- // excess values at the end (such as condition codes etc) just lop them off.
- // This doesn't need to worry about flags or chains, just explicit results.
- //
- // FIXME2: This doesn't work because there is currently no way to get an
- // accurate count of the # results the source pattern sets. This is because
- // of the "parallel" construct in X86 land, which looks like this:
- //
- //def : Pat<(parallel (X86and_flag GR8:$src1, GR8:$src2),
- // (implicit EFLAGS)),
- // (AND8rr GR8:$src1, GR8:$src2)>;
- //
- // This idiom means to match the two-result node X86and_flag (which is
- // declared as returning a single result, because we can't match multi-result
- // nodes yet). In this case, we would have to know that the input has two
- // results. However, mul8r is modelled exactly the same way, but without
- // implicit defs included. The fix is to support multiple results directly
- // and eliminate 'parallel'.
+ // excess values at the end (such as implicit defs of condition codes etc)
+ // just lop them off. This doesn't need to worry about glue or chains, just
+ // explicit results.