LGTM. Thanks for your time!

Addressed the comments of Jinsong and Hiroshi, will submit VSELECT related exploitation patch once this gets approved.
The irrelevant test case updates have been done by https://reviews.llvm.org/rL349325 and https://reviews.llvm.org/rL349329.

Addressed Jinsong's comments.

Hi Jinsong @jsji, thanks for your time! I'll break down it to some separate patches and add more comments later.

gentle ping.... this is a patch for Power9 setb exploitation

gentle ping...

gentle ping...

Addressed some comments from Nemanja.

In D53383#1276421, @stefanp wrote:

@jedilyn
You are correct that PIC is the only model on AIX. However, the object model (COFF) on AIX is different so the idea of PIC / no-PIC does not mean the same thing it does on Linux anyway. Also AIX is currently not supported for LLVM. If we decide to support it we can look at the options then. However, since AIX only uses one object model and there is no PIC/ no-PIC difference we will probably just end up ignoring the option anyway.

Thanks for clarification Stefan, it's a good idea just to ignore the PIC option on aix when it gets supported in future.

In D53383#1274877, @joerg wrote:

PPC as architecture is strongly geared towards PIC code"

Really some truth, as I know, AIX 64bit mode only supports PIC.

Thanks a lot for your time @nemanjai , such a good idea to use script to verify various combination. I'm sorry I can't actually get your point about simplifying the logic in the switch block. Each case check in the switch seems not too much and exactly follow the patterns in above comments nearing the case, people can easily map the checks to patterns there. Do I miss something there? Could you kindly clarify more on that? Thanks in advance!

In D50004#1185697, @nemanjai wrote:

+1, even for Power9 XSCPSGNDP makes pipeline busy longer than XXLOR, XXLOR is still better.

Can you clarify this please? Where is this information coming from? According to the UM, XXLOR takes up a whole superslice whereas XSCPSGNDP takes up a single slice so we can dispatch 2 of the former per cycle and 4 of the latter. And the "Pipe Busy Cycles" field for both is 1.

In D50004#1182972, @inouehrs wrote:

XSCPSGNDP has longer latency (6 cycles) than XXLOR (2 cycles) on POWER8 while it has higher throughput with the same latency on POWER9. So XXLOR is preferable for pre-P9.

Also, the two instructions have different behavior for a denormal input value in my understanding; XSCPSGNDP does normalization but XXLOR does not. Does this difference matter?

In D48813#1149078, @hfinkel wrote:

LGTM