From: Chinmay Rath <rathc@linux.ibm.com>
To: qemu-devel@nongnu.org, qemu-ppc@nongnu.org,
harshpb@linux.ibm.com, milesg@linux.ibm.com, npiggin@gmail.com,
richard.henderson@linaro.org,
Amit Machhiwal <amachhiw@linux.ibm.com>
Subject: Re: [PATCH v2 03/37] target/ppc: Convert cache instructions to decodetree
Date: Thu, 27 Aug 2026 14:07:07 +0530 [thread overview]
Message-ID: <65de3d79-da56-422b-8449-0b505a9eec3a@linux.ibm.com> (raw)
In-Reply-To: <20260826153132.6c0e83d2-05-amachhiw@linux.ibm.com>
On 8/26/26 15:50, Amit Machhiwal wrote:
> On 2026/08/26 10:38 AM, Chinmay Rath wrote:
>> From: Nikhil Kumar Singh <nikhilks@linux.ibm.com>
>>
>> Convert cache management instructions to decodetree using custom
>> X-form layouts (@X_ea, @X_l, @X_th) and dedicated translation functions.
>> This ensures reserved bits are accurately represented and ignored by
>> the decoder (using '-' in the patterns), rather than being erroneously
>> parsed into instruction fields.
>>
>> Legacy GEN_HANDLER-based implementations are removed.
>>
>> The implementation preserves legacy semantics, including:
>> - MMU-visible accesses for instructions treated as loads
>> (e.g. dcbt, dcbtst, dcbtep, dcbtstep)
>> - Supervisor and BookE-specific constraints
>> - Effective address computation via do_ea_calc()
>>
>> Instructions that were defined as no-ops in the legacy implementation
>> remain no-ops here.
>>
>> Note on intentional behavior changes:
>> - dcbf now validates the L field against the Power ISA v3.1 set of
>> legal values (0, 1, 3, 4, 6).
>> - Reserved bits that were previously encoded as part of the opcode
>> mask (and trapped if set) are now properly decoded as "don't care"
>> ('-' in the pattern).
>>
>> Testing:
>> - Verified TCG equivalence for all cache operations
>>
>> Signed-off-by: Nikhil Kumar Singh <nikhilks@linux.ibm.com>
>> Tested-by: Tasmiya Nalatwad <tasmiya@linux.ibm.com>
>> Reviewed-by: Glenn Miles <milesg@linux.ibm.com>
>> Signed-off-by: Chinmay Rath <rathc@linux.ibm.com>
>> ---
>> target/ppc/helper.h | 8 +-
>> target/ppc/insn32.decode | 35 +++
>> target/ppc/mem_helper.c | 8 +-
>> target/ppc/translate.c | 478 ++++++++++++++++++++-------------------
>> 4 files changed, 292 insertions(+), 237 deletions(-)
>>
>> diff --git a/target/ppc/helper.h b/target/ppc/helper.h
>> index 6b2d19a3ad..46ad78d1d4 100644
>> --- a/target/ppc/helper.h
>> +++ b/target/ppc/helper.h
>> @@ -48,12 +48,12 @@ DEF_HELPER_FLAGS_3(stmw, TCG_CALL_NO_WG, void, env, tl, i32)
>> DEF_HELPER_4(lsw, void, env, tl, i32, i32)
>> DEF_HELPER_5(lswx, void, env, tl, i32, i32, i32)
>> DEF_HELPER_FLAGS_4(stsw, TCG_CALL_NO_WG, void, env, tl, i32, i32)
>> -DEF_HELPER_FLAGS_3(dcbz, TCG_CALL_NO_WG, void, env, tl, int)
>> +DEF_HELPER_FLAGS_3(DCBZ, TCG_CALL_NO_WG, void, env, tl, int)
>> #ifdef TARGET_PPC64
>> -DEF_HELPER_FLAGS_2(dcbzl, TCG_CALL_NO_WG, void, env, tl)
>> +DEF_HELPER_FLAGS_2(DCBZL, TCG_CALL_NO_WG, void, env, tl)
>> #endif
>> -DEF_HELPER_FLAGS_2(icbi, TCG_CALL_NO_WG, void, env, tl)
>> -DEF_HELPER_FLAGS_2(icbiep, TCG_CALL_NO_WG, void, env, tl)
>> +DEF_HELPER_FLAGS_2(ICBI, TCG_CALL_NO_WG, void, env, tl)
>> +DEF_HELPER_FLAGS_2(ICBIEP, TCG_CALL_NO_WG, void, env, tl)
>>
>> #if defined(TARGET_PPC64)
>> DEF_HELPER_4(DIVDEU, i64, env, i64, i64, i32)
>> diff --git a/target/ppc/insn32.decode b/target/ppc/insn32.decode
>> index 26948e08a7..6b1cf9a0a9 100644
>> --- a/target/ppc/insn32.decode
>> +++ b/target/ppc/insn32.decode
