hbpintool returning 0.0 J inside docker container

[eafurst]

When running the hbpintool on a compiled version on /graphit/test/input/cc.gt, the tool always returns

HammerBlade Energy Cost: 0.000000e+00 J
Xeon E7-8894 v4 Energy Cost: 0.000000e+00 J

I've added __attribute__((noinline)) and this doesn't seem to fix the issue.
When I've run the hbpintool on this same file outside of the docker container, it does return nonzero estimates.

[mrutt92]

this sounds like it's being caused by #4 ?

[eafurst]

The compiled cc.gt file runs as expected inside the container. I only have gotten the issue from #4 with bfs_benchmark.gt (so far, these are the only two graphit programs I've tried compiling and running).

[mrutt92]

how are you running bfs_benchmark?

[eafurst]

I believe I have how I'm running bfs_benchmark in the body of issue #4.

One thing to note for #4 might be that cc.gt has the default schedule assigned to it and bfs_benchmark.gt has a written schedule.

[eafurst]

I am now getting nonzero values when running the hbpintool with the compiled /graphit/test/input/bfs_with_inputfilename_arg.gt and a modified /graphit/test/input/cc.gt that takes a graph filename as an arg (instead of hardcoding in the graph file). I think the issue might be related to having the input graph filename hardcoded into the program (for whatever reason).

[mrutt92]

Oh. It's probably because the working directory is in /hbpintool/ and that path is almost certainly a relative one.

[eafurst]

Oh sorry, to clarify #4 was caused just by trying to run the compiled program. I hadn't even gotten to trying to run bfs_benchmark with hbpintool at that point. I was running ./bfs.o ../../test/graphs/4.el from the /graphit/build/bin directory.