[jgabry]

AI disclaimer: I used Claude to help investigate this but I made the small code change myself. Claude helped with testing too (see below).

Summary

When TBB_LIB is set, make/compiler_flags links with -Wl,-rpath plus -Wl,--disable-new-dtags. Without it set the TBB in lib/tbb only adds -Wl,-rpath.

On Linux -rpath without --disable-new-dtags writes DT_RUNPATH, which the loader checks after LD_LIBRARY_PATH (for DT_RPATH it's the other way around). So anything linked with the bundled TBB runs on whatever libtbb.so.2 is in an LD_LIBRARY_PATH directory.

@avehtari noticed this in stan-dev/cmdstanr#1177 (last comment). R sets LD_LIBRARY_PATH for every process it starts and on Debian/Ubuntu that includes /usr/lib/x86_64-linux-gnu, so with "Ubuntu's libtbb2 package
installed every CmdStan model run from R uses the system TBB, not the one it was built with. Right now it doesn't matter because both are 2020.x, which is probably why nobody noticed.

This PR moves the LDFLAGS_TBB_DTAGS definition above ifdef TBB_LIB so it applies either way, and adds it next to the -rpath for the bundled TBB.

Tests

No new test, it's a makefile change only. Like I did for #3415, I had Claude print LDFLAGS_TBB for each OS from a small makefile that includes make/compiler_flags. Only the Linux row without TBB_LIB changes:

before
Linux:              [-Wl,-L,<math>/lib/tbb   -Wl,-rpath,<math>/lib/tbb]
Linux TBB_LIB:      [-Wl,-L,/x/lib -Wl,--disable-new-dtags -Wl,-rpath,/x/lib]
Darwin:             [-Wl,-L,<math>/lib/tbb   -Wl,-rpath,<math>/lib/tbb]
Darwin TBB_LIB:     [-Wl,-L,/x/lib  -Wl,-rpath,/x/lib]
Windows_NT:         [-Wl,-L,<math>/lib/tbb  ]
Windows_NT TBB_LIB: [-Wl,-L,/x/lib  ]

after
Linux:              [-Wl,-L,<math>/lib/tbb   -Wl,--disable-new-dtags -Wl,-rpath,<math>/lib/tbb]
Linux TBB_LIB:      [-Wl,-L,/x/lib -Wl,--disable-new-dtags -Wl,-rpath,/x/lib]
Darwin:             [-Wl,-L,<math>/lib/tbb    -Wl,-rpath,<math>/lib/tbb]
Darwin TBB_LIB:     [-Wl,-L,/x/lib  -Wl,-rpath,/x/lib]
Windows_NT:         [-Wl,-L,<math>/lib/tbb  ]
Windows_NT TBB_LIB: [-Wl,-L,/x/lib  ]

I also checked it in Docker (rocker/r-ver:4.3, Ubuntu 22.04, R 4.3.3, CmdStan 2.40.0 but with this change included, Ubuntu's libtbb2 2020.3).

Release notes

Executables and libraries linked with the bundled TBB now use it even if a different TBB is on LD_LIBRARY_PATH, like they already did when TBB_LIB is set.

Checklist

I left some boxes unchecked since this is just a makefile change

• Copyright holder: Jonah Gabry

  The copyright holder is typically you or your assignee, such as a university or company. By submitting this pull request, the copyright holder is agreeing to the license the submitted work under the following licenses:
  - Code: BSD 3-clause (https://opensource.org/licenses/BSD-3-Clause)
  - Documentation: CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/)

• the basic tests are passing

  • unit tests pass (to run, use: ./runTests.py test/unit)
  • header checks pass, (make test-headers)
  • dependencies checks pass, (make test-math-dependencies)
  • docs build, (make doxygen)
  • code passes the built in C++ standards checks (make cpplint)

• the code is written in idiomatic C++ and changes are documented in the doxygen

• the new changes are tested

[WardBrian]

This seems redundant with line 309?

[jgabry]

I don't think so (although I could be wrong). Line 309 is the case where a user points TBB_LIB at their own TBB, and it already included --disable-new-dtags. Line 335 is for the TBB in lib/tbb that CmdStan builds, which had -rpath without it. So this line is actually the fix, the rest of the diff is just because I moved the LDFLAGS_TBB_DTAGS definition above the ifdef so both cases can use it.