>> @@ -1306,6 +1306,41 @@ XVF64GERPN 111011 ... -- .... 0 ..... 10111010 ..- @XX3_at xa=%xx_xa_pair
>> XVF64GERNP 111011 ... -- .... 0 ..... 01111010 ..- @XX3_at xa=%xx_xa_pair
>> XVF64GERNN 111011 ... -- .... 0 ..... 11111010 ..- @XX3_at xa=%xx_xa_pair
>>
>> +## Cache Management Instructions
>> +
>> +&X_ea ra rb
>> +@X_ea ...... ..... ra:5 rb:5 .......... . &X_ea
>> +
>> +&X_l l ra rb
>> +@X_l ...... .. l:3 ra:5 rb:5 .......... . &X_l
>> +
>> +&X_th th ra rb
>> +@X_th ...... th:5 ra:5 rb:5 .......... . &X_th
>> +
>> +DCBF 011111 -- ... ..... ..... 0001010110 - @X_l
>> +DCBFEP 011111 ----- ..... ..... 0001111111 - @X_ea
>> +DCBI 011111 ----- ..... ..... 0111010110 - @X_ea
>> +DCBST 011111 ----- ..... ..... 0000110110 - @X_ea
>> +DCBSTEP 011111 ----- ..... ..... 0000111111 - @X_ea
>> +DCBA 011111 ----- ..... ..... 1011110110 - @X_ea
>> +
>> +DCBT 011111 ..... ..... ..... 0100010110 - @X_th
>> +DCBTEP 011111 ----- ..... ..... 0100111111 - @X_ea
>> +DCBTST 011111 ..... ..... ..... 0011110110 - @X_th
>> +DCBTSTEP 011111 ----- ..... ..... 0011111111 - @X_ea
>> +
>> +DCBTLS 011111 ..... ..... ..... 0010100110 - @X_th
>> +DCBLC 011111 ..... ..... ..... 0110000110 - @X_th
>> +
>> +DCBZ 011111 ----- ..... ..... 1111110110 - @X_ea
>> +DCBZEP 011111 ----- ..... ..... 1111111111 - @X_ea
>> +ICBI 011111 ----- ..... ..... 1111010110 - @X_ea
>> +ICBIEP 011111 ----- ..... ..... 1111011111 - @X_ea
>> +
>> +DST 011111 ----- ..... ..... 0101010110 - @X_ea
>> +DSTST 011111 ----- ..... ..... 0101110110 - @X_ea
> Per **PowerISA v3.1 §4.3.2.1**, dst, dstst, and dss are obsolete and
> must be treated as no-ops independent of whether Vector instructions are
> available. The new REQUIRE_INSNS_FLAGS(ctx, ALTIVEC) guard in all three
> handlers is incorrect — it causes these opcodes to be treated as illegal
> instructions on non-Altivec CPUs, contradicting the ISA mandate. These
> handlers should unconditionally return true with no flag check, matching
> the ISA's "always a no-op" requirement. The ra == 0 INVAL check in
> trans_DST and trans_DSTST should also be dropped for the same reason —
> if the instruction is a no-op, the operand constraint has no effect.
>
>> +DSS 011111 ----- ..... ..... 1100110110 - @X_ea
>> +
>> ##Extend Sign Word and Shift Left Immediate XS-form
>> EXTSWSLI 011111 ..... ..... ..... 110111101 . . @XS
>>
>> diff --git a/target/ppc/mem_helper.c b/target/ppc/mem_helper.c
>> index 787ba2cdec..95a224e74c 100644
>> --- a/target/ppc/mem_helper.c
>> +++ b/target/ppc/mem_helper.c
>> @@ -319,13 +319,13 @@ static void dcbz_common(CPUPPCState *env, target_ulong addr,
>> clear_helper_retaddr();
>> }
>>
>> -void helper_dcbz(CPUPPCState *env, target_ulong addr, int mmu_idx)
>> +void helper_DCBZ(CPUPPCState *env, target_ulong addr, int mmu_idx)
>> {
>> dcbz_common(env, addr, mmu_idx, env->dcache_line_size, GETPC());
>> }
>>
>> #ifdef TARGET_PPC64
>> -void helper_dcbzl(CPUPPCState *env, target_ulong addr)
>> +void helper_DCBZL(CPUPPCState *env, target_ulong addr)
>> {
>> int dcbz_size = env->dcache_line_size;
>>
>> @@ -341,7 +341,7 @@ void helper_dcbzl(CPUPPCState *env, target_ulong addr)
>> }
>> #endif
>>
>> -void helper_icbi(CPUPPCState *env, target_ulong addr)
>> +void helper_ICBI(CPUPPCState *env, target_ulong addr)
>> {
>> unsigned mmu_idx = cpu_mmu_index(env_cpu(env), false);
>> MemOpIdx oi = make_memop_idx(MO_UL | MO_UNALN, mmu_idx);
