Compilation w/ LLVM 3.0, status of trace generators

[Damien]

Dear all,

I'm working on a CUDA/PTX backend for a DSL, and I'd like to use
Ocelot as a source of performance metrics and other statistics to
drive the optimisation process. I have a couple of questions:

1. I compiled the current SVN version with LLVM 3.0 (gcc4.5.3,
cuda4.1, ubuntu11.10) and I had to change a number of occurrences of
'error.print' to 'error.Print' for it to compile properly.
- Tests run by .'/build.py -t full' pass except TestPTXAssembly which
fails.
- gpuocelot/tests/cuda4.1sdk/Reduction fails with:
GPU result = 43749422
CPU result = 2139353471
Is this expected, or is my build probably broken?

2. What is the status of the trace generators in gpuocelot/trace-
generators? I'm especially interested in PerformanceBoundGenerator,
LoadBalanceGenerator and MemoryTraceGenerator. Currently I need to
comment out the following when running scons in the trace-generators
directory:
# "traces/implementation/LoadBalanceGenerator.cpp",
# "traces/implementation/BasicBlockCountGenerator.cpp",
# "traces/implementation/TemporalSIMTGenerator.cpp"
These fail to compile with apparently non-environment related errors
(eg "error: no match for 'operator!='")

3. I saw mentions of something called IPDOM in the slides and in the
code. Could you point me to what that is?

Thanks & cheers,
Damien