lib/Target/X86/README-UNIMPLEMENTED.txt

   1 //===---------------------------------------------------------------------===//
   2 // Testcases that crash the X86 backend because they aren't implemented
   3 //===---------------------------------------------------------------------===//
   4
   5 These are cases we know the X86 backend doesn't handle.  Patches are welcome
   6 and appreciated, because no one has signed up to implemented these yet.
   7 Implementing these would allow elimination of the corresponding intrinsics,
   8 which would be great.
   9
  10 1) vector shifts
  11 2) vector comparisons
  12 3) vector fp<->int conversions: PR2683, PR2684, PR2685, PR2686, PR2688
  13 4) bitcasts from vectors to scalars: PR2804
  14 5) llvm.atomic.cmp.swap.i128.p0i128: PR3462