{2023.06}[2022b,a64fx] HarfBuzz 5.3.1 (originally built with EB 4.8.2)

[trz42]

Separated from #1051

[trz42]

bot: build instance:eessi-bot-deucalion repository:eessi.io-2023.06-software architecture:aarch64/a64fx

[eessi-bot-deucalion]

New job on instance eessi-bot-deucalion for CPU micro-architecture aarch64-a64fx for repository eessi.io-2023.06-software in job dir /home/eessibot/new-bot/jobs/2025.05/pr_1098/438526

date	job status	comment
May 27 21:16:21 UTC 2025	submitted	job id 438526 awaits release by job manager
May 27 21:17:21 UTC 2025	released	job awaits launch by Slurm scheduler
May 27 21:18:24 UTC 2025	running	job 438526 is running
May 27 22:14:47 UTC 2025	finished	😢 FAILURE (click triangle for details) Details ✅ job output file slurm-438526.out ✅ no message matching FATAL: ❌ found message matching ERROR: ❌ found message matching FAILED: ❌ found message matching required modules missing: ❌ no message matching No missing installations ✅ found message matching .tar.gz created! Artefacts eessi-2023.06-software-linux-aarch64-a64fx-17483835730.tar.gz size: 21 MiB (22037723 bytes) entries: 899 modules under 2023.06/software/linux/aarch64/a64fx/modules/all Bison/3.8.2-GCCcore-12.2.0.lua ICU/72.1-GCCcore-12.2.0.lua Ninja/1.11.1-GCCcore-12.2.0.lua PCRE2/10.40-GCCcore-12.2.0.lua software under 2023.06/software/linux/aarch64/a64fx/software Bison/3.8.2-GCCcore-12.2.0 ICU/72.1-GCCcore-12.2.0 Ninja/1.11.1-GCCcore-12.2.0 PCRE2/10.40-GCCcore-12.2.0 other under 2023.06/software/linux/aarch64/a64fx no other files in tarball
May 27 22:14:47 UTC 2025	test result	😁 SUCCESS (click triangle for details) ReFrame Summary [ SKIP ] (1/9) Skipping test: nodes in this partition only have 30720 MiB memory available (per node) accodring to the current ReFrame configuration, but 49152 MiB is needed [ SKIP ] (2/9) Skipping test: nodes in this partition only have 30720 MiB memory available (per node) accodring to the current ReFrame configuration, but 49152 MiB is needed [ SKIP ] (3/9) Skipping test: nodes in this partition only have 30720 MiB memory available (per node) accodring to the current ReFrame configuration, but 49152 MiB is needed [ SKIP ] (4/9) Skipping test: nodes in this partition only have 30720 MiB memory available (per node) accodring to the current ReFrame configuration, but 49152 MiB is needed [ OK ] (5/9) EESSI_LAMMPS_lj %device_type=cpu %module_name=LAMMPS/2Aug2023_update2-foss-2023a-kokkos %scale=1_node /04ff9ece @BotBuildTests:aarch64_a64fx+default P: perf: 579.282 timesteps/s (r:0, l:None, u:None) [ OK ] (6/9) EESSI_OSU_pt2pt_CPU %benchmark_info=mpi.pt2pt.osu_latency %module_name=OSU-Micro-Benchmarks/7.2-gompi-2023b %scale=1_node /15cad6c4 @BotBuildTests:aarch64_a64fx+default P: latency: 1.74 us (r:0, l:None, u:None) [ OK ] (7/9) EESSI_OSU_pt2pt_CPU %benchmark_info=mpi.pt2pt.osu_latency %module_name=OSU-Micro-Benchmarks/7.1-1-gompi-2023a %scale=1_node /6672deda @BotBuildTests:aarch64_a64fx+default P: latency: 1.77 us (r:0, l:None, u:None) [ OK ] (8/9) EESSI_OSU_pt2pt_CPU %benchmark_info=mpi.pt2pt.osu_bw %module_name=OSU-Micro-Benchmarks/7.2-gompi-2023b %scale=1_node /2a9a47b1 @BotBuildTests:aarch64_a64fx+default P: bandwidth: 7667.77 MB/s (r:0, l:None, u:None) [ OK ] (9/9) EESSI_OSU_pt2pt_CPU %benchmark_info=mpi.pt2pt.osu_bw %module_name=OSU-Micro-Benchmarks/7.1-1-gompi-2023a %scale=1_node /1b24ab8e @BotBuildTests:aarch64_a64fx+default P: bandwidth: 7612.59 MB/s (r:0, l:None, u:None) [ PASSED ] Ran 5/9 test case(s) from 9 check(s) (0 failure(s), 4 skipped, 0 aborted) Details ✅ job output file slurm-438526.out ❌ found message matching ERROR: ✅ no message matching [\s*FAILED\s*].*Ran .* test case

