tcg: Merge INDEX_op_sar_{i32,i64}

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This commit is contained in:
Richard Henderson 2025-01-08 08:05:18 -08:00
parent b5aafbaa83
commit 3949f365eb
7 changed files with 18 additions and 26 deletions

View file

@ -389,10 +389,10 @@ Shifts/Rotates
- | *t0* = *t1* >> *t2* (unsigned) - | *t0* = *t1* >> *t2* (unsigned)
| Unspecified behavior for negative or out-of-range shifts. | Unspecified behavior for negative or out-of-range shifts.
* - sar_i32/i64 *t0*, *t1*, *t2* * - sar *t0*, *t1*, *t2*
- | *t0* = *t1* >> *t2* (signed) - | *t0* = *t1* >> *t2* (signed)
| Unspecified behavior if *t2* < 0 or *t2* >= 32 (resp 64) | Unspecified behavior for negative or out-of-range shifts.
* - rotl_i32/i64 *t0*, *t1*, *t2* * - rotl_i32/i64 *t0*, *t1*, *t2*

View file

@ -58,6 +58,7 @@ DEF(or, 1, 2, 0, TCG_OPF_INT)
DEF(orc, 1, 2, 0, TCG_OPF_INT) DEF(orc, 1, 2, 0, TCG_OPF_INT)
DEF(rems, 1, 2, 0, TCG_OPF_INT) DEF(rems, 1, 2, 0, TCG_OPF_INT)
DEF(remu, 1, 2, 0, TCG_OPF_INT) DEF(remu, 1, 2, 0, TCG_OPF_INT)
DEF(sar, 1, 2, 0, TCG_OPF_INT)
DEF(shl, 1, 2, 0, TCG_OPF_INT) DEF(shl, 1, 2, 0, TCG_OPF_INT)
DEF(shr, 1, 2, 0, TCG_OPF_INT) DEF(shr, 1, 2, 0, TCG_OPF_INT)
DEF(sub, 1, 2, 0, TCG_OPF_INT) DEF(sub, 1, 2, 0, TCG_OPF_INT)
@ -76,7 +77,6 @@ DEF(st8_i32, 0, 2, 1, 0)
DEF(st16_i32, 0, 2, 1, 0) DEF(st16_i32, 0, 2, 1, 0)
DEF(st_i32, 0, 2, 1, 0) DEF(st_i32, 0, 2, 1, 0)
/* shifts/rotates */ /* shifts/rotates */
DEF(sar_i32, 1, 2, 0, 0)
DEF(rotl_i32, 1, 2, 0, 0) DEF(rotl_i32, 1, 2, 0, 0)
DEF(rotr_i32, 1, 2, 0, 0) DEF(rotr_i32, 1, 2, 0, 0)
DEF(deposit_i32, 1, 2, 2, 0) DEF(deposit_i32, 1, 2, 2, 0)
@ -115,7 +115,6 @@ DEF(st16_i64, 0, 2, 1, 0)
DEF(st32_i64, 0, 2, 1, 0) DEF(st32_i64, 0, 2, 1, 0)
DEF(st_i64, 0, 2, 1, 0) DEF(st_i64, 0, 2, 1, 0)
/* shifts/rotates */ /* shifts/rotates */
DEF(sar_i64, 1, 2, 0, 0)
DEF(rotl_i64, 1, 2, 0, 0) DEF(rotl_i64, 1, 2, 0, 0)
DEF(rotr_i64, 1, 2, 0, 0) DEF(rotr_i64, 1, 2, 0, 0)
DEF(deposit_i64, 1, 2, 2, 0) DEF(deposit_i64, 1, 2, 2, 0)