>> @@ -357,7 +357,7 @@ void helper_icbi(CPUPPCState *env, target_ulong addr)
>> cpu_ldl_mmu(env, addr, oi, GETPC());
>> }
>>
>> -void helper_icbiep(CPUPPCState *env, target_ulong addr)
>> +void helper_ICBIEP(CPUPPCState *env, target_ulong addr)
>> {
>> #if !defined(CONFIG_USER_ONLY)
>> MemOpIdx oi = make_memop_idx(MO_UL | MO_UNALN, PPC_TLB_EPID_LOAD);
>> diff --git a/target/ppc/translate.c b/target/ppc/translate.c
>> index cc9287dcc5..60c88a6153 100644
>> --- a/target/ppc/translate.c
>> +++ b/target/ppc/translate.c
>> @@ -4296,216 +4296,6 @@ static void gen_setb(DisasContext *ctx)
>> }
>> #endif
>>
>> -/*** Cache management ***/
>> -
>> -/* dcbf */
>> -static void gen_dcbf(DisasContext *ctx)
>> -{
>> - /* XXX: specification says this is treated as a load by the MMU */
>> - TCGv t0;
>> - gen_set_access_type(ctx, ACCESS_CACHE);
>> - t0 = tcg_temp_new();
>> - gen_addr_reg_index(ctx, t0);
>> - gen_qemu_ld8u(ctx, t0, t0);
>> -}
>> -
>> -/* dcbfep (external PID dcbf) */
>> -static void gen_dcbfep(DisasContext *ctx)
>> -{
>> - /* XXX: specification says this is treated as a load by the MMU */
>> - TCGv t0;
>> - CHK_SV(ctx);
>> - gen_set_access_type(ctx, ACCESS_CACHE);
>> - t0 = tcg_temp_new();
>> - gen_addr_reg_index(ctx, t0);
>> - tcg_gen_qemu_ld_tl(t0, t0, PPC_TLB_EPID_LOAD, DEF_MEMOP(MO_UB));
>> -}
>> -
>> -/* dcbi (Supervisor only) */
>> -static void gen_dcbi(DisasContext *ctx)
>> -{
>> -#if defined(CONFIG_USER_ONLY)
>> - GEN_PRIV(ctx);
>> -#else
>> - TCGv EA, val;
>> -
>> - CHK_SV(ctx);
>> - EA = tcg_temp_new();
>> - gen_set_access_type(ctx, ACCESS_CACHE);
>> - gen_addr_reg_index(ctx, EA);
>> - val = tcg_temp_new();
>> - /* XXX: specification says this should be treated as a store by the MMU */
>> - gen_qemu_ld8u(ctx, val, EA);
>> - gen_qemu_st8(ctx, val, EA);
>> -#endif /* defined(CONFIG_USER_ONLY) */
>> -}
>> -
>> -/* dcdst */
>> -static void gen_dcbst(DisasContext *ctx)
>> -{
>> - /* XXX: specification say this is treated as a load by the MMU */
>> - TCGv t0;
>> - gen_set_access_type(ctx, ACCESS_CACHE);
>> - t0 = tcg_temp_new();
>> - gen_addr_reg_index(ctx, t0);
>> - gen_qemu_ld8u(ctx, t0, t0);
>> -}
>> -
>> -/* dcbstep (dcbstep External PID version) */
>> -static void gen_dcbstep(DisasContext *ctx)
>> -{
>> - /* XXX: specification say this is treated as a load by the MMU */
>> - TCGv t0;
>> - gen_set_access_type(ctx, ACCESS_CACHE);
>> - t0 = tcg_temp_new();
>> - gen_addr_reg_index(ctx, t0);
>> - tcg_gen_qemu_ld_tl(t0, t0, PPC_TLB_EPID_LOAD, DEF_MEMOP(MO_UB));
>> -}
>> -
>> -/* dcbt */
>> -static void gen_dcbt(DisasContext *ctx)
>> -{
>> - /*
>> - * interpreted as no-op
>> - * XXX: specification say this is treated as a load by the MMU but
>> - * does not generate any exception
>> - */
>> -}
>> -
>> -/* dcbtep */
>> -static void gen_dcbtep(DisasContext *ctx)
>> -{
>> - /*
>> - * interpreted as no-op
>> - * XXX: specification say this is treated as a load by the MMU but
>> - * does not generate any exception
>> - */
>> -}
>> -
>> -/* dcbtst */
>> -static void gen_dcbtst(DisasContext *ctx)
>> -{
>> - /*
>> - * interpreted as no-op
>> - * XXX: specification say this is treated as a load by the MMU but
>> - * does not generate any exception
>> - */
>> -}
>> -
>> -/* dcbtstep */
>> -static void gen_dcbtstep(DisasContext *ctx)
>> -{
>> - /*
>> - * interpreted as no-op
>> - * XXX: specification say this is treated as a load by the MMU but
>> - * does not generate any exception
