tcg: Merge INDEX_op_and_{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 2025-01-06 10:32:44 -08:00
parent bf40f915fc
commit c3b920b3d6
8 changed files with 24 additions and 45 deletions

View file

@ -303,7 +303,7 @@ Logical
.. list-table:: .. list-table::
* - and_i32/i64 *t0*, *t1*, *t2* * - and *t0*, *t1*, *t2*
- | *t0* = *t1* & *t2* - | *t0* = *t1* & *t2*

View file

@ -40,6 +40,7 @@ DEF(mb, 0, 0, 1, TCG_OPF_NOT_PRESENT)
DEF(mov, 1, 1, 0, TCG_OPF_INT | TCG_OPF_NOT_PRESENT) DEF(mov, 1, 1, 0, TCG_OPF_INT | TCG_OPF_NOT_PRESENT)
DEF(add, 1, 2, 0, TCG_OPF_INT) DEF(add, 1, 2, 0, TCG_OPF_INT)
DEF(and, 1, 2, 0, TCG_OPF_INT)
DEF(setcond_i32, 1, 2, 1, 0) DEF(setcond_i32, 1, 2, 1, 0)
DEF(negsetcond_i32, 1, 2, 1, 0) DEF(negsetcond_i32, 1, 2, 1, 0)
@ -62,7 +63,6 @@ DEF(rem_i32, 1, 2, 0, 0)
DEF(remu_i32, 1, 2, 0, 0) DEF(remu_i32, 1, 2, 0, 0)
DEF(div2_i32, 2, 3, 0, 0) DEF(div2_i32, 2, 3, 0, 0)
DEF(divu2_i32, 2, 3, 0, 0) DEF(divu2_i32, 2, 3, 0, 0)
DEF(and_i32, 1, 2, 0, 0)
DEF(or_i32, 1, 2, 0, 0) DEF(or_i32, 1, 2, 0, 0)
DEF(xor_i32, 1, 2, 0, 0) DEF(xor_i32, 1, 2, 0, 0)
/* shifts/rotates */ /* shifts/rotates */
@ -124,7 +124,6 @@ DEF(rem_i64, 1, 2, 0, 0)
DEF(remu_i64, 1, 2, 0, 0) DEF(remu_i64, 1, 2, 0, 0)
DEF(div2_i64, 2, 3, 0, 0) DEF(div2_i64, 2, 3, 0, 0)
DEF(divu2_i64, 2, 3, 0, 0) DEF(divu2_i64, 2, 3, 0, 0)
DEF(and_i64, 1, 2, 0, 0)
DEF(or_i64, 1, 2, 0, 0) DEF(or_i64, 1, 2, 0, 0)
DEF(xor_i64, 1, 2, 0, 0) DEF(xor_i64, 1, 2, 0, 0)
/* shifts/rotates */ /* shifts/rotates */

View file

@ -1943,7 +1943,7 @@ static void decode_gusa(DisasContext *ctx, CPUSH4State *env)
op_opc = INDEX_op_add; op_opc = INDEX_op_add;
goto do_reg_op; goto do_reg_op;
case 0x2009: /* and Rm,Rn */ case 0x2009: /* and Rm,Rn */
op_opc = INDEX_op_and_i32; op_opc = INDEX_op_and;
goto do_reg_op; goto do_reg_op;
case 0x200a: /* xor Rm,Rn */ case 0x200a: /* xor Rm,Rn */
op_opc = INDEX_op_xor_i32; op_opc = INDEX_op_xor_i32;
@ -2105,7 +2105,7 @@ static void decode_gusa(DisasContext *ctx, CPUSH4State *env)
} }
break; break;
case INDEX_op_and_i32: case INDEX_op_and:
if (op_dst != st_src) { if (op_dst != st_src) {
goto fail; goto fail;
} }

