On 12/21/19 4:25 AM, Alex Bennée wrote:
>> -#define CPU_MMU_INDEX (cpu_mmu_index(env, true))
>> -#define MEMSUFFIX _code
>> -#define SOFTMMU_CODE_ACCESS
>> +uint32_t cpu_ldub_code(CPUArchState *env, abi_ptr addr);
>> +uint32_t cpu_lduw_code(CPUArchState *env, abi_ptr addr);
>> +uint32_t cpu_ldl_code(CPUArchState *env, abi_ptr addr);
>> +uint64_t cpu_ldq_code(CPUArchState *env, abi_ptr addr);
>>
>> -#define DATA_SIZE 1
>> -#include "exec/cpu_ldst_template.h"
>> +static inline int cpu_ldsb_code(CPUArchState *env, abi_ptr addr)
>> +{
>> + return (int8_t)cpu_ldub_code(env, addr);
>> +}
...
>> -uint8_t helper_ret_ldub_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -int8_t helper_ret_ldsb_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -uint16_t helper_le_lduw_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -int16_t helper_le_ldsw_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -uint32_t helper_le_ldl_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -uint64_t helper_le_ldq_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -uint16_t helper_be_lduw_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -int16_t helper_be_ldsw_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -uint32_t helper_be_ldl_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -uint64_t helper_be_ldq_cmmu(CPUArchState *env, target_ulong addr,
>> - TCGMemOpIdx oi, uintptr_t retaddr);
>> -
> <snip>
>
> Why have we dropped the final type and stuck to uint32/64? Do we rely on
> the caller truncating any data by casting the result of their eventual
> cpu_ldFoo?
I have *preserved* the existing type of cpu_ld*_code(), the actual interface
that is being used.
Cleaning up the load return and store value types is another project. I would
not limit that to *_code, though. And I'd want to think of some way to prove
that all callers are getting the promotion that they expected.
r~