>> - */
>> -}
>> -
>> -/* dcbtls */
>> -static void gen_dcbtls(DisasContext *ctx)
>> -{
>> - /* Always fails locking the cache */
>> - TCGv t0 = tcg_temp_new();
>> - gen_load_spr(t0, SPR_Exxx_L1CSR0);
>> - tcg_gen_ori_tl(t0, t0, L1CSR0_CUL);
>> - gen_store_spr(SPR_Exxx_L1CSR0, t0);
>> -}
>> -
>> -/* dcblc */
>> -static void gen_dcblc(DisasContext *ctx)
>> -{
>> - /*
>> - * interpreted as no-op
>> - */
>> -}
>> -
>> -/* dcbz */
>> -static void gen_dcbz(DisasContext *ctx)
>> -{
>> - TCGv tcgv_addr = tcg_temp_new();
>> -
>> - gen_set_access_type(ctx, ACCESS_CACHE);
>> - gen_addr_reg_index(ctx, tcgv_addr);
>> -
>> -#ifdef TARGET_PPC64
>> - if (ctx->excp_model == POWERPC_EXCP_970 && !(ctx->opcode & 0x00200000)) {
>> - gen_helper_dcbzl(tcg_env, tcgv_addr);
>> - return;
>> - }
>> -#endif
>> -
>> - gen_helper_dcbz(tcg_env, tcgv_addr, tcg_constant_i32(ctx->mem_idx));
>> -}
>> -
>> -/* dcbzep */
>> -static void gen_dcbzep(DisasContext *ctx)
>> -{
>> - TCGv tcgv_addr = tcg_temp_new();
>> -
>> - gen_set_access_type(ctx, ACCESS_CACHE);
>> - gen_addr_reg_index(ctx, tcgv_addr);
>> - gen_helper_dcbz(tcg_env, tcgv_addr, tcg_constant_i32(PPC_TLB_EPID_STORE));
>> -}
>> -
>> -/* dst / dstt */
>> -static void gen_dst(DisasContext *ctx)
>> -{
>> - if (rA(ctx->opcode) == 0) {
>> - gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);
>> - } else {
>> - /* interpreted as no-op */
>> - }
>> -}
>> -
>> -/* dstst /dststt */
>> -static void gen_dstst(DisasContext *ctx)
>> -{
>> - if (rA(ctx->opcode) == 0) {
>> - gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);
>> - } else {
>> - /* interpreted as no-op */
>> - }
>> -
>> -}
>> -
>> -/* dss / dssall */
>> -static void gen_dss(DisasContext *ctx)
>> -{
>> - /* interpreted as no-op */
>> -}
>> -
>> -/* icbi */
>> -static void gen_icbi(DisasContext *ctx)
>> -{
>> - TCGv t0;
>> - gen_set_access_type(ctx, ACCESS_CACHE);
>> - t0 = tcg_temp_new();
>> - gen_addr_reg_index(ctx, t0);
>> - gen_helper_icbi(tcg_env, t0);
>> -}
>> -
>> -/* icbiep */
>> -static void gen_icbiep(DisasContext *ctx)
>> -{
>> - TCGv t0;
>> - gen_set_access_type(ctx, ACCESS_CACHE);
>> - t0 = tcg_temp_new();
>> - gen_addr_reg_index(ctx, t0);
>> - gen_helper_icbiep(tcg_env, t0);
>> -}
>> -
>> -/* Optional: */
>> -/* dcba */
>> -static void gen_dcba(DisasContext *ctx)
>> -{
>> - /*
>> - * interpreted as no-op
>> - * XXX: specification say this is treated as a store by the MMU
>> - * but does not generate any exception
>> - */
>> -}
>> -
>> /*** Segment register manipulation ***/
>> /* Supervisor only: */
>>
>> @@ -5746,6 +5536,255 @@ static bool trans_LQARX(DisasContext *ctx, arg_LQARX *a)
>> return true;
>> }
>>
>> +/*
>> + * Cache Management Instructions (decodetree)
>> + */
>> +
>> +static bool trans_DCBA(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + REQUIRE_INSNS_FLAGS(ctx, CACHE_DCBA);
>> + return true;
>> +}
>> +
>> +static bool trans_DCBT(DisasContext *ctx, arg_X_th *a)
>> +{
>> + REQUIRE_INSNS_FLAGS(ctx, CACHE);
>> + return true;
>> +}
>> +
>> +static bool trans_DCBTEP(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + REQUIRE_INSNS_FLAGS2(ctx, BOOKE206);
>> + return true;
>> +}
>> +
>> +static bool trans_DCBTST(DisasContext *ctx, arg_X_th *a)
>> +{
>> + REQUIRE_INSNS_FLAGS(ctx, CACHE);
>> + return true;
>> +}
>> +
>> +static bool trans_DCBTSTEP(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + REQUIRE_INSNS_FLAGS2(ctx, BOOKE206);
