Re: [Qemu-devel] [PATCH v2] RISC-V: Fix riscv_isa_string, use popcount to count bits

[Peter Maydell]

On 10 March 2018 at 21:25, Philippe Mathieu-Daudé <address@hidden> wrote:
> On 03/09/2018 10:01 PM, Michael Clark wrote:
>> Logic bug caused the string size calculation for the RISC-V
>> format ISA string to be small. This fix allows slack for rv128.
>>
>> Cc: Palmer Dabbelt <address@hidden>
>> Cc: Peter Maydell <address@hidden>
>> Cc: Eric Blake <address@hidden>
>> Signed-off-by: Michael Clark <address@hidden>
>> ---
>>  target/riscv/cpu.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
>> index 4851890..1456535 100644
>> --- a/target/riscv/cpu.c
>> +++ b/target/riscv/cpu.c
>> @@ -391,7 +391,7 @@ static const TypeInfo riscv_cpu_type_info = {
>>  char *riscv_isa_string(RISCVCPU *cpu)
>>  {
>>      int i;
>> -    size_t maxlen = 5 + ctz32(cpu->env.misa);
>> +    size_t maxlen = 8 + ctpop64(cpu->env.misa);
>
> Can you document the magic 5/8?
>
> This looks nice, but this seems to me too much optimization to save few
> bytes, using sizeof(riscv_exts) is overflow-free.
>
> Maybe this is enough and self-explanatory:
>
>        const size_t maxlen = sizeof("rv128") + sizeof(riscv_exts);
>
>>      char *isa_string = g_new0(char, maxlen);
>>      snprintf(isa_string, maxlen, "rv%d", TARGET_LONG_BITS);
>
> Also, if you keep the snprintf() return value, you can (naming it 'n')
> simplify (also easier to review):
>
>>      for (i = 0; i < sizeof(riscv_exts); i++) {
>>
>         if (cpu->env.misa & RV(riscv_exts[i])) {
> -           isa_string[strlen(isa_string)] = riscv_exts[i] - 'A' + 'a';
> +           isa_string[n++] = tolower(riscv_exts[i]);
>         }
>     }
>
> and simply use g_new() with:
>
> +   isa_string[n] = '\0';
>
>     return isa_string;
> }

Hi -- any chance of a respin of this patch that addresses Philippe's
review comments so we can fix it before rc0? This is causing
my merge-build tests to fail about 50% of the time on OpenBSD
at the moment...

thanks
-- PMM