2008-10-16 08:46:12 +02:00
|
|
|
//===---------------------------------------------------------------------===//
|
|
|
|
|
// Testcases that crash the X86 backend because they aren't implemented
|
|
|
|
|
//===---------------------------------------------------------------------===//
|
|
|
|
|
|
|
|
|
|
These are cases we know the X86 backend doesn't handle. Patches are welcome
|
|
|
|
|
and appreciated, because no one has signed up to implemented these yet.
|
|
|
|
|
Implementing these would allow elimination of the corresponding intrinsics,
|
|
|
|
|
which would be great.
|
|
|
|
|
|
|
|
|
|
1) vector shifts
|
|
|
|
|
2) vector comparisons
|
2008-10-16 19:02:50 +02:00
|
|
|
3) vector fp<->int conversions: PR2683, PR2684, PR2685, PR2686, PR2688
|
2008-10-16 09:04:06 +02:00
|
|
|
4) bitcasts from vectors to scalars: PR2804
|
2010-02-09 06:55:14 +01:00
|
|
|
5) llvm.atomic.cmp.swap.i128.p0i128: PR3462
|