>> + return true;
>> +}
>> +
>> +static bool trans_DCBLC(DisasContext *ctx, arg_X_th *a)
>> +{
>> + /* Requires either PPC_BOOKE or PPC2_BOOKE206 */
>> + if (!(ctx->insns_flags & PPC_BOOKE) &&
>> + !(ctx->insns_flags2 & PPC2_BOOKE206)) {
>> + return false;
>> + }
>> + return true;
>> +}
>> +
>> +static bool trans_DSS(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + REQUIRE_INSNS_FLAGS(ctx, ALTIVEC);
>> + return true;
>> +}
>> +
>> +static bool trans_DCBTLS(DisasContext *ctx, arg_X_th *a)
>> +{
>> + TCGv t0 = tcg_temp_new();
>> +
>> + /* Requires either PPC_BOOKE or PPC2_BOOKE206 */
>> + if (!(ctx->insns_flags & PPC_BOOKE) &&
>> + !(ctx->insns_flags2 & PPC2_BOOKE206)) {
>> + return false;
>> + }
>> +
>> + gen_load_spr(t0, SPR_Exxx_L1CSR0);
>> + tcg_gen_ori_tl(t0, t0, L1CSR0_CUL);
>> + gen_store_spr(SPR_Exxx_L1CSR0, t0);
>> +
>> + return true;
>> +}
>> +
>> +static bool trans_DST(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + REQUIRE_INSNS_FLAGS(ctx, ALTIVEC);
>> +
>> + if (a->ra == 0) {
>> + gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);
>> + }
>> +
>> + return true;
>> +}
>> +
>> +static bool trans_DSTST(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + REQUIRE_INSNS_FLAGS(ctx, ALTIVEC);
>> +
>> + if (a->ra == 0) {
>> + gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);
>> + }
>> +
>> + return true;
>> +}
>> +
>> +static bool trans_DCBF(DisasContext *ctx, arg_X_l *a)
>> +{
>> + TCGv EA;
>> +
>> + REQUIRE_INSNS_FLAGS(ctx, CACHE);
>> +
>> + /*
>> + * As per PowerISA v3.1, the L field can have values 0, 1, 3, 4, or 6.
>> + * Other values are Undefined Behavior (UB).
>> + */
>> + switch (a->l) {
>> + case 0: /* dcbf */
>> + case 1: /* dcbfl */
>> + case 3: /* dcbflp */
>> + case 4: /* dcbfps */
>> + case 6: /* dcbstps */
>> + break;
>> + default:
>> + gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);
>> + return true;
>> + }
>> +
>> + gen_set_access_type(ctx, ACCESS_CACHE);
>> + EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
>> + gen_qemu_ld8u(ctx, EA, EA);
>> +
>> + return true;
>> +}
>> +
>> +static bool trans_DCBST(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + TCGv EA;
>> +
>> + REQUIRE_INSNS_FLAGS(ctx, CACHE);
>> +
>> + gen_set_access_type(ctx, ACCESS_CACHE);
>> + EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
>> + gen_qemu_ld8u(ctx, EA, EA);
>> +
>> + return true;
>> +}
>> +
>> +static bool trans_DCBFEP(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + TCGv EA;
>> +
>> + REQUIRE_INSNS_FLAGS2(ctx, BOOKE206);
>> +
>> + REQUIRE_SV(ctx);
>> +
>> + gen_set_access_type(ctx, ACCESS_CACHE);
>> + EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
>> +
>> + tcg_gen_qemu_ld_tl(EA, EA, PPC_TLB_EPID_LOAD, DEF_MEMOP(MO_UB));
>> +
>> + return true;
>> +}
>> +
>> +static bool trans_DCBSTEP(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + TCGv EA;
>> +
>> + REQUIRE_INSNS_FLAGS2(ctx, BOOKE206);
>> +
>> + gen_set_access_type(ctx, ACCESS_CACHE);
>> + EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
>> +
>> + tcg_gen_qemu_ld_tl(EA, EA, PPC_TLB_EPID_LOAD, DEF_MEMOP(MO_UB));
>> +
>> + return true;
>> +}
>> +
>> +static bool trans_DCBI(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + REQUIRE_INSNS_FLAGS(ctx, CACHE);
>> +
>> +#if defined(CONFIG_USER_ONLY)
>> + gen_priv_opc(ctx);
>> + return true;
> The #if defined(CONFIG_USER_ONLY) block looks redundant to me.
