Add support for llvm.fshl.i32

[weaversa]

or llvm.fshl.i64, or any llvm.fshl.i# really...

Symbolic execution failed.
Abort due to false assumption:
  Pointer is not a function pointer:
(2, 0x0:[64])
  in XXH_rotl32 at xxhash/xxhash32-ref.c:60:36
Stack frame
  Allocations:
    
  Writes:
    
Base memory
  Allocations:
    GlobalAlloc 7 0x0:[64] Immutable 1-byte-aligned XXH_rotl32
    GlobalAlloc 6 0x0:[64] Immutable 1-byte-aligned XXH_read32
    GlobalAlloc 5 0x0:[64] Immutable 1-byte-aligned XXH32_round
    GlobalAlloc 4 0x0:[64] Immutable 1-byte-aligned XXH32_avalanche
    GlobalAlloc 3 0x0:[64] Immutable 1-byte-aligned XXH32
    GlobalAlloc 2 0x0:[64] Immutable 1-byte-aligned llvm.fshl.i32
    GlobalAlloc 1 0x0:[64] Immutable 1-byte-aligned llvm.dbg.value
  Writes:

[weaversa]

Here's the intrinsic page: https://llvm.org/docs/LangRef.html#llvm-fshl-intrinsic

It turns out that if you implement the rotate function using the (shift-left | shift-right) method, llvm with -O1 (at least) just changes it into rotate.

[robdockins]

Interesting. I haven't bumped into funnel shifts before. Looks like it should be pretty easy to implement.

[robdockins]

This should be fixed once we propagate GaloisInc/crucible@e1cb3b5 into SAWScript.

[robdockins]

I believe this is fixed now.