Add LoadNonVolatileOperation::NormalizeLoadMemoryStateJoin() reduction

[phate]

No description provided.

[phate]

This seem to have created an endless loop.

[haved]

Looks good!
EDIT: Oh wait. Yeah I can see that it loops forever...

[phate]

@haved Could you do me the favor and fire off a single core(!) run on Idun. We would like to find out:

1. Whether the endless loop is at compile-time or at runtime of a benchmark.
2. Which benchmark/file it is.

Let's hope this gives us enough hints to fix it.

[haved]

@phate already on it

[haved]

@phate on my machine it didn't time out when doing AndersenRegionAware, but the opt one seems to run forever

[phate]

@haved What do you mean with "the opt one seems to run forever"?

[haved]

@phate when I do

./scripts/run-llvm-test-suite.sh --make-target llvm-run-andersen-region-aware

but when I do

./scripts/run-llvm-test-suite.sh --make-target llvm-run-opt

it hangs

[haved]

@phate It is the invocation of jlc that hangs. It seems to happen for 3 of the files:

jlc -DNDEBUG -JInvariantValueRedirection -JNodeReduction -JDeadNodeElimination -JThetaGammaInversion -JInvariantValueRedirection -JDeadNodeElimination -JInvariantValueRedirection -JDeadNodeElimination -JNodeReduction -JCommonNodeElimination -JDeadNodeElimination -JNodePullIn -JInvariantValueRedirection -JDeadNodeElimination -JLoopUnrolling -JInvariantValueRedirection -w -Werror=date-time -D__PRINT_NORM -MD -MT MultiSource/Benchmarks/DOE-ProxyApps-C/miniGMG/CMakeFiles/miniGMG.dir/operators.ompif.c.o -MF CMakeFiles/miniGMG.dir/operators.ompif.c.o.d -o CMakeFiles/miniGMG.dir/operators.ompif.c.o -c /home/havard/Development/PhD/jlm/usr/llvm-test-suite/MultiSource/Benchmarks/DOE-ProxyApps-C/miniGMG/operators.ompif.c

jlc -DNDEBUG -JInvariantValueRedirection -JNodeReduction -JDeadNodeElimination -JThetaGammaInversion -JInvariantValueRedirection -JDeadNodeElimination -JInvariantValueRedirection -JDeadNodeElimination -JNodeReduction -JCommonNodeElimination -JDeadNodeElimination -JNodePullIn -JInvariantValueRedirection -JDeadNodeElimination -JLoopUnrolling -JInvariantValueRedirection -w -Werror=date-time -DHAVEMPGLIB -DLAMEPARSE -DNDEBUG -D__NO_MATH_INLINES -DLAMESNDFILE -MD -MT MultiSource/Benchmarks/MiBench/consumer-lame/CMakeFiles/consumer-lame.dir/layer3.c.o -MF CMakeFiles/consumer-lame.dir/layer3.c.o.d -o CMakeFiles/consumer-lame.dir/layer3.c.o -c /home/havard/Development/PhD/jlm/usr/llvm-test-suite/MultiSource/Benchmarks/MiBench/consumer-lame/layer3.c

jlc -DNDEBUG -JInvariantValueRedirection -JNodeReduction -JDeadNodeElimination -JThetaGammaInversion -JInvariantValueRedirection -JDeadNodeElimination -JInvariantValueRedirection -JDeadNodeElimination -JNodeReduction -JCommonNodeElimination -JDeadNodeElimination -JNodePullIn -JInvariantValueRedirection -JDeadNodeElimination -JLoopUnrolling -JInvariantValueRedirection -w -Werror=date-time -Wno-implicit-int -MD -MT SingleSource/Benchmarks/Misc/CMakeFiles/himenobmtxpa.dir/himenobmtxpa.c.o -MF CMakeFiles/himenobmtxpa.dir/himenobmtxpa.c.o.d -o CMakeFiles/himenobmtxpa.dir/himenobmtxpa.c.o -c /home/havard/Development/PhD/jlm/usr/llvm-test-suite/SingleSource/Benchmarks/Misc/himenobmtxpa.c

[phate]

@haved We are lucky. It is jlm and not a benchmark executable. I will do the rest from here on. Thanks a lot!