> REQUIRE_SV is already defined for both build configurations — in
> user-only builds it unconditionally calls gen_priv_opc() and returns,
> and in system builds it checks ctx->pr. The manual #if split just
> duplicates what REQUIRE_SV already does. The function can be simplified
> to:
>
> static bool trans_DCBI(DisasContext *ctx, arg_X_ea *a)
> {
> REQUIRE_INSNS_FLAGS(ctx, CACHE);
> REQUIRE_SV(ctx);
>
> gen_set_access_type(ctx, ACCESS_CACHE);
> TCGv EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
> TCGv val = tcg_temp_new();
>
> gen_qemu_ld8u(ctx, val, EA);
> gen_qemu_st8(ctx, val, EA);
>
> return true;
> }
>
>> +#else
>> + TCGv EA, val;
>> +
>> + REQUIRE_SV(ctx);
>> +
>> + gen_set_access_type(ctx, ACCESS_CACHE);
>> + EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
>> + val = tcg_temp_new();
>> +
>> + gen_qemu_ld8u(ctx, val, EA);
>> + gen_qemu_st8(ctx, val, EA);
>> +
>> + return true;
>> +#endif
>> +}
>> +
>> +static bool trans_ICBI(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + TCGv EA;
>> +
>> + REQUIRE_INSNS_FLAGS(ctx, CACHE_ICBI);
>> +
>> + gen_set_access_type(ctx, ACCESS_CACHE);
>> +
>> + EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
>> + gen_helper_ICBI(tcg_env, EA);
>> +
>> + return true;
>> +}
>> +
>> +static bool trans_ICBIEP(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + TCGv EA;
>> +
>> + REQUIRE_INSNS_FLAGS2(ctx, BOOKE206);
>> + REQUIRE_SV(ctx);
>> +
>> + gen_set_access_type(ctx, ACCESS_CACHE);
>> + EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
>> +
>> + gen_helper_ICBIEP(tcg_env, EA);
>> + return true;
>> +}
>> +
>> +static bool trans_DCBZ(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + TCGv EA;
>> +
>> + REQUIRE_INSNS_FLAGS(ctx, CACHE_DCBZ);
>> +
>> + gen_set_access_type(ctx, ACCESS_CACHE);
>> +
>> + EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
>> +
>> +#if defined(TARGET_PPC64)
>> + if (ctx->excp_model == POWERPC_EXCP_970 &&
>> + !(ctx->opcode & 0x00200000)) {
Sorry I missed this comment.
> trans_DCBZ reads ctx->opcode directly to distinguish dcbz from dcbzl on
> the 970.
>
> In a decodetree translator ctx->opcode is not reliable — the instruction
I agree, in decodetree translation, it is not reliable to use
ctx->opcode, though it is technically usable and populated.
> fields are decoded into arg_* before the translator is called. Bit 21
> (0x00200000) is the LSB of the RT field: dcbz encodes RT=0 and the
> 970-specific dcbzl encodes RT=1. Since the current decode pattern
>
> 011111----- ..... ..... 1111110110
>
> treats all of RT as don't-care, both variants land in trans_DCBZ with
> the distinguishing bit lost.
>
> I think a better way would to split the decode entry so the table itself
> separates the two:
>
> DCBZ 011111 00000 ..... ..... 1111110110 - @X_ea
> DCBZL 011111 00001 ..... ..... 1111110110 - @X_ea
I agree with the idea of splitting the two insns. However, DCBZ should
have '-' for reserved bits staying true to the ISA and DCBZL could use
it's own format. However I am unable to find any official doc/ISA for
DCBZL yet. Hence will drop this insn from the patch in the interest of
time. Will post a followup patch for this.
Thanks,
Chinmay
>
> Then trans_DCBZ needs no ctx->opcode access, and trans_DCBZL handles the
> 970 path cleanly with its own REQUIRE_64BIT + excp_model check.
>
>> + gen_helper_DCBZL(tcg_env, EA);
>> + return true;
>> + }
>> +#endif
>> +
>> + gen_helper_DCBZ(tcg_env, EA, tcg_constant_i32(ctx->mem_idx));
>> + return true;
>> +}
>> +
>> +static bool trans_DCBZEP(DisasContext *ctx, arg_X_ea *a)
>> +{
>> + TCGv EA;
>> +
>> + REQUIRE_INSNS_FLAGS2(ctx, BOOKE206);
>> +
>> + REQUIRE_SV(ctx);
>> +
>> + gen_set_access_type(ctx, ACCESS_CACHE);
>> + EA = do_ea_calc(ctx, a->ra, cpu_gpr[a->rb]);
>> +
>> + gen_helper_DCBZ(tcg_env, EA, tcg_constant_i32(PPC_TLB_EPID_STORE));
>> + return true;
>> +}
>> +
>> #include "translate/fixedpoint-impl.c.inc"
>>
>> #include "translate/fp-impl.c.inc"
