ad embedded expression tests to expect_ad testing framework

[SteveBronder]

Summary

This adds a simple version of the expression testing framework directly into the call to expect_ad(). This helps fix #2736 by making sure any function that goes through the expression tests will also go check that it is safe to pass Eigen expressions to.

The files for all of this are in test/unit/math/expr_tests.hpp with the main function being check_expr_test(...). This function takes in any set of arguments and for each type that is an Eigen matrix or vector it checks if the function is 'safe' to pass expressions to. The word 'safe' in this context means that, if an unary expression was passed as an argument to the function, is it only called as many times as there are individual elements in the underlying Eigen object? The original Eigen object is wrapped in an unaryFunction that calls the counterOp functor from the original expression testing framework.

template <typename Scal>
struct counterOp {
  int* counter_;
  counterOp(int* counter) { counter_ = counter; }
  const Scal& operator()(const Scal& a) const {
    (*counter_)++;
    return a;
  }
};
// In the code this is used on a matrix x like 
int call_counter;
auto x_op = x.unaryExpr(counterOp(&call_counter));
double b = x_op[0];
double c = x_op[0];
double d = x_op[0];
std::cout << call_counter << std::endl;
// prints "3"

Tests

Tests can be run with

./runTests.py test/unit/math/expr_tests_test.cpp

The tests include examples for one, two, and three inputs to functions. They check that for each combination of inputs we fail the correct number of Eigen objects passed to each function. Each set of inputs also has an associated test that just runs a 'good' funtion which does not fail the tests.

Side Effects

This will increase compile times a bit

Release notes

Adds a simple expression tests framework to the autodiff test framework

Checklist

• Math issue Document Stan Math expressions requirements  #2736

• Copyright holder: Steve Bronder

  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]

The loss of two hours of wait time to prevent upstream failures is worth it

I just wanted to make sure it was clear: these upstream failures are in PRs, not in anything user-facing. So it prevents some developer frustration and churn here in that situation. The same thing could be prevented if the user was told anywhere how to preemptively run the tests, which is what the discussion in #2736 was about

[andrjohns]

Needing to compile all of mix the tests first, only to have a failure in one or two can lead to a bit of wasted computation.

The --jumbo flag already compiles in batches, would it be possible to change the behaviour to run a given batch as soon as it's compiled? That could help things fail faster, especially if we look into tuning the size of the batches

[WardBrian]

It would be possible, but I don't think that is an easy change in the current test infrastructure, since make doesn't tell you when some targets are done. I'd be happy to be proven wrong there.

Care would need to be taken to make sure the output was still readable from this

[andrjohns]

It would be possible, but I don't think that is an easy change in the current test infrastructure, since make doesn't tell you when some targets are done. I'd be happy to be proven wrong there.

Care would need to be taken to make sure the output was still readable from this

I was thinking of the steps:

1. Generate batched test files (prob_1.cpp, prob_2.cpp, etc.), but don't compile
2. Use python's (or bash's) threading to call runTests.py -j1 x on batches in parallel

That way the compilation is still parallelised, but each call is only building and immediately running one test. Whether or not this would be particularly 'clean', I'm not sure

[WardBrian]

I think something of that flavor could work as long as anything shared between tests (like TBB or Sundials) was already built, otherwise you'd get filesystem races. Worth exploring things like that.

[stan-buildbot]

---

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
arma/arma.stan	0.17	0.17	1.04	4.3% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	1.05	4.76% faster
gp_regr/gen_gp_data.stan	0.02	0.02	1.06	5.77% faster
gp_regr/gp_regr.stan	0.1	0.1	1.02	2.33% faster
sir/sir.stan	75.82	74.29	1.02	2.02% faster
irt_2pl/irt_2pl.stan	3.59	3.71	0.97	-3.39% slower
eight_schools/eight_schools.stan	0.05	0.05	1.03	2.5% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.23	0.23	1.01	0.62% faster
pkpd/one_comp_mm_elim_abs.stan	17.06	16.73	1.02	1.94% faster
garch/garch.stan	0.43	0.43	1.0	0.21% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.67	2.65	1.01	0.64% faster
arK/arK.stan	1.54	1.55	0.99	-0.69% slower
gp_pois_regr/gp_pois_regr.stan	2.39	2.4	1.0	-0.25% slower
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.69	8.62	1.01	0.82% faster
performance.compilation	176.23	177.15	0.99	-0.52% slower
Mean result: 1.0147850941500054

---

Jenkins Console Log
Blue Ocean
Commit hash: bc29abd03525f49136326f8e3ab87d96c3be2e56

---

Machine information

No LSB modules are available. 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: 46 bits physical, 48 bits virtual
CPU(s): 80
On-line CPU(s) list: 0-79
Thread(s) per core: 2
Core(s) per socket: 20
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Stepping: 4
CPU MHz: 2400.000
CPU max MHz: 3700.0000
CPU min MHz: 1000.0000
BogoMIPS: 4800.00
Virtualization: VT-x
L1d cache: 1.3 MiB
L1i cache: 1.3 MiB
L2 cache: 40 MiB
L3 cache: 55 MiB
NUMA node0 CPU(s): 0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78
NUMA node1 CPU(s): 1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79
Vulnerability Gather data sampling: Mitigation; Microcode
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Retbleed: Mitigation; IBRS
Vulnerability Spec rstack overflow: Not affected
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; IBRS, IBPB conditional, STIBP conditional, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Mitigation; Clear CPU buffers; SMT vulnerable
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single pti intel_ppin ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req pku ospke md_clear flush_l1d arch_capabilities

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

[SteveBronder]

@WardBrian do you have any idea what is going on with the benchmark tests failing? I'm not seeing an error in the logs?

https://jenkins.flatironinstitute.org/blue/rest/organizations/jenkins/pipelines/Stan/pipelines/CmdStan%20Performance%20Tests/branches/downstream_tests/runs/953/log/?start=0

[SteveBronder]

It looks like the failure is coming from

+ curl -s -S 'https://jenkins.flatironinstitute.org/job/Stan/job/CmdStan%20Performance%20Tests/job/downstream_tests/953//logText/progressiveText?start=0'
curl: (7) Failed to connect to jenkins.flatironinstitute.org port 443: Connection timed out

Seems weird?

