ARITH_ADDCC);
+
+ if (use_vis3_instructions && !is_sub) {
+ /* Note that ADDXC doesn't accept immediates. */
+ if (bhconst && bh != 0) {
+ tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_T2, bh);
+ bh = TCG_REG_T2;
+ }
+ tcg_out_arith(s, rh, ah, bh, ARITH_ADDXC);
+ } else if (bh == TCG_REG_G0) {
+ /* If we have a zero, we can perform the operation in two insns,
+ with the arithmetic first, and a conditional move into place. */
+ if (rh == ah) {
+ tcg_out_arithi(s, TCG_REG_T2, ah, 1,
+ is_sub ? ARITH_SUB : ARITH_ADD);
+ tcg_out_movcc(s, TCG_COND_LTU, MOVCC_XCC, rh, TCG_REG_T2, 0);
+ } else {
+ tcg_out_arithi(s, rh, ah, 1, is_sub ? ARITH_SUB : ARITH_ADD);
+ tcg_out_movcc(s, TCG_COND_GEU, MOVCC_XCC, rh, ah, 0);
+ }
+ } else {
+ /* Otherwise adjust BH as if there is carry into T2 ... */
+ if (bhconst) {
+ tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_T2, bh + (is_sub ? -1 : 1));
+ } else {
+ tcg_out_arithi(s, TCG_REG_T2, bh, 1,
+ is_sub ? ARITH_SUB : ARITH_ADD);
+ }
+ /* ... smoosh T2 back to original BH if carry is clear ... */
+ tcg_out_movcc(s, TCG_COND_GEU, MOVCC_XCC, TCG_REG_T2, bh, bhconst);
+ /* ... and finally perform the arithmetic with the new operand. */
+ tcg_out_arith(s, rh, ah, TCG_REG_T2, is_sub ? ARITH_SUB : ARITH_ADD);
+ }
+
+ tcg_out_mov(s, TCG_TYPE_I64, rl, tmp);
+}
+
static void tcg_out_call_nodelay(TCGContext *s, tcg_insn_unit *dest)
{
ptrdiff_t disp = tcg_pcrel_diff(s, dest);
@@ -1264,12 +1339,14 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
break;
case INDEX_op_add2_i32:
- tcg_out_addsub2(s, a0, a1, a2, args[3], args[4], const_args[4],
- args[5], const_args[5], ARITH_ADDCC, ARITH_ADDX);
+ tcg_out_addsub2_i32(s, args[0], args[1], args[2], args[3],
+ args[4], const_args[4], args[5], const_args[5],
+ ARITH_ADDCC, ARITH_ADDC);
break;
case INDEX_op_sub2_i32:
- tcg_out_addsub2(s, a0, a1, a2, args[3], args[4], const_args[4],
- args[5], const_args[5], ARITH_SUBCC, ARITH_SUBX);
+ tcg_out_addsub2_i32(s, args[0], args[1], args[2], args[3],
+ args[4], const_args[4], args[5], const_args[5],
+ ARITH_SUBCC, ARITH_SUBC);
break;
case INDEX_op_mulu2_i32:
c = ARITH_UMUL;
@@ -1351,6 +1428,17 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
case INDEX_op_movcond_i64:
tcg_out_movcond_i64(s, args[5], a0, a1, a2, c2, args[3], const_args[3]);
break;
+ case INDEX_op_add2_i64:
+ tcg_out_addsub2_i64(s, args[0], args[1], args[2], args[3], args[4],
+ const_args[4], args[5], const_args[5], false);
+ break;
+ case INDEX_op_sub2_i64:
+ tcg_out_addsub2_i64(s, args[0], args[1], args[2], args[3], args[4],
+ const_args[4], args[5], const_args[5], true);
+ break;
+ case INDEX_op_muluh_i64:
+ tcg_out_arith(s, args[0], args[1], args[2], ARITH_UMULXHI);
+ break;
gen_arith:
tcg_out_arithc(s, a0, a1, a2, c2, c);
@@ -1449,6 +1537,10 @@ static const TCGTargetOpDef sparc_op_defs[] = {
{ INDEX_op_setcond_i64, { "R", "RZ", "RJ" } },
{ INDEX_op_movcond_i64, { "R", "RZ", "RJ", "RI", "0" } },
+ { INDEX_op_add2_i64, { "R", "R", "RZ", "RZ", "RJ", "RI" } },
+ { INDEX_op_sub2_i64, { "R", "R", "RZ", "RZ", "RJ", "RI" } },
+ { INDEX_op_muluh_i64, { "R", "RZ", "RZ" } },
+
{ INDEX_op_qemu_ld_i32, { "r", "A" } },
{ INDEX_op_qemu_ld_i64, { "R", "A" } },
{ INDEX_op_qemu_st_i32, { "sZ", "A" } },
@@ -1459,6 +1551,15 @@ static const TCGTargetOpDef sparc_op_defs[] = {
static void tcg_target_init(TCGContext *s)
{
+ /* Only probe for the platform and capabilities if we havn't already
+ determined maximum values at compile time. */
+#ifndef use_vis3_instructions
+ {
+ unsigned long hwcap = qemu_getauxval(AT_HWCAP);
+ use_vis3_instructions = (hwcap & HWCAP_SPARC_VIS3) != 0;
+ }
+#endif
+
tcg_regset_set32(tcg_target_available_regs[TCG_TYPE_I32], 0, 0xffffffff);
tcg_regset_set32(tcg_target_available_regs[TCG_TYPE_I64], 0, ALL_64);
generated by cgit v1.2.3 (git 2.26.2) at 2025-02-09 16:42:21 +0000