Signed-off-by: Richard Henderson <richard.hender...@linaro.org> --- include/tcg/tcg-opc.h | 5 +---- tcg/optimize.c | 10 +++++----- tcg/tcg-op.c | 4 ++-- tcg/tcg.c | 6 ++---- target/i386/tcg/emit.c.inc | 12 +----------- tcg/aarch64/tcg-target.c.inc | 6 ++---- tcg/arm/tcg-target.c.inc | 4 ++-- tcg/i386/tcg-target.c.inc | 5 ++--- 8 files changed, 17 insertions(+), 35 deletions(-)
diff --git a/include/tcg/tcg-opc.h b/include/tcg/tcg-opc.h index d390d639ee..c9069525a1 100644 --- a/include/tcg/tcg-opc.h +++ b/include/tcg/tcg-opc.h @@ -72,6 +72,7 @@ DEF(rotr, 1, 2, 0, TCG_OPF_INT) DEF(extract, 1, 1, 2, TCG_OPF_INT) DEF(sextract, 1, 1, 2, TCG_OPF_INT) DEF(deposit, 1, 2, 2, TCG_OPF_INT) +DEF(extract2, 1, 2, 1, TCG_OPF_INT) DEF(brcond, 0, 2, 2, TCG_OPF_BB_END | TCG_OPF_COND_BRANCH | TCG_OPF_INT) DEF(setcond, 1, 2, 1, TCG_OPF_INT) @@ -81,8 +82,6 @@ DEF(movcond, 1, 4, 1, TCG_OPF_INT) /* load/store */ DEF(ld_i32, 1, 1, 2, 0) DEF(st_i32, 0, 2, 2, 0) -/* shifts/rotates */ -DEF(extract2_i32, 1, 2, 1, 0) DEF(brcond2_i32, 0, 4, 2, TCG_OPF_BB_END | TCG_OPF_COND_BRANCH) DEF(setcond2_i32, 1, 4, 1, 0) @@ -96,8 +95,6 @@ DEF(ctpop_i32, 1, 1, 0, 0) /* load/store */ DEF(ld_i64, 1, 1, 2, 0) DEF(st_i64, 0, 2, 2, 0) -/* shifts/rotates */ -DEF(extract2_i64, 1, 2, 1, 0) /* size changing ops */ DEF(ext_i32_i64, 1, 1, 0, 0) diff --git a/tcg/optimize.c b/tcg/optimize.c index 25ab293a73..82606a00f5 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -1831,12 +1831,12 @@ static bool fold_extract2(OptContext *ctx, TCGOp *op) uint64_t v2 = arg_info(op->args[2])->val; int shr = op->args[3]; - if (op->opc == INDEX_op_extract2_i64) { - v1 >>= shr; - v2 <<= 64 - shr; - } else { + if (op->type == TCG_TYPE_I32) { v1 = (uint32_t)v1 >> shr; v2 = (uint64_t)((int32_t)v2 << (32 - shr)); + } else { + v1 >>= shr; + v2 <<= 64 - shr; } return tcg_opt_gen_movi(ctx, op, op->args[0], v1 | v2); } @@ -2766,7 +2766,7 @@ void tcg_optimize(TCGContext *s) case INDEX_op_extract: done = fold_extract(&ctx, op); break; - CASE_OP_32_64(extract2): + case INDEX_op_extract2: done = fold_extract2(&ctx, op); break; case INDEX_op_ext_i32_i64: diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c index 65a6031eaf..f3f758b15b 100644 --- a/tcg/tcg-op.c +++ b/tcg/tcg-op.c @@ -1070,7 +1070,7 @@ void tcg_gen_extract2_i32(TCGv_i32 ret, TCGv_i32 al, TCGv_i32 ah, } else if (al == ah) { tcg_gen_rotri_i32(ret, al, ofs); } else if (TCG_TARGET_HAS_extract2(TCG_TYPE_I32)) { - tcg_gen_op4i_i32(INDEX_op_extract2_i32, ret, al, ah, ofs); + tcg_gen_op4i_i32(INDEX_op_extract2, ret, al, ah, ofs); } else { TCGv_i32 t0 = tcg_temp_ebb_new_i32(); tcg_gen_shri_i32(t0, al, ofs); @@ -2802,7 +2802,7 @@ void tcg_gen_extract2_i64(TCGv_i64 ret, TCGv_i64 al, TCGv_i64 ah, } else if (al == ah) { tcg_gen_rotri_i64(ret, al, ofs); } else if (TCG_TARGET_HAS_extract2(TCG_TYPE_I64)) { - tcg_gen_op4i_i64(INDEX_op_extract2_i64, ret, al, ah, ofs); + tcg_gen_op4i_i64(INDEX_op_extract2, ret, al, ah, ofs); } else { TCGv_i64 t0 = tcg_temp_ebb_new_i64(); tcg_gen_shri_i64(t0, al, ofs); diff --git a/tcg/tcg.c b/tcg/tcg.c index 77f28147a1..e8377a9bbe 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -2224,6 +2224,8 @@ bool tcg_op_supported(TCGOpcode op, TCGType type) return has_type && TCG_TARGET_HAS_div2(type); case INDEX_op_eqv: return has_type && TCG_TARGET_HAS_eqv(type); + case INDEX_op_extract2: + return has_type && TCG_TARGET_HAS_extract2(type); case INDEX_op_muls2: return has_type && TCG_TARGET_HAS_muls2(type); case INDEX_op_mulsh: @@ -2251,8 +2253,6 @@ bool tcg_op_supported(TCGOpcode op, TCGType type) case INDEX_op_sub2: return has_type && TCG_TARGET_HAS_sub2(type); - case INDEX_op_extract2_i32: - return TCG_TARGET_HAS_extract2(TCG_TYPE_I32); case INDEX_op_bswap16_i32: