Add QE PW test suite - #2373
Conversation
There was a problem hiding this comment.
Pull request overview
Adds a Quantum ESPRESSO (PWscf) test-suite runner to the AOMP test harness so QE PW can be cloned via the test manifest and exercised with the nightly compiler toolchain.
Changes:
- Extended
manifests/test_24.0.xmlwith a newqeompremote and a QE OMP-offload fork project entry. - Added
bin/run_qe.shto configure/build QE PW with AOMP + ROCm and run the PW test-suite under MPI with per-rank GPU pinning.
Reviewed changes
Copilot reviewed 1 out of 2 changed files in this pull request and generated no comments.
| File | Description |
|---|---|
| manifests/test_24.0.xml | Adds the QE fork remote/project so clone_test.sh can fetch the sources into the test workspace. |
| bin/run_qe.sh | New end-to-end build + PW test-suite runner, including generated GPU affinity launcher and OpenMPI binding setup. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
dpalermo
left a comment
There was a problem hiding this comment.
When I run the script I see QE build, but I don't see the tests run:
run on r11:
AOMP=/COD/LATEST/aomp/llvm ROCM_PATH=/COD/LATEST/aomp AOMP_SUPP=/COD/LATEST/aomp/local ./run_qe.sh
If I then run with rerun, the tests are run, but are all failing:
AOMP=/COD/LATEST/aomp/llvm ROCM_PATH=/COD/LATEST/aomp AOMP_SUPP=/COD/LATEST/aomp/local ./run_qe.sh rerun
| export LIBOMPTARGET_KERNEL_TRACE=${LIBOMPTARGET_KERNEL_TRACE:-0} | ||
|
|
||
| # Setup AOMP variables | ||
| AOMP=${AOMP:-/usr/lib/aomp} |
There was a problem hiding this comment.
change to:
AOMP=${AOMP:-$HOME/rocm/aomp/llvm}
|
|
||
| # ROCm root passed to QE configure (--with-rocm). ROCM_DIR is set by | ||
| # aomp_common_vars; fall back to /opt/rocm for a standard ROCm install. | ||
| ROCM_PATH=${ROCM_PATH:-${ROCM_DIR:-/opt/rocm}} |
There was a problem hiding this comment.
change to:
ROCM_PATH=${ROCM_PATH:-$(realpath -m $(realpath -m $AOMP)/../..)}
|
@dpalermo could you please share some log files? You can find both .out and .err inside each test folder contained in the test-suite from the QE root directory. What about the rerun issue? Any log file from the first failing (AOMP=/COD/LATEST/aomp/llvm ROCM_PATH=/COD/LATEST/aomp AOMP_SUPP=/COD/LATEST/aomp/local ./run_qe.sh) command? |
I put copies of the log files here (captured stdout/stderr from running run_qe.sh): [r11 ~/git/aomp24.0/aomp/bin]$ ll ~dpalermo/4Giacomo/ I also see many *.out files on r11 in ~dpalermo/git/aomp-test/q-e-omp that seem to show the tests actually ran. I won't rerun for now if you want to take a look at what is currently there. |
dpalermo
left a comment
There was a problem hiding this comment.
Met with Giacomo and am able to run now (I was missing rocblas with the build I was trying to run). Previous review comments are just the defaults, can be fixed later.
Motivation
Add Quantum Espresso PW test suite run script to test this app with the nightly compiler.
Technical Details
Added run_qe.sh script to bin/ and updated manifest/test_24.0.xml
Test Plan
I tested the script on r11, launching it with the command: AOMP=/mnt/COT/multiarch-2026-07-28 ROCM_PATH=/mnt/COT/multiarch-2026-07-28 ./run_qe.sh
Test Result
The code has been succesfully cloned, build and test suite has been executed.