diff options
author | Chinmay Rath <rathc@linux.ibm.com> | 2024-04-23 12:02:34 +0530 |
---|---|---|
committer | Nicholas Piggin <npiggin@gmail.com> | 2024-05-24 08:57:50 +1000 |
commit | 948e257c48cf8e387e12b81f4b96195d52b68455 (patch) | |
tree | 1e417223912cc656de2b6218bac386e7766ea4a7 /target/ppc/translate/fixedpoint-impl.c.inc | |
parent | ae556c6a49d47208e4cbc70efab3dfd5bb2ac309 (diff) |
target/ppc: Move logical fixed-point instructions to decodetree.
Moving the below instructions to decodetree specification :
andi[s]., {ori, xori}[s] : D-form
{and, andc, nand, or, orc, nor, xor, eqv}[.],
exts{b, h, w}[.], cnt{l, t}z{w, d}[.],
popcnt{b, w, d}, prty{w, d}, cmp, bpermd : X-form
With this patch, all the fixed-point logical instructions have been
moved to decodetree.
The changes were verified by validating that the tcg ops generated by those
instructions remain the same, which were captured with the '-d in_asm,op' flag.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Chinmay Rath <rathc@linux.ibm.com>
[np: 32-bit compile fix]
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Diffstat (limited to 'target/ppc/translate/fixedpoint-impl.c.inc')
-rw-r--r-- | target/ppc/translate/fixedpoint-impl.c.inc | 300 |
1 files changed, 300 insertions, 0 deletions
diff --git a/target/ppc/translate/fixedpoint-impl.c.inc b/target/ppc/translate/fixedpoint-impl.c.inc index 872fed664d..fa0191e866 100644 --- a/target/ppc/translate/fixedpoint-impl.c.inc +++ b/target/ppc/translate/fixedpoint-impl.c.inc @@ -856,6 +856,285 @@ TRANS(SETBCR, do_set_bool_cond, false, true) TRANS(SETNBC, do_set_bool_cond, true, false) TRANS(SETNBCR, do_set_bool_cond, true, true) +/* + * Fixed-Point Logical Instructions + */ + +static bool do_addi_(DisasContext *ctx, arg_D_ui *a, bool shift) +{ + tcg_gen_andi_tl(cpu_gpr[a->ra], cpu_gpr[a->rt], shift ? a->ui << 16 : a->ui); + gen_set_Rc0(ctx, cpu_gpr[a->ra]); + return true; +} + +static bool do_ori(DisasContext *ctx, arg_D_ui *a, bool shift) +{ + if (a->rt == a->ra && a->ui == 0) { + /* NOP */ + return true; + } + tcg_gen_ori_tl(cpu_gpr[a->ra], cpu_gpr[a->rt], shift ? a->ui << 16 : a->ui); + return true; +} + +static bool do_xori(DisasContext *ctx, arg_D_ui *a, bool shift) +{ + if (a->rt == a->ra && a->ui == 0) { + /* NOP */ + return true; + } + tcg_gen_xori_tl(cpu_gpr[a->ra], cpu_gpr[a->rt], shift ? a->ui << 16 : a->ui); + return true; +} + +static bool do_logical1(DisasContext *ctx, arg_X_sa_rc *a, + void (*helper)(TCGv, TCGv)) +{ + helper(cpu_gpr[a->ra], cpu_gpr[a->rs]); + if (unlikely(a->rc)) { + gen_set_Rc0(ctx, cpu_gpr[a->ra]); + } + return true; +} + +static bool do_logical2(DisasContext *ctx, arg_X_rc *a, + void (*helper)(TCGv, TCGv, TCGv)) +{ + helper(cpu_gpr[a->ra], cpu_gpr[a->rt], cpu_gpr[a->rb]); + if (unlikely(a->rc)) { + gen_set_Rc0(ctx, cpu_gpr[a->ra]); + } + return true; +} + +static bool trans_OR(DisasContext *ctx, arg_OR *a) +{ + /* Optimisation for mr. ri case */ + if (a->rt != a->ra || a->rt != a->rb) { + if (a->rt != a->rb) { + tcg_gen_or_tl(cpu_gpr[a->ra], cpu_gpr[a->rt], cpu_gpr[a->rb]); + } else { + tcg_gen_mov_tl(cpu_gpr[a->ra], cpu_gpr[a->rt]); + } + if (unlikely(a->rc)) { + gen_set_Rc0(ctx, cpu_gpr[a->ra]); + } + } else if (unlikely(a->rc)) { + gen_set_Rc0(ctx, cpu_gpr[a->rt]); +#if defined(TARGET_PPC64) + } else if (a->rt != 0) { /* 0 is nop */ + int prio = 0; + + switch (a->rt) { + case 1: + /* Set process priority to low */ + prio = 2; + break; + case 6: + /* Set process priority to medium-low */ + prio = 3; + break; + case 2: + /* Set process priority to normal */ + prio = 4; + break; +#if !defined(CONFIG_USER_ONLY) + case 31: + if (!ctx->pr) { + /* Set process priority to very low */ + prio = 1; + } + break; + case 5: + if (!ctx->pr) { + /* Set process priority to medium-hight */ + prio = 5; + } + break; + case 3: + if (!ctx->pr) { + /* Set process priority to high */ + prio = 6; + } + break; + case 7: + if (ctx->hv && !ctx->pr) { + /* Set process priority to very high */ + prio = 7; + } + break; +#endif + default: + break; + } + if (prio) { + TCGv t0 = tcg_temp_new(); + gen_load_spr(t0, SPR_PPR); + tcg_gen_andi_tl(t0, t0, ~0x001C000000000000ULL); + tcg_gen_ori_tl(t0, t0, ((uint64_t)prio) << 50); + gen_store_spr(SPR_PPR, t0); + } +#if !defined(CONFIG_USER_ONLY) + /* + * Pause out of TCG otherwise spin loops with smt_low eat too + * much CPU and the kernel hangs. This applies to all + * encodings other than no-op, e.g., miso(rs=26), yield(27), + * mdoio(29), mdoom(30), and all currently undefined. + */ + gen_pause(ctx); +#endif +#endif + } + + return true; +} + +static bool trans_XOR(DisasContext *ctx, arg_XOR *a) +{ + /* Optimisation for "set to zero" case */ + if (a->rt != a->rb) { + tcg_gen_xor_tl(cpu_gpr[a->ra], cpu_gpr[a->rt], cpu_gpr[a->rb]); + } else { + tcg_gen_movi_tl(cpu_gpr[a->ra], 0); + } + if (unlikely(a->rc)) { + gen_set_Rc0(ctx, cpu_gpr[a->ra]); + } + return true; +} + +static bool trans_CMPB(DisasContext *ctx, arg_CMPB *a) +{ + REQUIRE_INSNS_FLAGS2(ctx, ISA205); + gen_helper_CMPB(cpu_gpr[a->ra], cpu_gpr[a->rt], cpu_gpr[a->rb]); + return true; +} + +static bool do_cntzw(DisasContext *ctx, arg_X_sa_rc *a, + void (*helper)(TCGv_i32, TCGv_i32, uint32_t)) +{ + TCGv_i32 t = tcg_temp_new_i32(); + + tcg_gen_trunc_tl_i32(t, cpu_gpr[a->rs]); + helper(t, t, 32); + tcg_gen_extu_i32_tl(cpu_gpr[a->ra], t); + + if (unlikely(a->rc)) { + gen_set_Rc0(ctx, cpu_gpr[a->ra]); + } + return true; +} + +#if defined(TARGET_PPC64) +static bool do_cntzd(DisasContext *ctx, arg_X_sa_rc *a, + void (*helper)(TCGv_i64, TCGv_i64, uint64_t)) +{ + helper(cpu_gpr[a->ra], cpu_gpr[a->rs], 64); + if (unlikely(a->rc)) { + gen_set_Rc0(ctx, cpu_gpr[a->ra]); + } + return true; +} +#endif + +static bool trans_CNTLZD(DisasContext *ctx, arg_CNTLZD *a) +{ + REQUIRE_64BIT(ctx); +#if defined(TARGET_PPC64) + do_cntzd(ctx, a, tcg_gen_clzi_i64); +#else + qemu_build_not_reached(); +#endif + return true; +} + +static bool trans_CNTTZD(DisasContext *ctx, arg_CNTTZD *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_INSNS_FLAGS2(ctx, ISA300); +#if defined(TARGET_PPC64) + do_cntzd(ctx, a, tcg_gen_ctzi_i64); +#else + qemu_build_not_reached(); +#endif + return true; +} + +static bool trans_POPCNTB(DisasContext *ctx, arg_POPCNTB *a) +{ + REQUIRE_INSNS_FLAGS(ctx, POPCNTB); + gen_helper_POPCNTB(cpu_gpr[a->ra], cpu_gpr[a->rs]); + return true; +} + +static bool trans_POPCNTW(DisasContext *ctx, arg_POPCNTW *a) +{ + REQUIRE_INSNS_FLAGS(ctx, POPCNTWD); +#if defined(TARGET_PPC64) + gen_helper_POPCNTW(cpu_gpr[a->ra], cpu_gpr[a->rs]); +#else + tcg_gen_ctpop_i32(cpu_gpr[a->ra], cpu_gpr[a->rs]); +#endif + return true; +} + +static bool trans_POPCNTD(DisasContext *ctx, arg_POPCNTD *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_INSNS_FLAGS(ctx, POPCNTWD); +#if defined(TARGET_PPC64) + tcg_gen_ctpop_i64(cpu_gpr[a->ra], cpu_gpr[a->rs]); +#else + qemu_build_not_reached(); +#endif + return true; +} + +static bool trans_PRTYW(DisasContext *ctx, arg_PRTYW *a) +{ + TCGv ra = cpu_gpr[a->ra]; + TCGv rs = cpu_gpr[a->rs]; + TCGv t0 = tcg_temp_new(); + + REQUIRE_INSNS_FLAGS2(ctx, ISA205); + tcg_gen_shri_tl(t0, rs, 16); + tcg_gen_xor_tl(ra, rs, t0); + tcg_gen_shri_tl(t0, ra, 8); + tcg_gen_xor_tl(ra, ra, t0); + tcg_gen_andi_tl(ra, ra, (target_ulong)0x100000001ULL); + return true; +} + +static bool trans_PRTYD(DisasContext *ctx, arg_PRTYD *a) +{ + TCGv ra = cpu_gpr[a->ra]; + TCGv rs = cpu_gpr[a->rs]; + TCGv t0 = tcg_temp_new(); + + REQUIRE_64BIT(ctx); + REQUIRE_INSNS_FLAGS2(ctx, ISA205); + tcg_gen_shri_tl(t0, rs, 32); + tcg_gen_xor_tl(ra, rs, t0); + tcg_gen_shri_tl(t0, ra, 16); + tcg_gen_xor_tl(ra, ra, t0); + tcg_gen_shri_tl(t0, ra, 8); + tcg_gen_xor_tl(ra, ra, t0); + tcg_gen_andi_tl(ra, ra, 1); + return true; +} + +static bool trans_BPERMD(DisasContext *ctx, arg_BPERMD *a) +{ + REQUIRE_64BIT(ctx); + REQUIRE_INSNS_FLAGS2(ctx, PERM_ISA206); +#if defined(TARGET_PPC64) + gen_helper_BPERMD(cpu_gpr[a->ra], cpu_gpr[a->rt], cpu_gpr[a->rb]); +#else + qemu_build_not_reached(); +#endif + return true; +} + static bool trans_CFUGED(DisasContext *ctx, arg_X *a) { REQUIRE_64BIT(ctx); @@ -944,6 +1223,27 @@ static bool trans_PEXTD(DisasContext *ctx, arg_X *a) return true; } +TRANS(ANDI_, do_addi_, false); +TRANS(ANDIS_, do_addi_, true); +TRANS(ORI, do_ori, false); +TRANS(ORIS, do_ori, true); +TRANS(XORI, do_xori, false); +TRANS(XORIS, do_xori, true); + +TRANS(AND, do_logical2, tcg_gen_and_tl); +TRANS(ANDC, do_logical2, tcg_gen_andc_tl); +TRANS(NAND, do_logical2, tcg_gen_nand_tl); +TRANS(ORC, do_logical2, tcg_gen_orc_tl); +TRANS(NOR, do_logical2, tcg_gen_nor_tl); +TRANS(EQV, do_logical2, tcg_gen_eqv_tl); +TRANS(EXTSB, do_logical1, tcg_gen_ext8s_tl); +TRANS(EXTSH, do_logical1, tcg_gen_ext16s_tl); + +TRANS(CNTLZW, do_cntzw, tcg_gen_clzi_i32); +TRANS_FLAGS2(ISA300, CNTTZW, do_cntzw, tcg_gen_ctzi_i32); + +TRANS64(EXTSW, do_logical1, tcg_gen_ext32s_tl); + static bool trans_ADDG6S(DisasContext *ctx, arg_X *a) { const target_ulong carry_bits = (target_ulong)-1 / 0xf; |