Re: [PATCH v2 09/42] include/exec: Inline *_data memory operations

[Richard Henderson <richard.henderson@linaro.org>]

On 4/1/25 01:24, Philippe Mathieu-Daudé wrote:
> On 18/3/25 22:31, Richard Henderson wrote:
>> These need to be per-target for 'abi_ptr'.  Expand inline to
>> the *_data_ra api with ra == 0.
>>
>> Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
>> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
>> ---
>>   include/exec/cpu_ldst.h     | 123 ++++++++++++++++++++++++++++++------
>>   accel/tcg/ldst_common.c.inc |  89 --------------------------
>>   2 files changed, 104 insertions(+), 108 deletions(-)
>>
>> diff --git a/include/exec/cpu_ldst.h b/include/exec/cpu_ldst.h
>> index a2a90c7554..d084da0b5f 100644
>> --- a/include/exec/cpu_ldst.h
>> +++ b/include/exec/cpu_ldst.h
>> @@ -74,25 +74,6 @@
>>   #include "user/guest-host.h"
>>   #endif /* CONFIG_USER_ONLY */
>> -uint32_t cpu_ldub_data(CPUArchState *env, abi_ptr ptr);
>> -int cpu_ldsb_data(CPUArchState *env, abi_ptr ptr);
>> -uint32_t cpu_lduw_be_data(CPUArchState *env, abi_ptr ptr);
>> -int cpu_ldsw_be_data(CPUArchState *env, abi_ptr ptr);
>> -uint32_t cpu_ldl_be_data(CPUArchState *env, abi_ptr ptr);
>> -uint64_t cpu_ldq_be_data(CPUArchState *env, abi_ptr ptr);
>> -uint32_t cpu_lduw_le_data(CPUArchState *env, abi_ptr ptr);
>> -int cpu_ldsw_le_data(CPUArchState *env, abi_ptr ptr);
>> -uint32_t cpu_ldl_le_data(CPUArchState *env, abi_ptr ptr);
>> -uint64_t cpu_ldq_le_data(CPUArchState *env, abi_ptr ptr);
>> -
>> -void cpu_stb_data(CPUArchState *env, abi_ptr ptr, uint32_t val);
>> -void cpu_stw_be_data(CPUArchState *env, abi_ptr ptr, uint32_t val);
>> -void cpu_stl_be_data(CPUArchState *env, abi_ptr ptr, uint32_t val);
>> -void cpu_stq_be_data(CPUArchState *env, abi_ptr ptr, uint64_t val);
>> -void cpu_stw_le_data(CPUArchState *env, abi_ptr ptr, uint32_t val);
>> -void cpu_stl_le_data(CPUArchState *env, abi_ptr ptr, uint32_t val);
>> -void cpu_stq_le_data(CPUArchState *env, abi_ptr ptr, uint64_t val);
>> -
>>   static inline uint32_t
>>   cpu_ldub_mmuidx_ra(CPUArchState *env, abi_ptr addr, int mmu_idx, uintptr_t ra)
>>   {
>> @@ -342,6 +323,110 @@ cpu_stq_le_data_ra(CPUArchState *env, abi_ptr addr, uint64_t val, 
>> uintptr_t ra)
>>       cpu_stq_le_mmuidx_ra(env, addr, val, mmu_index, ra);
>>   }
>> +/*--------------------------*/
>> +
>> +static inline uint32_t
> 
> Pre-existing, but why don't we use 'unsigned int' like for the signed
> variants, or the plain int8_t type?

It's the original type used by the api.

You'd never change this.  If you want to audit all of the uses vs the api change, you'd 
update to one of the newer function groups.


r~