[PATCH] D56082: [X86][SLP] Enable SLP vectorization for 128-bit horizontal X86 instructions (add, sub)

Wed Jan 16 11:17:35 PST 2019

anton-afanasyev added a comment.

In D56082#1360129 <https://reviews.llvm.org/D56082#1360129>, @craig.topper wrote:

> There are a lot of test changes here that have nothing to do with the add/sub operations mentioned in the title. How do we know these changes are good? Was any benchmarking done with this patch?

Yes, I have done a lot of benchmarking using test suite while issue investigation.
I cannot attach results right now (working on other issue), can do it later.

The most suspicios was `size..text` metrics of `dijkstra` test (https://gcc.godbolt.org/z/3E1AmW), but I have investigated that is not bug, but rightful work of Loop Unrolling Pass.

Is `size..text` and `exec_time` enough? Can you advise any other benchmarking?

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D56082/new/

https://reviews.llvm.org/D56082