极术小姐姐 · 2020年02月03日

The "usage model" of ARMv8 SVE contiguous "non-fault"

Hello,

What exactly is the "usage model" of ARMv8 SVE contiguous "non-fault" load instructions ?

I understand the usage model of "first-fault" SVE instrcutions (which is described in many white papers) but the "usage model" of ARMv8 SVE contiguous "non-fault" load instructions is not clear to me.

What I am looking for is the example of why this instruction is useful (i.e. why it has been introduced in the SVE instruction set).

For instance, for the "first-fault" load SVE instructions the usage model is the implementation of the loops with the "dynamic" exit criteria detection (as it is explained in the multiple ARM white papers on SVE)

I am looking for the similar explanation and/or example of the "usage model" for the SVE "non-fault" load instructions

0 评论

1 个回答得票排序 · 时间排序

棋子 · 2020年02月03日

Thanks for the reply. However, what I am looking for is not the "definition" of the instruction but rather the "usage scenario i.e. the explanation "why" the instruction has been introduced. In other words, what kind of the algorithms it is useful for.
For instance,there are multiple available “white papers” which provide the examples of the “usage scenarios” for SVE “first -faulting” Load instructions (LDFF) (for instance, the vectorized strlen() implementation). However, it appears, that they do not provide the “usage scenarios” for SVE “non-faulting” Load instructions (LDNF).

I have updated the question to clarify what information I am looking for

0 评论已采纳