[libc-commits] [libc] [libc][math][c23] Add expf16 C23 math function (PR #100632)

[via libc-commits]

overmighty wrote:

Before, on i7-13700H:

- With Clang 18, without `-march=native`:  
  ```
   Performance tests with inputs in denormal range:
  -- My function --
       Total time      : 1626076534 ns 
       Average runtime : 79.4759 ns/op 
       Ops per second  : 12582433 op/s 
  -- Other function --
       Total time      : 21810517 ns 
       Average runtime : 1.06601 ns/op 
       Ops per second  : 938079551 op/s 
  -- Average runtime ratio --
       Mine / Other's  : 74.5547 
  
   Performance tests with inputs in normal range:
  -- My function --
       Total time      : 23652101150 ns 
       Average runtime : 38.4975 ns/op 
       Ops per second  : 25975704 op/s 
  -- Other function --
       Total time      : 648820903 ns 
       Average runtime : 1.05606 ns/op 
       Ops per second  : 946917704 op/s 
  -- Average runtime ratio --
       Mine / Other's  : 36.454 
  ```

- With Clang 18, with `-march=native`:  
  ```
   Performance tests with inputs in denormal range:
  -- My function --
       Total time      : 329939627 ns 
       Average runtime : 16.1261 ns/op 
       Ops per second  : 62011344 op/s 
  -- Other function --
       Total time      : 21790717 ns 
       Average runtime : 1.06504 ns/op 
       Ops per second  : 938931931 op/s 
  -- Average runtime ratio --
       Mine / Other's  : 15.1413 

   Performance tests with inputs in normal range:
  -- My function --
       Total time      : 22716554242 ns 
       Average runtime : 36.9748 ns/op 
       Ops per second  : 27045475 op/s 
  -- Other function --
       Total time      : 648758548 ns 
       Average runtime : 1.05596 ns/op 
       Ops per second  : 947008716 op/s 
  -- Average runtime ratio --
       Mine / Other's  : 35.0154 
  ```

After:

- With Clang 18, without `-march=native`:  
  ```
   Performance tests with inputs in denormal range:
  -- My function --
       Total time      : 1515840736 ns 
       Average runtime : 74.088 ns/op 
       Ops per second  : 13497460 op/s 
  -- Other function --
       Total time      : 27118777 ns 
       Average runtime : 1.32545 ns/op 
       Ops per second  : 754458801 op/s 
  -- Average runtime ratio --
       Mine / Other's  : 55.8964 
  
   Performance tests with inputs in normal range:
  -- My function --
       Total time      : 23072321763 ns 
       Average runtime : 37.5538 ns/op 
       Ops per second  : 26628442 op/s 
  -- Other function --
       Total time      : 810717705 ns 
       Average runtime : 1.31957 ns/op 
       Ops per second  : 757822354 op/s 
  -- Average runtime ratio --
       Mine / Other's  : 28.4591 
  ```

- With Clang 18, with `-march=native`:  
  ```
   Performance tests with inputs in denormal range:
  -- My function --
       Total time      : 224244813 ns 
       Average runtime : 10.9602 ns/op 
       Ops per second  : 91239568 op/s 
  -- Other function --
       Total time      : 21774833 ns 
       Average runtime : 1.06426 ns/op 
       Ops per second  : 939616850 op/s 
  -- Average runtime ratio --
       Mine / Other's  : 10.2983 
  
   Performance tests with inputs in normal range:
  -- My function --
       Total time      : 21790749065 ns 
       Average runtime : 35.4679 ns/op 
       Ops per second  : 28194533 op/s 
  -- Other function --
       Total time      : 648719445 ns 
       Average runtime : 1.05589 ns/op 
       Ops per second  : 947065799 op/s 
  -- Average runtime ratio --
       Mine / Other's  : 33.5904 
  ```


https://github.com/llvm/llvm-project/pull/100632