>> @@ -5910,25 +5949,6 @@ GEN_HANDLER_E(mcrxrx, 0x1F, 0x00, 0x12, 0x007FF801, PPC_NONE, PPC2_ISA300),
>> #endif
>> GEN_HANDLER(mtmsr, 0x1F, 0x12, 0x04, 0x001EF801, PPC_MISC),
>> GEN_HANDLER(mtspr, 0x1F, 0x13, 0x0E, 0x00000000, PPC_MISC),
>> -GEN_HANDLER(dcbf, 0x1F, 0x16, 0x02, 0x03C00001, PPC_CACHE),
>> -GEN_HANDLER_E(dcbfep, 0x1F, 0x1F, 0x03, 0x03C00001, PPC_NONE, PPC2_BOOKE206),
>> -GEN_HANDLER(dcbi, 0x1F, 0x16, 0x0E, 0x03E00001, PPC_CACHE),
>> -GEN_HANDLER(dcbst, 0x1F, 0x16, 0x01, 0x03E00001, PPC_CACHE),
>> -GEN_HANDLER_E(dcbstep, 0x1F, 0x1F, 0x01, 0x03E00001, PPC_NONE, PPC2_BOOKE206),
>> -GEN_HANDLER(dcbt, 0x1F, 0x16, 0x08, 0x00000001, PPC_CACHE),
>> -GEN_HANDLER_E(dcbtep, 0x1F, 0x1F, 0x09, 0x00000001, PPC_NONE, PPC2_BOOKE206),
>> -GEN_HANDLER(dcbtst, 0x1F, 0x16, 0x07, 0x00000001, PPC_CACHE),
>> -GEN_HANDLER_E(dcbtstep, 0x1F, 0x1F, 0x07, 0x00000001, PPC_NONE, PPC2_BOOKE206),
>> -GEN_HANDLER_E(dcbtls, 0x1F, 0x06, 0x05, 0x02000001, PPC_BOOKE, PPC2_BOOKE206),
>> -GEN_HANDLER_E(dcblc, 0x1F, 0x06, 0x0c, 0x02000001, PPC_BOOKE, PPC2_BOOKE206),
>> -GEN_HANDLER(dcbz, 0x1F, 0x16, 0x1F, 0x03C00001, PPC_CACHE_DCBZ),
>> -GEN_HANDLER_E(dcbzep, 0x1F, 0x1F, 0x1F, 0x03C00001, PPC_NONE, PPC2_BOOKE206),
>> -GEN_HANDLER(dst, 0x1F, 0x16, 0x0A, 0x01800001, PPC_ALTIVEC),
>> -GEN_HANDLER(dstst, 0x1F, 0x16, 0x0B, 0x01800001, PPC_ALTIVEC),
>> -GEN_HANDLER(dss, 0x1F, 0x16, 0x19, 0x019FF801, PPC_ALTIVEC),
>> -GEN_HANDLER(icbi, 0x1F, 0x16, 0x1E, 0x03E00001, PPC_CACHE_ICBI),
>> -GEN_HANDLER_E(icbiep, 0x1F, 0x1F, 0x1E, 0x03E00001, PPC_NONE, PPC2_BOOKE206),
>> -GEN_HANDLER(dcba, 0x1F, 0x16, 0x17, 0x03E00001, PPC_CACHE_DCBA),
>> GEN_HANDLER(mfsr, 0x1F, 0x13, 0x12, 0x0010F801, PPC_SEGMENT),
>> GEN_HANDLER(mfsrin, 0x1F, 0x13, 0x14, 0x001F0001, PPC_SEGMENT),
>> GEN_HANDLER(mtsr, 0x1F, 0x12, 0x06, 0x0010F801, PPC_SEGMENT),
>> --
>> 2.55.0
>>
next prev parent reply other threads:[~2026-08-27 8:38 UTC|newest]
Thread overview: 83+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-08-26 5:08 [PATCH v2 00/37] target/ppc: PPC TCG Improvements (decodetree migrations + ISA 2.07 flag updates) Chinmay Rath
2026-08-26 5:08 ` [PATCH v2 01/37] target/ppc: Migrate extswsli to decodetree Chinmay Rath
2026-08-26 7:33 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 02/37] target/ppc: Migrate atomic loads " Chinmay Rath
2026-08-26 9:20 ` Amit Machhiwal
2026-08-26 12:28 ` Chinmay Rath
2026-08-26 12:30 ` Chinmay Rath
2026-08-26 5:08 ` [PATCH v2 03/37] target/ppc: Convert cache instructions " Chinmay Rath
2026-08-26 10:20 ` Amit Machhiwal
2026-08-26 13:09 ` Chinmay Rath
2026-08-26 13:17 ` Amit Machhiwal
2026-08-27 9:13 ` Chinmay Rath
2026-08-27 8:37 ` Chinmay Rath [this message]
2026-08-26 5:08 ` [PATCH v2 04/37] target/ppc: Move vector merge " Chinmay Rath
2026-08-26 10:57 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 05/37] target/ppc: Move vector pack " Chinmay Rath
2026-08-26 11:04 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 06/37] target/ppc: Move st{b, h, w, d, q}cx " Chinmay Rath
2026-08-26 11:18 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 07/37] target/ppc: convert slw, srw instruction via decode spec Chinmay Rath
2026-08-26 11:28 ` Amit Machhiwal
2026-08-26 11:30 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 08/37] target/ppc: convert sraw[i] " Chinmay Rath
2026-08-26 11:41 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 09/37] target/ppc: Convert mcrf to decode tree Chinmay Rath
2026-08-26 11:52 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 10/37] target/ppc: Move fixed-point Shift insns to decodetree Chinmay Rath
2026-08-26 5:08 ` [PATCH v2 11/37] target/ppc: Move fixed-point byte-reversal store " Chinmay Rath
2026-08-26 5:08 ` [PATCH v2 12/37] target/ppc: Move GPR atomic load/store instructions " Chinmay Rath
2026-08-26 12:05 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 13/37] target/ppc: Move isync instruction " Chinmay Rath