case INDEX_op_bswap32_i32: return TCG_TARGET_HAS_bswap(TCG_TYPE_I32); @@ -2271,8 +2271,6 @@ bool tcg_op_supported(TCGOpcode op, TCGType type) case INDEX_op_extrh_i64_i32: return TCG_TARGET_REG_BITS == 64; - case INDEX_op_extract2_i64: - return TCG_TARGET_REG_BITS == 64 && TCG_TARGET_HAS_extract2(TCG_TYPE_I64); case INDEX_op_bswap16_i64: case INDEX_op_bswap32_i64: case INDEX_op_bswap64_i64: diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc index 00f7a6123a..c8fb3c36d9 100644 --- a/target/i386/tcg/emit.c.inc +++ b/target/i386/tcg/emit.c.inc @@ -19,16 +19,6 @@ * License along with this library; if not, see <http://www.gnu.org/licenses/>. */ -/* - * Sometimes, knowing what the backend has can produce better code. - * The exact opcode to check depends on 32- vs. 64-bit. - */ -#ifdef TARGET_X86_64 -#define INDEX_op_extract2_tl INDEX_op_extract2_i64 -#else -#define INDEX_op_extract2_tl INDEX_op_extract2_i32 -#endif - #define MMX_OFFSET(reg) \ ({ assert((reg) >= 0 && (reg) <= 7); \ offsetof(CPUX86State, fpregs[reg].mmx); }) @@ -2993,7 +2983,7 @@ static void gen_PMOVMSKB(DisasContext *s, X86DecodedInsn *decode) tcg_gen_ld8u_tl(s->T0, tcg_env, offsetof(CPUX86State, xmm_t0.ZMM_B(vec_len - 1))); while (vec_len > 8) { vec_len -= 8; - if (tcg_op_supported(INDEX_op_extract2_tl, TCG_TYPE_TL)) { + if (tcg_op_supported(INDEX_op_extract2, TCG_TYPE_TL)) { /* * Load the next byte of the result into the high byte of T. * TCG does a similar expansion of deposit to shl+extract2; by diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc index 57f4e05b5c..d982ac38a1 100644 --- a/tcg/aarch64/tcg-target.c.inc +++ b/tcg/aarch64/tcg-target.c.inc @@ -2426,8 +2426,7 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType ext, tcg_out_sbfm(s, ext, a0, a1, a2, a2 + args[3] - 1); break; - case INDEX_op_extract2_i64: - case INDEX_op_extract2_i32: + case INDEX_op_extract2: tcg_out_extr(s, ext, a0, REG0(2), REG0(1), args[3]); break; @@ -2998,8 +2997,7 @@ static TCGConstraintSetIndex tcg_target_op_def(const TCGOp *op) case INDEX_op_deposit: return C_O1_I2(r, 0, rZ); - case INDEX_op_extract2_i32: - case INDEX_op_extract2_i64: + case INDEX_op_extract2: return C_O1_I2(r, rZ, rZ); case INDEX_op_add2: diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc index 1bf020ed67..ef33cf0ac7 100644 --- a/tcg/arm/tcg-target.c.inc +++ b/tcg/arm/tcg-target.c.inc @@ -2133,7 +2133,7 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type, case INDEX_op_sextract: tcg_out_sextract(s, COND_AL, args[0], args[1], args[2], args[3]); break; - case INDEX_op_extract2_i32: + case INDEX_op_extract2: /* ??? These optimization vs zero should be generic. */ /* ??? But we can't substitute 2 for 1 in the opcode stream yet. */ if (const_args[1]) { @@ -2229,7 +2229,7 @@ static TCGConstraintSetIndex tcg_target_op_def(const TCGOp *op) return C_O0_I2(r, rIN); case INDEX_op_deposit: return C_O1_I2(r, 0, rZ); - case INDEX_op_extract2_i32: + case INDEX_op_extract2: return C_O1_I2(r, rZ, rZ); case INDEX_op_movcond: return C_O1_I4(r, r, rIN, rIK, 0); diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc index 917775b7f6..61acf23708 100644 --- a/tcg/i386/tcg-target.c.inc +++ b/tcg/i386/tcg-target.c.inc @@ -3093,7 +3093,7 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, TCGType type, } break; - OP_32_64(extract2): + case INDEX_op_extract2: /* Note that SHRD outputs to the r/m operand. */ tcg_out_modrm(s, OPC_SHRD_Ib + rexw, a2, a0); tcg_out8(s, args[3]); @@ -3726,8 +3726,7 @@ static TCGConstraintSetIndex tcg_target_op_def(const TCGOp *op) case INDEX_op_ctpop_i64: return C_O1_I1(r, r); - case INDEX_op_extract2_i32: - case INDEX_op_extract2_i64: + case INDEX_op_extract2: return C_O1_I2(r, 0, r); case INDEX_op_deposit: -- 2.43.0