[boegel]

bot: build instance:eessi-bot-deucalion repository:eessi.io-2023.06-software architecture:aarch64/a64fx

[eessi-bot-deucalion]

New job on instance eessi-bot-deucalion for CPU micro-architecture aarch64-a64fx for repository eessi.io-2023.06-software in job dir /home/eessibot/new-bot/jobs/2025.06/pr_1098/461125

date	job status	comment
Jun 15 14:49:08 UTC 2025	submitted	job id 461125 awaits release by job manager
Jun 15 14:49:41 UTC 2025	released	job awaits launch by Slurm scheduler
Jun 15 14:50:44 UTC 2025	running	job 461125 is running
Jun 15 15:42:51 UTC 2025	finished	😢 FAILURE (click triangle for details) Details ✅ job output file slurm-461125.out ✅ no message matching FATAL: ❌ found message matching ERROR: ❌ found message matching FAILED: ❌ found message matching required modules missing: ❌ no message matching No missing installations ✅ found message matching .tar.gz created! Artefacts eessi-2023.06-software-linux-aarch64-a64fx-17500016520.tar.gz size: 21 MiB (22045670 bytes) entries: 899 modules under 2023.06/software/linux/aarch64/a64fx/modules/all Bison/3.8.2-GCCcore-12.2.0.lua ICU/72.1-GCCcore-12.2.0.lua Ninja/1.11.1-GCCcore-12.2.0.lua PCRE2/10.40-GCCcore-12.2.0.lua software under 2023.06/software/linux/aarch64/a64fx/software Bison/3.8.2-GCCcore-12.2.0 ICU/72.1-GCCcore-12.2.0 Ninja/1.11.1-GCCcore-12.2.0 PCRE2/10.40-GCCcore-12.2.0 other under 2023.06/software/linux/aarch64/a64fx no other files in tarball
Jun 15 15:42:51 UTC 2025	test result	😁 SUCCESS (click triangle for details) ReFrame Summary [ SKIP ] (1/9) Skipping test: nodes in this partition only have 30720 MiB memory available (per node) accodring to the current ReFrame configuration, but 49152 MiB is needed [ SKIP ] (2/9) Skipping test: nodes in this partition only have 30720 MiB memory available (per node) accodring to the current ReFrame configuration, but 49152 MiB is needed [ SKIP ] (3/9) Skipping test: nodes in this partition only have 30720 MiB memory available (per node) accodring to the current ReFrame configuration, but 49152 MiB is needed [ SKIP ] (4/9) Skipping test: nodes in this partition only have 30720 MiB memory available (per node) accodring to the current ReFrame configuration, but 49152 MiB is needed [ OK ] (5/9) EESSI_OSU_pt2pt_CPU %benchmark_info=mpi.pt2pt.osu_latency %module_name=OSU-Micro-Benchmarks/7.2-gompi-2023b %scale=1_node /15cad6c4 @BotBuildTests:aarch64_a64fx+default P: latency: 1.73 us (r:0, l:None, u:None) [ OK ] (6/9) EESSI_OSU_pt2pt_CPU %benchmark_info=mpi.pt2pt.osu_latency %module_name=OSU-Micro-Benchmarks/7.1-1-gompi-2023a %scale=1_node /6672deda @BotBuildTests:aarch64_a64fx+default P: latency: 1.81 us (r:0, l:None, u:None) [ OK ] (7/9) EESSI_OSU_pt2pt_CPU %benchmark_info=mpi.pt2pt.osu_bw %module_name=OSU-Micro-Benchmarks/7.2-gompi-2023b %scale=1_node /2a9a47b1 @BotBuildTests:aarch64_a64fx+default P: bandwidth: 7588.3 MB/s (r:0, l:None, u:None) [ OK ] (8/9) EESSI_OSU_pt2pt_CPU %benchmark_info=mpi.pt2pt.osu_bw %module_name=OSU-Micro-Benchmarks/7.1-1-gompi-2023a %scale=1_node /1b24ab8e @BotBuildTests:aarch64_a64fx+default P: bandwidth: 7766.1 MB/s (r:0, l:None, u:None) [ OK ] (9/9) EESSI_LAMMPS_lj %device_type=cpu %module_name=LAMMPS/2Aug2023_update2-foss-2023a-kokkos %scale=1_node /04ff9ece @BotBuildTests:aarch64_a64fx+default P: perf: 416.528 timesteps/s (r:0, l:None, u:None) [ PASSED ] Ran 5/9 test case(s) from 9 check(s) (0 failure(s), 4 skipped, 0 aborted) Details ✅ job output file slurm-461125.out ❌ found message matching ERROR: ✅ no message matching [\s*FAILED\s*].*Ran .* test case

