Re: [PATCH v1 1/4] exfat: refactor nls.c (tables)

[David Timber <dxdt@dev.snart.me>]

On 4/30/26 18:22, David Timber wrote:
> On 4/30/26 16:53, Yuezhang.Mo@sony.com wrote:
>>> +/* exfat/tables.c */
>>> +/* Upcase table macro */
>>> +#define EXFAT_NUM_UPCASE     (2918)
>>> +#define EXFAT_UTBL_COUNT     (0x10000)
>>> +
>>> +extern const unsigned short exfat_uni_def_upcase[EXFAT_NUM_UPCASE];
>>> +extern const unsigned short exfat_bad_uni_chars[];
>> The global variable exfat_bad_uni_chars[] is unnecessary, as is the 
>> generic function exfat_wstrchr().
>>
>> We can implement a dedicated function `exfat_is_bad_uni_char()`, in
>> which we can define static `exfat_bad_uni_chars`, or we can use a
>> switch...case... statement.
> I agree.
>
> I too think that switch a case statement is suitable here because
> there's only a few __u16's to compare so there's really no need to leave
> room for cache misses. Let the CPU prefetch the characters along with
> the instructions.
>
> Will make a separate commit if no objection.
Did some fiddling and found out that that could risk performance
regression. But it's a problem specific to Clang, which is not so
popular choice of compiler in the embedded world where it could have a
significant impact.

If it doesn't bother anyone, I'll send out the reroll.

https://github.com/llvm/llvm-project/issues/127365#issuecomment-4362781091