View file

@ -433,7 +433,8 @@ static uint64_t do_constant_folding_2(TCGOpcode op, uint64_t x, uint64_t y)
CASE_OP_32_64(mul): CASE_OP_32_64(mul):
return x * y; return x * y;
CASE_OP_32_64_VEC(and): case INDEX_op_and:
case INDEX_op_and_vec:
return x & y; return x & y;
CASE_OP_32_64_VEC(or): CASE_OP_32_64_VEC(or):
@ -802,9 +803,7 @@ static int do_constant_folding_cond1(OptContext *ctx, TCGOp *op, TCGArg dest,
/* Expand to AND with a temporary if no backend support. */ /* Expand to AND with a temporary if no backend support. */
if (!TCG_TARGET_HAS_tst) { if (!TCG_TARGET_HAS_tst) {
TCGOpcode and_opc = (ctx->type == TCG_TYPE_I32 TCGOp *op2 = opt_insert_before(ctx, op, INDEX_op_and, 3);
? INDEX_op_and_i32 : INDEX_op_and_i64);
TCGOp *op2 = opt_insert_before(ctx, op, and_opc, 3);
TCGArg tmp = arg_new_temp(ctx); TCGArg tmp = arg_new_temp(ctx);
op2->args[0] = tmp; op2->args[0] = tmp;
@ -897,8 +896,8 @@ static int do_constant_folding_cond2(OptContext *ctx, TCGOp *op, TCGArg *args)
/* Expand to AND with a temporary if no backend support. */ /* Expand to AND with a temporary if no backend support. */
if (!TCG_TARGET_HAS_tst && is_tst_cond(c)) { if (!TCG_TARGET_HAS_tst && is_tst_cond(c)) {
TCGOp *op1 = opt_insert_before(ctx, op, INDEX_op_and_i32, 3); TCGOp *op1 = opt_insert_before(ctx, op, INDEX_op_and, 3);
TCGOp *op2 = opt_insert_before(ctx, op, INDEX_op_and_i32, 3); TCGOp *op2 = opt_insert_before(ctx, op, INDEX_op_and, 3);
TCGArg t1 = arg_new_temp(ctx); TCGArg t1 = arg_new_temp(ctx);
TCGArg t2 = arg_new_temp(ctx); TCGArg t2 = arg_new_temp(ctx);
@ -1709,8 +1708,7 @@ static bool fold_deposit(OptContext *ctx, TCGOp *op)
TempOptInfo *t2 = arg_info(op->args[2]); TempOptInfo *t2 = arg_info(op->args[2]);
int ofs = op->args[3]; int ofs = op->args[3];
int len = op->args[4]; int len = op->args[4];
int width; int width = 8 * tcg_type_size(ctx->type);
TCGOpcode and_opc;
uint64_t z_mask, s_mask; uint64_t z_mask, s_mask;
if (ti_is_const(t1) && ti_is_const(t2)) { if (ti_is_const(t1) && ti_is_const(t2)) {
@ -1719,24 +1717,11 @@ static bool fold_deposit(OptContext *ctx, TCGOp *op)
ti_const_val(t2))); ti_const_val(t2)));
} }
switch (ctx->type) {
case TCG_TYPE_I32:
and_opc = INDEX_op_and_i32;
width = 32;
break;
case TCG_TYPE_I64:
and_opc = INDEX_op_and_i64;
width = 64;
break;
default:
g_assert_not_reached();
}
/* Inserting a value into zero at offset 0. */ /* Inserting a value into zero at offset 0. */
if (ti_is_const_val(t1, 0) && ofs == 0) { if (ti_is_const_val(t1, 0) && ofs == 0) {
uint64_t mask = MAKE_64BIT_MASK(0, len); uint64_t mask = MAKE_64BIT_MASK(0, len);
