fix gamma lccdf for alpha > 30 and overflow

[spinkney]

Summary

This pr closes #3257. The gamma_lccdf experiences numerical instability for alpha approximately greater than 30. It’s due to gamma_q or grad_reg_inc_gamma. This pr rewrites the code to use gamma_p and grad_reg_lower_inc_gamma, which avoids potential instability in a tgamma(alpha) and digamma(alpha) call. It seems to sample a bit faster than just wrapping gamma_lcdf with log1m_exp.

Tests

I've implemented new tests in the problematic regime.

Release notes

The gamma_lccdf is made more robust for alpha values. The issue was overflow in the derivatives when alpha was greater than 30.

Checklist

• Copyright holder: Publicis Groupe
  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

[spinkney]

@WardBrian I think the test failure is a GitHub actions failure. Can this be restarted?

[stan-buildbot]

---

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_regr/gp_regr.stan	0.09	0.09	1.0	-0.24% slower
gp_regr/gen_gp_data.stan	0.02	0.02	1.02	1.85% faster
arK/arK.stan	1.72	1.7	1.01	0.97% faster
eight_schools/eight_schools.stan	0.05	0.05	1.03	2.92% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.47	8.35	1.02	1.48% faster
pkpd/one_comp_mm_elim_abs.stan	18.72	18.35	1.02	1.99% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.24	0.24	1.02	1.58% faster
sir/sir.stan	68.96	67.36	1.02	2.31% faster
gp_pois_regr/gp_pois_regr.stan	2.73	2.73	1.0	0.02% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.6	2.56	1.01	1.38% faster
irt_2pl/irt_2pl.stan	3.99	3.95	1.01	1.15% faster
arma/arma.stan	0.28	0.27	1.03	2.53% faster
garch/garch.stan	0.4	0.4	1.0	0.42% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	1.0	-0.36% slower
performance.compilation	213.18	219.77	0.97	-3.09% slower
Mean result: 1.0102555202019106

---

Jenkins Console Log
Blue Ocean
Commit hash: c6a125f2a31a30b9377a95fa7a388007e625cfcb

---

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: 1096.333
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
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 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

[WardBrian]

@spinkney that did go away, but there are some jenkins failures that look legitimate: https://jenkins.flatironinstitute.org/blue/organizations/jenkins/Stan%2FMath/detail/PR-3264/3/pipeline/591

Locally run

./runTests.py -j8 test/prob/gamma/gamma_ccdf_log_0*_test.cpp

I do get some failures, but seemingly fewer than CI, when I run this.

[spinkney]

@spinkney that did go away, but there are some jenkins failures that look legitimate: https://jenkins.flatironinstitute.org/blue/organizations/jenkins/Stan%2FMath/detail/PR-3264/3/pipeline/591

Locally run

./runTests.py -j8 test/prob/gamma/gamma_ccdf_log_0*_test.cpp

I do get some failures, but seemingly fewer than CI, when I run this.

Those are legit. Is this all around having a 0 as input?

[WardBrian]

I don't think we're passing zero in this test: https://github.com/stan-dev/math/blob/develop/test/prob/gamma/gamma_ccdf_log_test.hpp

It sounds like it's an issue with vector/matrix inputs

[spinkney]

@WardBrian I'll look at this tomorrow. I suspect it's the addition of grad_reg_lower_inc_gamma. I don't think this is exposed so it might not have all the necessary templating to handle the matrix/vector inputs.

[spinkney]

@WardBrian the recent push passes locally. Fixed two vectorization bugs.

The first, when alpha is a constant vector it was treated like a scalar, so lgamma(alpha) would get cached from the first element, ie alpha[0], and reused for all elements (breaks gradients for vector/matrix inputs).

The second happened with y == 0 which now sums correctly over vectors.

I was able to simplify the expression with [[maybe_unused]], which I found described at https://stackoverflow.com/a/49321420/1207834. The [[maybe_unused]] lets us avoid duplicating the loop or adding (void)alpha_is_scalar;. Let me know if that it ok as it's c++17 stuff and I haven't seen it used in the rest of Stan-math.

[stan-buildbot]

---

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_regr/gp_regr.stan	0.1	0.09	1.03	3.25% faster
gp_regr/gen_gp_data.stan	0.03	0.02	1.02	2.28% faster
arK/arK.stan	1.94	1.82	1.07	6.26% faster
eight_schools/eight_schools.stan	0.06	0.05	1.06	5.95% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	9.42	8.88	1.06	5.76% faster
pkpd/one_comp_mm_elim_abs.stan	20.63	19.51	1.06	5.42% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.28	0.25	1.13	11.12% faster
sir/sir.stan	74.37	70.65	1.05	5.01% faster
gp_pois_regr/gp_pois_regr.stan	2.99	2.86	1.04	4.1% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.79	2.74	1.02	1.89% faster
irt_2pl/irt_2pl.stan	4.5	4.38	1.03	2.78% faster
arma/arma.stan	0.3	0.3	1.01	0.8% faster
garch/garch.stan	0.46	0.44	1.05	4.92% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	1.02	1.53% faster
performance.compilation	233.43	236.6	0.99	-1.36% slower
Mean result: 1.0424062039904969

---

Jenkins Console Log
Blue Ocean
Commit hash: c6a125f2a31a30b9377a95fa7a388007e625cfcb

---

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: 3289.842
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
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 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

[spinkney]

This needs #3265 before merging.