2026-08-26 12:20 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 14/37] target/ppc: Convert b{a, l, la} to decode tree Chinmay Rath
2026-08-26 12:24 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 15/37] target/ppc: move various conditional branch insns to decodetree Chinmay Rath
2026-08-26 12:38 ` Amit Machhiwal
2026-08-26 12:45 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 16/37] target/ppc: Fix TRANS* macro variadic arguments handling Chinmay Rath
2026-08-26 12:50 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 17/37] target/ppc: Move wait instruction to decodetree Chinmay Rath
2026-08-26 13:12 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 18/37] target/ppc: Move sleep & friends " Chinmay Rath
2026-08-26 13:26 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 19/37] target/ppc: Refactor sleep and its variants to use a common helper Chinmay Rath
2026-08-26 13:38 ` Amit Machhiwal
2026-08-26 14:30 ` Miles Glenn
2026-08-26 5:08 ` [PATCH v2 20/37] target/ppc: Move Condition Register access instructions to decodetree Chinmay Rath
2026-08-26 14:24 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 21/37] target/ppc: Move Condition Register logical " Chinmay Rath
2026-08-26 14:43 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 22/37] target/ppc: make do_ea_calc_ra available for 32 bit builds Chinmay Rath
2026-08-26 14:48 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 23/37] target/ppc: Move Fixed-Point Load/Store String instructions to decodetree Chinmay Rath
2026-08-26 15:19 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 24/37] target/ppc: Move VMX integer arithmetic and BCD " Chinmay Rath
2026-08-26 15:30 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 25/37] target/ppc: Move rlwimi, rlwinm " Chinmay Rath
2026-08-26 15:41 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 26/37] target/ppc: Move lmw, stmw " Chinmay Rath
2026-08-26 15:44 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 27/37] target/ppc: Move mfmsr, mtmsr[d] " Chinmay Rath
2026-08-26 15:49 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 28/37] target/ppc: Move byte-reverse " Chinmay Rath
2026-08-26 15:54 ` Amit Machhiwal
2026-08-27 10:52 ` Chinmay Rath
2026-08-26 5:08 ` [PATCH v2 29/37] target/ppc: Move system call and rfi " Chinmay Rath
2026-08-26 16:54 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 30/37] target/ppc: Replace PPC2_VSX207 flag with PPC2_ISA207 Chinmay Rath
2026-08-26 16:56 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 31/37] target/ppc: Use PPC2_ISA207 instead of PPC2_BCTAR_ISA207 Chinmay Rath
2026-08-26 16:58 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 32/37] target/ppc: Use PPC2_ISA207 instead of PPC2_LSQ_ISA207 Chinmay Rath
2026-08-26 17:01 ` Amit Machhiwal
2026-08-26 5:08 ` [PATCH v2 33/37] target/ppc: Use PPC2_ISA207 instead of PPC2_ALTIVEC_207 Chinmay Rath
2026-08-26 17:23 ` Amit Machhiwal
2026-08-26 5:09 ` [PATCH v2 34/37] target/ppc: Use PPC2_ISA207 instead of PPC2_ISA207S Chinmay Rath
2026-08-26 5:09 ` [PATCH v2 35/37] target/ppc: Reorder PPC2 flags Chinmay Rath
2026-08-26 5:09 ` [PATCH v2 36/37] target/ppc: Add ICBT support for ISA version 2.07 Chinmay Rath
2026-08-26 5:09 ` [PATCH v2 37/37] target/ppc: Add self as maintainer for PowerPC TCG CPUs Chinmay Rath
2026-08-26 5:41 ` Amit Machhiwal
2026-08-26 6:25 ` Gautam Menghani
2026-08-26 9:37 ` Aditya Gupta
2026-08-27 9:50 ` [PATCH v2 00/37] target/ppc: PPC TCG Improvements (decodetree migrations + ISA 2.07 flag updates) Aniket Sahu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=65de3d79-da56-422b-8449-0b505a9eec3a@linux.ibm.com \
--to=rathc@linux.ibm.com \
--cc=amachhiw@linux.ibm.com \
--cc=harshpb@linux.ibm.com \
--cc=milesg@linux.ibm.com \
--cc=npiggin@gmail.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-ppc@nongnu.org \
--cc=richard.henderson@linaro.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.