op->opc = and_opc; op->opc = INDEX_op_and;
op->args[1] = op->args[2]; op->args[1] = op->args[2];
op->args[2] = arg_new_constant(ctx, mask); op->args[2] = arg_new_constant(ctx, mask);
return fold_and(ctx, op); return fold_and(ctx, op);
@ -1746,7 +1731,7 @@ static bool fold_deposit(OptContext *ctx, TCGOp *op)
if (ti_is_const_val(t2, 0)) { if (ti_is_const_val(t2, 0)) {
uint64_t mask = deposit64(-1, ofs, len, 0); uint64_t mask = deposit64(-1, ofs, len, 0);
op->opc = and_opc; op->opc = INDEX_op_and;
op->args[2] = arg_new_constant(ctx, mask); op->args[2] = arg_new_constant(ctx, mask);
return fold_and(ctx, op); return fold_and(ctx, op);
} }
@ -2297,7 +2282,7 @@ static int fold_setcond_zmask(OptContext *ctx, TCGOp *op, bool neg)
static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg) static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg)
{ {
TCGOpcode and_opc, sub_opc, xor_opc, neg_opc, shr_opc; TCGOpcode sub_opc, xor_opc, neg_opc, shr_opc;
TCGOpcode uext_opc = 0, sext_opc = 0; TCGOpcode uext_opc = 0, sext_opc = 0;
TCGCond cond = op->args[3]; TCGCond cond = op->args[3];
TCGArg ret, src1, src2; TCGArg ret, src1, src2;
@ -2319,7 +2304,6 @@ static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg)
switch (ctx->type) { switch (ctx->type) {
case TCG_TYPE_I32: case TCG_TYPE_I32:
and_opc = INDEX_op_and_i32;
sub_opc = INDEX_op_sub_i32; sub_opc = INDEX_op_sub_i32;
xor_opc = INDEX_op_xor_i32; xor_opc = INDEX_op_xor_i32;
shr_opc = INDEX_op_shr_i32; shr_opc = INDEX_op_shr_i32;
@ -2332,7 +2316,6 @@ static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg)
} }
break; break;
case TCG_TYPE_I64: case TCG_TYPE_I64:
and_opc = INDEX_op_and_i64;
sub_opc = INDEX_op_sub_i64; sub_opc = INDEX_op_sub_i64;
xor_opc = INDEX_op_xor_i64; xor_opc = INDEX_op_xor_i64;
shr_opc = INDEX_op_shr_i64; shr_opc = INDEX_op_shr_i64;
@ -2371,7 +2354,7 @@ static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg)
op2->args[2] = arg_new_constant(ctx, sh); op2->args[2] = arg_new_constant(ctx, sh);
src1 = ret; src1 = ret;
} }
op->opc = and_opc; op->opc = INDEX_op_and;
op->args[1] = src1; op->args[1] = src1;
op->args[2] = arg_new_constant(ctx, 1); op->args[2] = arg_new_constant(ctx, 1);
} }
@ -2848,7 +2831,8 @@ void tcg_optimize(TCGContext *s)
CASE_OP_32_64(add2): CASE_OP_32_64(add2):
done = fold_add2(&ctx, op); done = fold_add2(&ctx, op);
break; break;
CASE_OP_32_64_VEC(and): case INDEX_op_and:
case INDEX_op_and_vec:
done = fold_and(&ctx, op); done = fold_and(&ctx, op);
break; break;
CASE_OP_32_64_VEC(andc): CASE_OP_32_64_VEC(andc):

