Simple for-loop optimizations in Assembly for AD

[GiudGiud]

refs #10256
what is a good case to check if it's impactful or not?
is there a reference AD + displaced mesh case we can try?

[moosebuild]

Job Documentation, step Docs: sync website on 0871122 wanted to post the following:

View the site here

This comment will be updated on new commits.

[moosebuild]

Job Coverage, step Generate coverage on 0871122 wanted to post the following:

Framework coverage

	3403d2	#30036 087112
	Total	Total	+/-	New
Rate	85.29%	85.29%	-0.00%	57.69%
Hits	109051	109057	+6	30
Misses	18808	18813	+5	22

Diff coverage report

Full coverage report

Modules coverage

Coverage did not change

Full coverage reports

Reports

• framework
• chemical_reactions
• combined
• contact
• electromagnetics
• external_petsc_solver
• fluid_properties
• fsi
• functional_expansion_tools
• geochemistry
• heat_transfer
• level_set
• misc
• navier_stokes
• optimization
• peridynamics
• phase_field
• porous_flow
• ray_tracing
• rdg
• reactor
• richards
• scalar_transport
• solid_mechanics
• solid_properties
• stochastic_tools
• subchannel
• thermal_hydraulics
• xfem

Warnings

• framework new line coverage rate 57.69% is less than the suggested 90.0%

This comment will be updated on new commits.

[lindsayad]

Have we figured out whether make_range vectorizes? If we haven't figured it out, can we? If we have figured it out and it vectorizes, let's switch to it

[GiudGiud]

nothing vectorizes at the moment as far I could tell when I tried.

My next hope for vectorizing is Logan switching out MooseArrays to std::vector for variables and a lot of other constructs

[GiudGiud]

changed to range loops for qps.
I tried on simple - diffusion + r 3.5 + displacements and no improvements on runtime

Though we do take a x2.2 slowdown when turning on the displaced mesh.

[lindsayad]

Though we do take a x2.2 slowdown when turning on the displaced mesh.

Wait waaat? Using make_range gives you a 2.2 factor slow-down?

[GiudGiud]

Wait waaat? Using make_range gives you a 2.2 factor slow-down?

no no no it's turning on the displaced mesh that does that

[GiudGiud]

This input shows it. You can comment out the global params to turn on/off

[Mesh]
  type = GeneratedMesh
  dim = 3
  nx = 15
  ny = 15
  nz = 15

  uniform_refine = 3
[]

[GlobalParams]
  use_displaced_mesh = true
  displacements = 'disp_x disp_y disp_z'
[]

[AuxVariables]
  [disp_x]
  []
  [disp_y]
  []
  [disp_z]
  []
[]

[Variables]
  [./u]
  [../]
[]

[Kernels]
  [./diff]
    type = ADDiffusion
    variable = u
  [../]
[]

[BCs]
  [./left]
    type = DirichletBC
    variable = u
    boundary = left
    value = 0
  [../]
  [./right]
    type = DirichletBC
    variable = u
    boundary = right
    value = 1
  [../]
[]

[Preconditioning]
  [./smp]
    type = SMP
    full = true
  [../]
[]

[Executioner]
  type = Steady

  # Preconditioned JFNK (default)
  # solve_type = 'Newton'


  petsc_options_iname = '-pc_type -pc_hypre_type'
  petsc_options_value = 'hypre boomeramg'

  l_tol = 1e-10
  nl_rel_tol = 1e-4
  nl_max_its = 1
[]

[Outputs]
  exodus = true
[]

[GiudGiud]

Doco failure unrelated