[WardBrian]

ah I see. These makefiles are really a mess that I think could be cleaned up significantly if only ever defined LDFLAGS_TBB in one place and just set the constituent parts conditionally, that way we wouldn’t have to totally duplicate the work in the case of a custom TBB location

[jgabry]

Definitely a mess. Do you want that before this fix or can this one go in before the makefile cleanup?

[WardBrian]

We can try to clean it up after this and #3426

[avehtari]

this worked correctly on my Ubuntu 22.04 laptop (g++ 11.4.0) and on a RHEL 9 cluster (GCC 13.3.0)

[stan-buildbot]

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
stat_comp_benchmarks/benchmarks/gp_regr/gp_regr.stan	0.22	0.22	1.01	0.61% faster
stat_comp_benchmarks/benchmarks/gp_regr/gen_gp_data.stan	0.06	0.06	0.99	-0.93% slower
stat_comp_benchmarks/benchmarks/low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	1.02	2.41% faster
stat_comp_benchmarks/benchmarks/arma/arma.stan	0.61	0.61	1.0	-0.22% slower
stat_comp_benchmarks/benchmarks/arK/arK.stan	3.18	3.18	1.0	-0.1% slower
stat_comp_benchmarks/benchmarks/pkpd/one_comp_mm_elim_abs.stan	43.88	43.13	1.02	1.72% faster
stat_comp_benchmarks/benchmarks/pkpd/sim_one_comp_mm_elim_abs.stan	0.58	0.59	0.99	-0.95% slower
stat_comp_benchmarks/benchmarks/gp_pois_regr/gp_pois_regr.stan	4.46	4.47	1.0	-0.31% slower
stat_comp_benchmarks/benchmarks/low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	20.97	20.94	1.0	0.12% faster
stat_comp_benchmarks/benchmarks/garch/garch.stan	0.89	0.9	0.99	-1.16% slower
stat_comp_benchmarks/benchmarks/eight_schools/eight_schools.stan	0.11	0.11	1.02	1.53% faster
stat_comp_benchmarks/benchmarks/irt_2pl/irt_2pl.stan	8.28	8.26	1.0	0.3% faster
stat_comp_benchmarks/benchmarks/low_dim_gauss_mix/low_dim_gauss_mix.stan	6.34	6.33	1.0	0.17% faster
stat_comp_benchmarks/benchmarks/sir/sir.stan	167.89	168.11	1.0	-0.13% slower
performance.compilation	386.93	383.31	1.01	0.94% faster
Mean result: 1.0027729423964387

Jenkins Console Log
Jenkins Build Stages
Commit hash: 017f091b63e5f89b01de871d2a7ead2d78127331

Machine information

Distributor ID:	Ubuntu
Description:	Ubuntu 20.04.3 LTS
Release:	20.04
Codename:	focal

CPU:

Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Byte Order:                              Little Endian
Address sizes:                           43 bits physical, 48 bits virtual
CPU(s):                                  256
On-line CPU(s) list:                     0-255
Thread(s) per core:                      2
Core(s) per socket:                      64
Socket(s):                               2
NUMA node(s):                            2
Vendor ID:                               AuthenticAMD
CPU family:                              23
Model:                                   49
Model name:                              AMD EPYC 7742 64-Core Processor
Stepping:                                0
Frequency boost:                         enabled
CPU MHz:                                 1494.854
CPU max MHz:                             3416.0681
CPU min MHz:                             1500.0000
BogoMIPS:                                4491.56
Virtualization:                          AMD-V
L1d cache:                               4 MiB
L1i cache:                               4 MiB
L2 cache:                                64 MiB
L3 cache:                                512 MiB
NUMA node0 CPU(s):                       0-63,128-191
NUMA node1 CPU(s):                       64-127,192-255
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Old microcode:             Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Mitigation; untrained return thunk; SMT enabled with STIBP protection
Vulnerability Spec rstack overflow:      Mitigation; Safe RET
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Retpolines; IBPB conditional; STIBP always-on; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Mitigation; IBPB before exit to userspace
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif v_spec_ctrl umip rdpid overflow_recov succor smca sev sev_es

G++:

g++ (Ubuntu 9.4.0-1ubuntu1~20.04) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Clang:

clang version 10.0.0-4ubuntu1 
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/bin