View file

@ -458,10 +458,10 @@ static uint64_t do_constant_folding_2(TCGOpcode op, TCGType type,
} }
return (uint64_t)x >> (y & 63); return (uint64_t)x >> (y & 63);
case INDEX_op_sar_i32: case INDEX_op_sar:
return (int32_t)x >> (y & 31); if (type == TCG_TYPE_I32) {
return (int32_t)x >> (y & 31);
case INDEX_op_sar_i64: }
return (int64_t)x >> (y & 63); return (int64_t)x >> (y & 63);
case INDEX_op_rotr_i32: case INDEX_op_rotr_i32:
@ -2600,7 +2600,7 @@ static bool fold_shift(OptContext *ctx, TCGOp *op)
} }
switch (op->opc) { switch (op->opc) {
CASE_OP_32_64(sar): case INDEX_op_sar:
/* /*
* Arithmetic right shift will not reduce the number of * Arithmetic right shift will not reduce the number of
* input sign repetitions. * input sign repetitions.
@ -3027,7 +3027,7 @@ void tcg_optimize(TCGContext *s)
break; break;
CASE_OP_32_64(rotl): CASE_OP_32_64(rotl):
CASE_OP_32_64(rotr): CASE_OP_32_64(rotr):
CASE_OP_32_64(sar): case INDEX_op_sar:
case INDEX_op_shl: case INDEX_op_shl:
case INDEX_op_shr: case INDEX_op_shr:
done = fold_shift(&ctx, op); done = fold_shift(&ctx, op);

View file

@ -511,7 +511,7 @@ void tcg_gen_shri_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
void tcg_gen_sar_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2) void tcg_gen_sar_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
{ {
tcg_gen_op3_i32(INDEX_op_sar_i32, ret, arg1, arg2); tcg_gen_op3_i32(INDEX_op_sar, ret, arg1, arg2);
} }
void tcg_gen_sari_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2) void tcg_gen_sari_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
@ -1624,7 +1624,7 @@ void tcg_gen_shr_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
void tcg_gen_sar_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2) void tcg_gen_sar_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
{ {
if (TCG_TARGET_REG_BITS == 64) { if (TCG_TARGET_REG_BITS == 64) {
tcg_gen_op3_i64(INDEX_op_sar_i64, ret, arg1, arg2); tcg_gen_op3_i64(INDEX_op_sar, ret, arg1, arg2);
} else { } else {
gen_helper_sar_i64(ret, arg1, arg2); gen_helper_sar_i64(ret, arg1, arg2);
} }

View file

@ -1042,8 +1042,7 @@ static const TCGOutOp * const all_outop[NB_OPS] = {
OUTOP(INDEX_op_orc, TCGOutOpBinary, outop_orc), OUTOP(INDEX_op_orc, TCGOutOpBinary, outop_orc),
OUTOP(INDEX_op_rems, TCGOutOpBinary, outop_rems), OUTOP(INDEX_op_rems, TCGOutOpBinary, outop_rems),
OUTOP(INDEX_op_remu, TCGOutOpBinary, outop_remu), OUTOP(INDEX_op_remu, TCGOutOpBinary, outop_remu),
OUTOP(INDEX_op_sar_i32, TCGOutOpBinary, outop_sar), OUTOP(INDEX_op_sar, TCGOutOpBinary, outop_sar),
OUTOP(INDEX_op_sar_i64, TCGOutOpBinary, outop_sar),
OUTOP(INDEX_op_shl, TCGOutOpBinary, outop_shl), OUTOP(INDEX_op_shl, TCGOutOpBinary, outop_shl),
OUTOP(INDEX_op_shr, TCGOutOpBinary, outop_shr), OUTOP(INDEX_op_shr, TCGOutOpBinary, outop_shr),
OUTOP(INDEX_op_sub, TCGOutOpSubtract, outop_sub), OUTOP(INDEX_op_sub, TCGOutOpSubtract, outop_sub),
@ -5421,8 +5420,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
case INDEX_op_orc: case INDEX_op_orc:
case INDEX_op_rems: case INDEX_op_rems:
case INDEX_op_remu: case INDEX_op_remu:
case INDEX_op_sar_i32: case INDEX_op_sar:
case INDEX_op_sar_i64:
case INDEX_op_shl: case INDEX_op_shl:
case INDEX_op_shr: case INDEX_op_shr:
case INDEX_op_xor: case INDEX_op_xor:

