Saw this just in time. If this true with gcc as well or does it include both clang and gcc?bob wrote:If you download the new Xcode 7.3 update you might end up with a problem. It will compile Crafty normally, but a profile-guided optimization crashes and burns during the first compile stage. I have submitted all the info to Apple. As of right now, on my MacBook, I can no longer do a PGO compile..
Just clang. gcc is an independent compiler, which I would assume you have installed via macports?MikeB wrote:Saw this just in time. If this true with gcc as well or does it include both clang and gcc?bob wrote:If you download the new Xcode 7.3 update you might end up with a problem. It will compile Crafty normally, but a profile-guided optimization crashes and burns during the first compile stage. I have submitted all the info to Apple. As of right now, on my MacBook, I can no longer do a PGO compile..
Correct.. 4.8bob wrote:Just clang. gcc is an independent compiler, which I would assume you have installed via macports?MikeB wrote:Saw this just in time. If this true with gcc as well or does it include both clang and gcc?bob wrote:If you download the new Xcode 7.3 update you might end up with a problem. It will compile Crafty normally, but a profile-guided optimization crashes and burns during the first compile stage. I have submitted all the info to Apple. As of right now, on my MacBook, I can no longer do a PGO compile..
For info, I upgraded XCode last night and tested PGO (native clang, not gcc). Works ok here.bob wrote:If you download the new Xcode 7.3 update you might end up with a problem. It will compile Crafty normally, but a profile-guided optimization crashes and burns during the first compile stage. I have submitted all the info to Apple. As of right now, on my MacBook, I can no longer do a PGO compile..
Code: Select all
floyd-pgo1: $(wildcard Source/*) Makefile versions.json
$(CC) $(CFLAGS) -DNDEBUG -o $@ $(uciSources) $(LDFLAGS) -fprofile-instr-generate
echo 'bench movetime 500 bestof 1' | ./$@ | grep result
floyd-pgo2: $(wildcard Source/*) Makefile versions.json floyd-pgo1
xcrun llvm-profdata show default.profraw
xcrun llvm-profdata merge -output=code.profdata default.profraw
$(CC) $(CFLAGS) -DNDEBUG -o $@ $(uciSources) $(LDFLAGS) -fprofile-instr-use=code.profdataLOL that is ridiculous ..sje wrote:I don't use Xcode's PGO, and it's a wonder that it ever did work as Apple has little interest in optimization.
If Apple did have an interest in optimization, then it wouldn't have included gigabytes of crap in its Xcode distribution for its watches, phones, TV boxes, tablets, etc. I was watching the files being installed over the hour or so it took to update; included were things like different sized icons for Seattle area public transit stations.
I now have about a half-dozen codes that cause the compiler itself to segfault, yet they compile just fine on Xcode 7.2. I have narrowed the problems down by compiling each source file of Crafty separately. Xcode 7.3 crashes when compiling either movgen.c or evaluate.c when using the profiler.mvk wrote:For info, I upgraded XCode last night and tested PGO (native clang, not gcc). Works ok here.bob wrote:If you download the new Xcode 7.3 update you might end up with a problem. It will compile Crafty normally, but a profile-guided optimization crashes and burns during the first compile stage. I have submitted all the info to Apple. As of right now, on my MacBook, I can no longer do a PGO compile..
Code: Select all
floyd-pgo1: $(wildcard Source/*) Makefile versions.json $(CC) $(CFLAGS) -DNDEBUG -o $@ $(uciSources) $(LDFLAGS) -fprofile-instr-generate echo 'bench movetime 500 bestof 1' | ./$@ | grep result floyd-pgo2: $(wildcard Source/*) Makefile versions.json floyd-pgo1 xcrun llvm-profdata show default.profraw xcrun llvm-profdata merge -output=code.profdata default.profraw $(CC) $(CFLAGS) -DNDEBUG -o $@ $(uciSources) $(LDFLAGS) -fprofile-instr-use=code.profdata
I think I am giving up on Xcode and removing it. gcc works perfectly for me, produces faster/cleaner code, and doesn't try to support everything from frying pans to flyswatters...MikeB wrote:LOL that is ridiculous ..sje wrote:I don't use Xcode's PGO, and it's a wonder that it ever did work as Apple has little interest in optimization.
If Apple did have an interest in optimization, then it wouldn't have included gigabytes of crap in its Xcode distribution for its watches, phones, TV boxes, tablets, etc. I was watching the files being installed over the hour or so it took to update; included were things like different sized icons for Seattle area public transit stations.
I've installed gcc 5.3 and it seems to work really well speed-wise. Significantly faster executable than that which clang produced, so I am back in business. Xcode is dead and buried here...bob wrote:I think I am giving up on Xcode and removing it. gcc works perfectly for me, produces faster/cleaner code, and doesn't try to support everything from frying pans to flyswatters...MikeB wrote:LOL that is ridiculous ..sje wrote:I don't use Xcode's PGO, and it's a wonder that it ever did work as Apple has little interest in optimization.
If Apple did have an interest in optimization, then it wouldn't have included gigabytes of crap in its Xcode distribution for its watches, phones, TV boxes, tablets, etc. I was watching the files being installed over the hour or so it took to update; included were things like different sized icons for Seattle area public transit stations.