Unpleasant: LLVM/Clang 3.2 On The ARM Cortex-A15

Since publishing LLVM/Clang 3.2 benchmarks a few days ago that showed the Clang C/C++ compiler competing with -- and in some cases outperforming -- the GCC compiler on Intel x86_64, several Phoronix readers have been asking how things compare on the ARM side.

While a few weeks back I published some early benchmarks of GCC vs. LLVM/Clang on the ARM Cortex-A15 using the Samsung Google Chromebook with the Exynos 5 Dual, it seems since the LLVM/Clang 3.0 release, there's some ARMv7 regressions that have come up.

After building the final release of LLVM/Clang 3.2 in the same way to previous LLVM/Clang releases on the same hardware and in a similar manner to other ARM Cotex-A9 development boards, I've been hitting some pain points. With the Samsung Exynos 5 Dual's ARM Cortex-A15, I'm now hitting new problems with LLVM/Clang 3.2 not encountered on the A9 hardware tested. Segmentation faults in Clang have become common in this latest A15 compiler testing on Linux:

4. Running pass 'ARM Instruction Selection' on function '@__cxx_global_var_init'
clang: error: unable to execute command: Segmentation fault (core dumped)
clang: error: clang frontend command failed due to signal (use -v to see invocation)
clang version 3.2 (tags/RELEASE_32/final)
Target: armv7l-unknown-linux-gnueabihf
Thread model: posix
clang: note: diagnostic msg: PLEASE submit a bug report to http://llvm.org/bugs/ and include the crash backtrace, preprocessed source, and associated run script.
clang: note: diagnostic msg:
********************
PLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT:
Preprocessed source(s) and associated run script(s) are located at:
clang: note: diagnostic msg: /tmp/smallpt-z07Dod.cpp
clang: note: diagnostic msg: /tmp/smallpt-z07Dod.sh
clang: note: diagnostic msg:
********************

The Clang segmentation faults have been common within the ARM Instruction Selection pass on this release that came out last week and has occurred for multiple test profiles on different functions. This A15 upset is sad to see with the ARM Cortex-A15 performance being a huge upgrade over the A9-based ARM SoCs.

LLVM 3.2 brought many new features, including on the ARM side, but unfortunately not without some fallout. LLVM/Clang 3.2 benchmarks from ARM Cortex-A9 hardware is forthcoming while it's likely that the Cortex-A15 benchmarks will be delayed to the next release unless a workaround/fix is soon discovered.