View file

@ -625,9 +625,10 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
tci_args_rrr(insn, &r0, &r1, &r2); tci_args_rrr(insn, &r0, &r1, &r2);
regs[r0] = regs[r1] >> (regs[r2] % TCG_TARGET_REG_BITS); regs[r0] = regs[r1] >> (regs[r2] % TCG_TARGET_REG_BITS);
break; break;
case INDEX_op_sar_i32: case INDEX_op_sar:
tci_args_rrr(insn, &r0, &r1, &r2); tci_args_rrr(insn, &r0, &r1, &r2);
regs[r0] = (int32_t)regs[r1] >> (regs[r2] & 31); regs[r0] = ((tcg_target_long)regs[r1]
>> (regs[r2] % TCG_TARGET_REG_BITS));
break; break;
#if TCG_TARGET_HAS_rot_i32 #if TCG_TARGET_HAS_rot_i32
case INDEX_op_rotl_i32: case INDEX_op_rotl_i32:
@ -787,10 +788,6 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
/* Shift/rotate operations (64 bit). */ /* Shift/rotate operations (64 bit). */
case INDEX_op_sar_i64:
tci_args_rrr(insn, &r0, &r1, &r2);
regs[r0] = (int64_t)regs[r1] >> (regs[r2] & 63);
break;
#if TCG_TARGET_HAS_rot_i64 #if TCG_TARGET_HAS_rot_i64
case INDEX_op_rotl_i64: case INDEX_op_rotl_i64:
tci_args_rrr(insn, &r0, &r1, &r2); tci_args_rrr(insn, &r0, &r1, &r2);
@ -1073,12 +1070,11 @@ int print_insn_tci(bfd_vma addr, disassemble_info *info)
case INDEX_op_orc: case INDEX_op_orc:
case INDEX_op_rems: case INDEX_op_rems:
case INDEX_op_remu: case INDEX_op_remu:
case INDEX_op_sar:
case INDEX_op_shl: case INDEX_op_shl:
case INDEX_op_shr: case INDEX_op_shr:
case INDEX_op_sub: case INDEX_op_sub:
case INDEX_op_xor: case INDEX_op_xor:
case INDEX_op_sar_i32:
case INDEX_op_sar_i64:
case INDEX_op_rotl_i32: case INDEX_op_rotl_i32:
case INDEX_op_rotl_i64: case INDEX_op_rotl_i64:
case INDEX_op_rotr_i32: case INDEX_op_rotr_i32:

View file

@ -779,7 +779,7 @@ static void tgen_sar(TCGContext *s, TCGType type,
tcg_out_ext32s(s, TCG_REG_TMP, a1); tcg_out_ext32s(s, TCG_REG_TMP, a1);
a1 = TCG_REG_TMP; a1 = TCG_REG_TMP;
} }
tcg_out_op_rrr(s, glue(INDEX_op_sar_i,TCG_TARGET_REG_BITS), a0, a1, a2); tcg_out_op_rrr(s, INDEX_op_sar, a0, a1, a2);
} }
static const TCGOutOpBinary outop_sar = { static const TCGOutOpBinary outop_sar = {
@ -897,7 +897,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type,
tcg_out_ldst(s, opc, args[0], args[1], args[2]); tcg_out_ldst(s, opc, args[0], args[1], args[2]);
break; break;
CASE_32_64(sar)
CASE_32_64(rotl) /* Optional (TCG_TARGET_HAS_rot_*). */ CASE_32_64(rotl) /* Optional (TCG_TARGET_HAS_rot_*). */
CASE_32_64(rotr) /* Optional (TCG_TARGET_HAS_rot_*). */ CASE_32_64(rotr) /* Optional (TCG_TARGET_HAS_rot_*). */
CASE_32_64(clz) /* Optional (TCG_TARGET_HAS_clz_*). */ CASE_32_64(clz) /* Optional (TCG_TARGET_HAS_clz_*). */