[stan-buildbot]

---

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
arma/arma.stan	0.33	0.43	0.77	-30.14% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	0.82	-21.51% slower
gp_regr/gen_gp_data.stan	0.02	0.03	0.85	-17.04% slower
gp_regr/gp_regr.stan	0.1	0.12	0.84	-19.44% slower
sir/sir.stan	72.63	74.07	0.98	-1.99% slower
irt_2pl/irt_2pl.stan	4.27	4.32	0.99	-1.05% slower
eight_schools/eight_schools.stan	0.06	0.06	1.07	6.43% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.26	0.25	1.05	4.78% faster
pkpd/one_comp_mm_elim_abs.stan	20.72	20.41	1.02	1.5% faster
garch/garch.stan	0.45	0.41	1.1	9.25% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.82	2.63	1.07	6.74% faster
arK/arK.stan	1.89	1.77	1.06	6.05% faster
gp_pois_regr/gp_pois_regr.stan	3.1	3.16	0.98	-1.85% slower
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	9.16	8.9	1.03	2.86% faster
performance.compilation	190.82	199.75	0.96	-4.68% slower
Mean result: 0.9728302747107814

---

Jenkins Console Log
Blue Ocean
Commit hash: fe1ecd9ed58c8acfabb1e8d59239ec6b930b570f

---

Machine information

No LSB modules are available. 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: 46 bits physical, 48 bits virtual
CPU(s): 80
On-line CPU(s) list: 0-79
Thread(s) per core: 2
Core(s) per socket: 20
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Stepping: 4
CPU MHz: 3640.454
CPU max MHz: 3700.0000
CPU min MHz: 1000.0000
BogoMIPS: 4800.00
Virtualization: VT-x
L1d cache: 1.3 MiB
L1i cache: 1.3 MiB
L2 cache: 40 MiB
L3 cache: 55 MiB
NUMA node0 CPU(s): 0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78
NUMA node1 CPU(s): 1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79
Vulnerability Gather data sampling: Mitigation; Microcode
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Retbleed: Mitigation; IBRS
Vulnerability Spec rstack overflow: Not affected
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; IBRS, IBPB conditional, STIBP conditional, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Mitigation; Clear CPU buffers; SMT vulnerable
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single pti intel_ppin ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req pku ospke md_clear flush_l1d arch_capabilities

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

[SteveBronder]

@WardBrian I think this is good to merge after the release is finished

[stan-buildbot]

---

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
arma/arma.stan	0.42	0.38	1.11	10.07% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	1.03	2.61% faster
gp_regr/gen_gp_data.stan	0.03	0.03	1.05	4.57% faster
gp_regr/gp_regr.stan	0.1	0.1	1.02	1.84% faster
sir/sir.stan	76.82	77.09	1.0	-0.34% slower
irt_2pl/irt_2pl.stan	5.14	4.83	1.06	6.04% faster
eight_schools/eight_schools.stan	0.06	0.06	0.96	-4.05% slower
pkpd/sim_one_comp_mm_elim_abs.stan	0.37	0.29	1.28	22.01% faster
pkpd/one_comp_mm_elim_abs.stan	20.89	21.91	0.95	-4.9% slower
garch/garch.stan	0.51	0.51	1.01	0.71% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.91	2.96	0.99	-1.47% slower
arK/arK.stan	2.06	2.06	1.0	-0.15% slower
gp_pois_regr/gp_pois_regr.stan	3.19	3.21	1.0	-0.47% slower
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	9.77	9.89	0.99	-1.28% slower
performance.compilation	199.37	205.47	0.97	-3.06% slower
Mean result: 1.0271448709133995

---

Jenkins Console Log
Blue Ocean
Commit hash: d669f2b029575d5191b8f394b58b652e4bc4c805

---

Machine information

No LSB modules are available. 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: 46 bits physical, 48 bits virtual
CPU(s): 80
On-line CPU(s) list: 0-79
Thread(s) per core: 2
Core(s) per socket: 20
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Stepping: 4
CPU MHz: 2400.000
CPU max MHz: 3700.0000
CPU min MHz: 1000.0000
BogoMIPS: 4800.00
Virtualization: VT-x
L1d cache: 1.3 MiB
L1i cache: 1.3 MiB
L2 cache: 40 MiB
L3 cache: 55 MiB
NUMA node0 CPU(s): 0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78
NUMA node1 CPU(s): 1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79
Vulnerability Gather data sampling: Mitigation; Microcode
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Mitigation; IBRS
Vulnerability Spec rstack overflow: Not affected
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; IBRS; IBPB conditional; STIBP conditional; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Mitigation; Clear CPU buffers; SMT vulnerable
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single pti intel_ppin ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req pku ospke md_clear flush_l1d arch_capabilities

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