[SteveBronder]

Summary

Each translation unit currently gets its own opencl_context wrapper and kernel-cache registry. Use one C++17 inline wrapper so device selection can invalidate kernels registered from every translation unit. Add a context-identity regression and kernel execution before and after selecting a device through another translation unit.

Part 2 of the CMake migration split from #3463.

Targets cmake-split/01-compile-hygiene (#3501) as the preceding layer of the native stack. The review contains only this layer's prepared patch and its supporting tests.

Tests

• The new context-identity test fails with the baseline static wrapper and passes with the inline wrapper.
• The independent Make shared-library regression passes.
• GPU kernel execution and device-switch runtime coverage remains for CI.

Side Effects

OpenCL kernel-cache registration is shared across translation units; this changes the linkage of the context wrapper to match grouped execution.

Release notes

Share OpenCL kernel-cache registration across translation units.

Checklist

• Copyright holder: Simons Foundation
• The complete basic-test and CI matrix is passing
• C++ idioms and Doxygen documentation have completed review
• Focused validation is passing as described above; remaining coverage is listed

Related PRs

• 1/7: [CMake 1/7] Fix test compilation with strict compiler warnings
• 2/7: [CMake 2/7] Share OpenCL kernel-cache registration across translation units (this PR)
• 3/7: [CMake 3/7] Isolate TBB construction tests from process-wide initialization
• 4/7: [CMake 4/7] Make probability test fixtures safe for grouped linking
• 5/7: [CMake 5/7] Make unit tests safe for grouped execution
• 6/7: [CMake 6/7] Initialize CVODES adjoint quadrature derivative buffers
• 7/7: [CMake 7/7] Add CMake builds and migrate the test runner and CI

Native stack: https://github.com/stan-dev/math/stack/3509

The merge-base changed after approval.

[stan-buildbot]

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
stat_comp_benchmarks/benchmarks/gp_regr/gen_gp_data.stan	0.04	0.04	1.02	1.6% faster
stat_comp_benchmarks/benchmarks/gp_regr/gp_regr.stan	0.18	0.18	1.0	-0.02% slower
stat_comp_benchmarks/benchmarks/garch/garch.stan	0.99	0.98	1.01	0.93% faster
stat_comp_benchmarks/benchmarks/low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	0.99	-1.38% slower
stat_comp_benchmarks/benchmarks/eight_schools/eight_schools.stan	0.11	0.11	1.0	-0.01% slower
stat_comp_benchmarks/benchmarks/low_dim_gauss_mix/low_dim_gauss_mix.stan	8.45	8.4	1.01	0.6% faster
stat_comp_benchmarks/benchmarks/gp_pois_regr/gp_pois_regr.stan	4.09	3.99	1.03	2.45% faster
stat_comp_benchmarks/benchmarks/pkpd/one_comp_mm_elim_abs.stan	31.64	31.87	0.99	-0.73% slower
stat_comp_benchmarks/benchmarks/pkpd/sim_one_comp_mm_elim_abs.stan	0.37	0.38	0.99	-0.96% slower
stat_comp_benchmarks/benchmarks/irt_2pl/irt_2pl.stan	7.96	8.03	0.99	-0.9% slower
stat_comp_benchmarks/benchmarks/sir/sir.stan	166.41	166.36	1.0	0.03% faster
stat_comp_benchmarks/benchmarks/arK/arK.stan	3.44	3.48	0.99	-1.04% slower
stat_comp_benchmarks/benchmarks/arma/arma.stan	0.58	0.57	1.01	1.29% faster
stat_comp_benchmarks/benchmarks/low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	25.27	25.24	1.0	0.12% faster
performance.compilation	383.55	394.42	0.97	-2.83% slower
Mean result: 0.9995966556824036

Jenkins Console Log
Jenkins Build Stages
Commit hash: 148669cf074a47c2622971790ef9133551135df7

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:                           52 bits physical, 57 bits virtual
CPU(s):                                  192
On-line CPU(s) list:                     0-191
Thread(s) per core:                      2
Core(s) per socket:                      48
Socket(s):                               2
NUMA node(s):                            2
Vendor ID:                               AuthenticAMD
CPU family:                              25
Model:                                   17
Model name:                              AMD EPYC 9474F 48-Core Processor
Stepping:                                1
Frequency boost:                         enabled
CPU MHz:                                 1497.201
CPU max MHz:                             4114.4229
CPU min MHz:                             1500.0000
BogoMIPS:                                7189.43
Virtualization:                          AMD-V
L1d cache:                               3 MiB
L1i cache:                               3 MiB
L2 cache:                                96 MiB
L3 cache:                                512 MiB
NUMA node0 CPU(s):                       0-47,96-143
NUMA node1 CPU(s):                       48-95,144-191
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:                  Not affected
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; Enhanced / Automatic IBRS; IBPB conditional; STIBP always-on; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Mitigation; Clear CPU buffers
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 amd_lbr_v2 nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 pcid 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 perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local user_shstk avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif x2avic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid overflow_recov succor smca fsrm flush_l1d debug_swap

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