Displaying 3 results from an estimated 3 matches for "d8943".
Did you mean:
18943
2015 Nov 19
5
[RFC] Introducing a vector reduction add instruction.
...1, %xmm0 # xmm0 = xmm1[2,3,0,1]
paddd %xmm1, %xmm0
pshufd $229, %xmm0, %xmm1 # xmm1 = xmm0[1,1,2,3]
paddd %xmm0, %xmm1
movd %xmm1, %eax
retq
Note that due to smaller VF we are using now (currently 4), we could
not explore the most benefit of psadbw. The patch in
http://reviews.llvm.org/D8943 has enables us to use bigger VFs based
on the smallest type in a loop. The follow-up work is refining the
cost model to let bigger VFs have less cost. For the example above the
best result is from VF >=16.
The draft of the patch is here: http://reviews.llvm.org/D14840
I will refine the patch l...
2015 Nov 25
2
[RFC] Introducing a vector reduction add instruction.
...%xmm0, %xmm1 # xmm1 = xmm0[1,1,2,3]
>> paddd %xmm0, %xmm1
>> movd %xmm1, %eax
>> retq
>>
>>
>> Note that due to smaller VF we are using now (currently 4), we could
>> not explore the most benefit of psadbw. The patch in
>> http://reviews.llvm.org/D8943 has enables us to use bigger VFs based
>> on the smallest type in a loop. The follow-up work is refining the
>> cost model to let bigger VFs have less cost. For the example above
>> the
>> best result is from VF >=16.
>>
>> The draft of the patch is here: http...
2015 Nov 25
2
[RFC] Introducing a vector reduction add instruction.
...addd %xmm0, %xmm1
> >> movd %xmm1, %eax
> >> retq
> >>
> >>
> >> Note that due to smaller VF we are using now (currently 4), we
> >> could
> >> not explore the most benefit of psadbw. The patch in
> >> http://reviews.llvm.org/D8943 has enables us to use bigger VFs
> >> based
> >> on the smallest type in a loop. The follow-up work is refining the
> >> cost model to let bigger VFs have less cost. For the example above
> >> the
> >> best result is from VF >=16.
> >>
>...