View file

@ -401,7 +401,7 @@ void tcg_gen_neg_i32(TCGv_i32 ret, TCGv_i32 arg)
void tcg_gen_and_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2) void tcg_gen_and_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
{ {
tcg_gen_op3_i32(INDEX_op_and_i32, ret, arg1, arg2); tcg_gen_op3_i32(INDEX_op_and, ret, arg1, arg2);
} }
void tcg_gen_andi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2) void tcg_gen_andi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
@ -1575,7 +1575,7 @@ void tcg_gen_sub_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
void tcg_gen_and_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2) void tcg_gen_and_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_and_i64, ret, arg1, arg2); tcg_gen_op3_i64(INDEX_op_and, ret, arg1, arg2);
} else { } else {
tcg_gen_and_i32(TCGV_LOW(ret), TCGV_LOW(arg1), TCGV_LOW(arg2)); tcg_gen_and_i32(TCGV_LOW(ret), TCGV_LOW(arg1), TCGV_LOW(arg2));
tcg_gen_and_i32(TCGV_HIGH(ret), TCGV_HIGH(arg1), TCGV_HIGH(arg2)); tcg_gen_and_i32(TCGV_HIGH(ret), TCGV_HIGH(arg1), TCGV_HIGH(arg2));

View file

@ -1005,8 +1005,7 @@ QEMU_BUILD_BUG_ON((int)(offsetof(CPUNegativeOffsetState, tlb.f[0]) -
/* Register allocation descriptions for every TCGOpcode. */ /* Register allocation descriptions for every TCGOpcode. */
static const TCGOutOp * const all_outop[NB_OPS] = { static const TCGOutOp * const all_outop[NB_OPS] = {
OUTOP(INDEX_op_add, TCGOutOpBinary, outop_add), OUTOP(INDEX_op_add, TCGOutOpBinary, outop_add),
OUTOP(INDEX_op_and_i32, TCGOutOpBinary, outop_and), OUTOP(INDEX_op_and, TCGOutOpBinary, outop_and),
OUTOP(INDEX_op_and_i64, TCGOutOpBinary, outop_and),
}; };
#undef OUTOP #undef OUTOP
@ -2208,6 +2207,7 @@ bool tcg_op_supported(TCGOpcode op, TCGType type, unsigned flags)
return TCG_TARGET_HAS_qemu_ldst_i128; return TCG_TARGET_HAS_qemu_ldst_i128;
case INDEX_op_add: case INDEX_op_add:
case INDEX_op_and:
case INDEX_op_mov: case INDEX_op_mov:
return has_type; return has_type;
@ -2225,7 +2225,6 @@ bool tcg_op_supported(TCGOpcode op, TCGType type, unsigned flags)
case INDEX_op_sub_i32: case INDEX_op_sub_i32:
case INDEX_op_neg_i32: case INDEX_op_neg_i32:
case INDEX_op_mul_i32: case INDEX_op_mul_i32:
case INDEX_op_and_i32:
case INDEX_op_or_i32: case INDEX_op_or_i32:
case INDEX_op_xor_i32: case INDEX_op_xor_i32:
case INDEX_op_shl_i32: case INDEX_op_shl_i32:
@ -2308,7 +2307,6 @@ bool tcg_op_supported(TCGOpcode op, TCGType type, unsigned flags)
case INDEX_op_sub_i64: case INDEX_op_sub_i64:
case INDEX_op_neg_i64: case INDEX_op_neg_i64:
case INDEX_op_mul_i64: case INDEX_op_mul_i64:
case INDEX_op_and_i64:
case INDEX_op_or_i64: case INDEX_op_or_i64:
case INDEX_op_xor_i64: case INDEX_op_xor_i64:
case INDEX_op_shl_i64: case INDEX_op_shl_i64:
@ -5444,8 +5442,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
break; break;
case INDEX_op_add: case INDEX_op_add:
case INDEX_op_and_i32: case INDEX_op_and:
case INDEX_op_and_i64:
{ {
const TCGOutOpBinary *out = const TCGOutOpBinary *out =
container_of(all_outop[op->opc], TCGOutOpBinary, base); container_of(all_outop[op->opc], TCGOutOpBinary, base);

View file

@ -535,7 +535,7 @@ 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]; regs[r0] = regs[r1] * regs[r2];
break; break;
CASE_32_64(and) case INDEX_op_and:
tci_args_rrr(insn, &r0, &r1, &r2); tci_args_rrr(insn, &r0, &r1, &r2);
regs[r0] = regs[r1] & regs[r2]; regs[r0] = regs[r1] & regs[r2];
break; break;
@ -1083,12 +1083,11 @@ int print_insn_tci(bfd_vma addr, disassemble_info *info)
break; break;
case INDEX_op_add: case INDEX_op_add:
case INDEX_op_and:
case INDEX_op_sub_i32: case INDEX_op_sub_i32:
case INDEX_op_sub_i64: case INDEX_op_sub_i64:
case INDEX_op_mul_i32: case INDEX_op_mul_i32:
case INDEX_op_mul_i64: case INDEX_op_mul_i64:
case INDEX_op_and_i32:
case INDEX_op_and_i64:
case INDEX_op_or_i32: case INDEX_op_or_i32:
case INDEX_op_or_i64: case INDEX_op_or_i64:
case INDEX_op_xor_i32: case INDEX_op_xor_i32:

View file

@ -651,7 +651,7 @@ static const TCGOutOpBinary outop_add = {
static void tgen_and(TCGContext *s, TCGType type, static void tgen_and(TCGContext *s, TCGType type,
TCGReg a0, TCGReg a1, TCGReg a2) TCGReg a0, TCGReg a1, TCGReg a2)
{ {
tcg_out_op_rrr(s, glue(INDEX_op_and_i,TCG_TARGET_REG_BITS), a0, a1, a2); tcg_out_op_rrr(s, INDEX_op_and, a0, a1, a2);
} }
static const TCGOutOpBinary outop_and = { static const TCGOutOpBinary outop_and = {