# intent behind that configuration, so that we can correctly judge
# whether to preserve that during maintenance decisions.
#
+# Other configurations might coincidentally test such configurations
+# (e.g. because they are the current default), but it is appropriate
+# to intend to teach each feature (or a feature combination) exactly
+# once, and for the intent to be reflected precisely in the
+# configuration syntax, so that the configurations are stable even
+# if the defaults change in future.
+#
# The phrase "at time of release" refers to either the initial
# GROMACS release from that branch, or the next expected release
# from that branch. (We will tend not to retroactively support
# Test newest gcc supported by newest CUDA at time of release
# Test thread-MPI with CUDA
-# Test SIMD (AVX2_256) GPU code-path
-gcc-5 gpu cuda-8.0 openmp release simd=avx2_256
+# Test SIMD implementation of pair search for GPU code-path
+gcc-5 gpu cuda-8.0 thread-mpi openmp release simd=avx2_256
# Test with ThreadSanitizer (without OpenMP, because of Redmine #1850)
# Test AVX2_256 SIMD
gcc-4.9 tsan fftpack simd=avx2_256
# Test newest gcc at time of release
-# Test on MacOS
+# Test on MacOS (because gcc-6 is only available there)
gcc-6 double
# Test older clang
# Test double precision
# Test without OpenMP
-clang-3.4 double no-openmp fftpack
+# Test thread-MPI
+clang-3.4 double thread-mpi no-openmp fftpack
# Test newest clang at time of release
# Test with AddressSanitizer (without OpenMP, see below)
# Test newest cmake at time of release
# Test MKL
# Test without any MPI
-# Test on CentOS
+# Test on CentOS (because icc-16.0 is only available there)
icc-16.0 no-thread-mpi openmp mkl cmake-3.6.1 simd=avx_256
# Test oldest supported cmake