The previous implementation has two problems:
1. At the liveness analysis phase, the liveIn and liveOut computation
is incorrect. The liveIn is not a static information it should be computed
along with the liveOut during the backward data flow analysis.
2. At the register allocation phase, it onl
Hi List,
Although I submitted a patch recently to set the barrier function to
noduplicate attribute, but it seems that it still has problem.
Considering the following type of code:
Label0:
call barrier;
Label1:
.
br i1 %cmp0, label %label2, %label0
label2:
.
.
b
I had the same problem when i was building the gen-backend. Make sure you
are using clang 3.3 when you compile. In my case, even though i had the
path to LLVM 3.3 setup, an older version of clang was also in my path, and
cmake was associating that as the C/C++ compiler.
-Rashid
On Sat, Dec 14, 2