On Mon, Jul 24, 2017 at 8:28 PM, Balbir Singh <bsinghar...@gmail.com> wrote:
> On Mon, Jul 24, 2017 at 11:01 AM, Matt Brown
> <matthew.brown....@gmail.com> wrote:
>> This adds emulations for the popcntb, popcntw, and popcntd instructions.
>> Tested for correctness against the popcnt{b,w,d} instructions on ppc64le.
>>
>> Signed-off-by: Matt Brown <matthew.brown....@gmail.com>
>> ---
>> v2:
>> - fixed opcodes
>> - fixed typecasting
>> - fixed bitshifting error for both 32 and 64bit arch
>> ---
>> arch/powerpc/lib/sstep.c | 43 ++++++++++++++++++++++++++++++++++++++++++-
>> 1 file changed, 42 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/powerpc/lib/sstep.c b/arch/powerpc/lib/sstep.c
>> index 87d277f..e6a16a3 100644
>> --- a/arch/powerpc/lib/sstep.c
>> +++ b/arch/powerpc/lib/sstep.c
>> @@ -612,6 +612,35 @@ static nokprobe_inline void do_cmpb(struct pt_regs
>> *regs, unsigned long v1,
>> regs->gpr[rd] = out_val;
>> }
>>
>> +/*
>> + * The size parameter is used to adjust the equivalent popcnt instruction.
>> + * popcntb = 8, popcntw = 32, popcntd = 64
>> + */
>> +static nokprobe_inline void do_popcnt(struct pt_regs *regs, unsigned long
>> v1,
>> + int size, int ra)
>> +{
>> + unsigned long long high, low, mask;
>> + unsigned int n;
>> + int i, j;
>> +
>> + high = 0;
>> + low = 0;
>> +
>> + for (i = 0; i < (64 / size); i++) {
>> + n = 0;
>> + for (j = 0; j < size; j++) {
>> + mask = 1UL << (j + (i * size));
>> + if (v1 & mask)
>> + n++;
>> + }
>> + if ((i * size) < 32)
>> + low |= n << (i * size);
>> + else
>> + high |= n << ((i * size) - 32);
>> + }
>> + regs->gpr[ra] = (high << 32) | low;
>> +}
>
> There's a way to do it in very efficient way via the Giles-Miller
> method of side-ways addition
>
> Please see
>
> http://opensourceforu.com/2012/06/power-programming-bitwise-tips-tricks/
> and lib/hweight.c, you can reuse the code from lib/hweight.c
Oh that's a really cool technique.
We could use that for the parity instructions too.
>
> Balbir Singh