[bedroge]

The build of Rust 1.65.0 failed with a weird error:

[ 54%] Built target LLVMTarget
[ 54%] Building CXX object lib/Transforms/Utils/CMakeFiles/LLVMTransformUtils.dir/PredicateInfo.cpp.o
[ 54%] Building CXX object lib/Transforms/Utils/CMakeFiles/LLVMTransformUtils.dir/PromoteMemoryToRegister.cpp.o
[ 54%] Building CXX object lib/Transforms/Utils/CMakeFiles/LLVMTransformUtils.dir/RelLookupTableConverter.cpp.o
[ 54%] Building CXX object lib/Transforms/Utils/CMakeFiles/LLVMTransformUtils.dir/ScalarEvolutionExpander.cpp.o
[ 54%] Building CXX object lib/Transforms/Utils/CMakeFiles/LLVMTransformUtils.dir/SCCPSolver.cpp.o
[ 55%] Building CXX object lib/Transforms/Utils/CMakeFiles/LLVMTransformUtils.dir/StripGCRelocates.cpp.o
[ 55%] Linking CXX executable ../../bin/llvm-tli-checker
[ 55%] Linking CXX executable ../../bin/llvm-sim
[ 55%] Linking CXX executable ../../bin/llvm-stress
[ 55%] Building CXX object lib/Transforms/Utils/CMakeFiles/LLVMTransformUtils.dir/SSAUpdater.cpp.o
[ 55%] Building CXX object lib/Transforms/Utils/CMakeFiles/LLVMTransformUtils.dir/SSAUpdaterBulk.cpp.o
[ 55%] Built target llvm-stress
[ 55%] Built target llvm-sim
[ 55%] Building CXX object lib/Transforms/Utils/CMakeFiles/LLVMTransformUtils.dir/SampleProfileInference.cpp.o
[ 55%] Built target llvm-tli-checker
[ 55%] Building CXX object lib/DWP/CMakeFiles/LLVMDWP.dir/DWP.cpp.o
[ 55%] Building CXX object lib/DWP/CMakeFiles/LLVMDWP.dir/DWPError.cpp.o
/tmp/eb-ujicqi2g/eb-q95hr60e/ccMOYzFH.s: Assembler messages:
/tmp/eb-ujicqi2g/eb-q95hr60e/ccMOYzFH.s:18063: Error: immediate offset out of range 0 to 31 at operand 3 -- `st1b z2.d,p0,[z0.d,#88]'

[ocaisa]

EasyBuild LLM integration would have probably helped here:

This is an ARM SVE (Scalable Vector Extension) assembly instruction (st1b) where the offset #88 is too large for what the instruction allows.

Specifically, st1b stores bytes to memory using predicate-controlled SVE addressing, and the immediate offset must be within the range [0, 31] (i.e., it's encoded in 5 bits). An offset of #88 is out of range.

[bedroge]

EasyBuild LLM integration would have probably helped here:

This is an ARM SVE (Scalable Vector Extension) assembly instruction (st1b) where the offset #88 is too large for what the instruction allows.

Specifically, st1b stores bytes to memory using predicate-controlled SVE addressing, and the immediate offset must be within the range [0, 31] (i.e., it's encoded in 5 bits). An offset of #88 is out of range.

That is indeed helpful 😄 But can it also provide a patch? 😅

[ocaisa]

It tries, one thing it says is

Make sure your assembler (binutils) supports the instruction set your compiler is generating.

and I wonder if the vector widths of a64fx (which IIRC are especially wide) are actually supported.

The other option is to switch off SVE somehow.

[bedroge]

It tries, one thing it says is

Make sure your assembler (binutils) supports the instruction set your compiler is generating.

and I wonder if the vector widths of a64fx (which IIRC are especially wide) are actually supported.

The other option is to switch off SVE somehow.

A slightly newer Rust (1.70.0 instead of 1.65.0) built with GCCcore-12.3.0 instead of GCCcore 12.2.0 does work fine. And the binutils version is fixed, since that's taken from the compat layer.