Merge branch 'broadcast_changes'

Author: ChrisRackauckas

REQUIRE

@@ -1,5 +1,5 @@
 julia 0.6.0-pre
-DiffEqBase 1.5.0
+DiffEqBase 1.10.0
 Parameters 0.5.0
 ForwardDiff 0.5.0
 GenericSVD 0.0.2

src/caches.jl

@@ -925,7 +925,7 @@ end
 alg_cache(alg::Feagin14,u,rate_prototype,uEltypeNoUnits,tTypeNoUnits,uprev,uprev2,f,t,::Type{Val{false}}) = Feagin14ConstantCache(realtype(uEltypeNoUnits),realtype(tTypeNoUnits))
 
 
-type Rosenbrock23Cache{uType,uArrayType,rateType,du2Type,vecuType,JType,TabType,TFType,UFType,F,JCType} <: OrdinaryDiffEqMutableCache
+type Rosenbrock23Cache{uType,uArrayType,rateType,du2Type,LinuType,vecuType,JType,TabType,TFType,UFType,F,JCType} <: OrdinaryDiffEqMutableCache
   u::uType
   uprev::uType
   k₁::rateType
@@ -946,7 +946,8 @@ type Rosenbrock23Cache{uType,uArrayType,rateType,du2Type,vecuType,JType,TabType,
   tab::TabType
   tf::TFType
   uf::UFType
-  linsolve_tmp::vecuType
+  linsolve_tmp::LinuType
+  linsolve_tmp_vec::vecuType
   linsolve::F
   jac_config::JCType
 end
@@ -956,7 +957,7 @@ du_cache(c::Rosenbrock23Cache) = (c.k₁,c.k₂,c.k₃,c.du1,c.du2,c.f₁,c.fsal
 jac_cache(c::Rosenbrock23Cache) = (c.J,c.W)
 vecu_cache(c::Rosenbrock23Cache) = (c.vectmp,c.vectmp2,c.vectmp3)
 
-type Rosenbrock32Cache{uType,uArrayType,rateType,du2Type,vecuType,JType,TabType,TFType,UFType,F,JCType} <: OrdinaryDiffEqMutableCache
+type Rosenbrock32Cache{uType,uArrayType,rateType,du2Type,LinuType,vecuType,JType,TabType,TFType,UFType,F,JCType} <: OrdinaryDiffEqMutableCache
   u::uType
   uprev::uType
   k₁::rateType
@@ -977,7 +978,8 @@ type Rosenbrock32Cache{uType,uArrayType,rateType,du2Type,vecuType,JType,TabType,
   tab::TabType
   tf::TFType
   uf::UFType
-  linsolve_tmp::vecuType
+  linsolve_tmp::LinuType
+  linsolve_tmp_vec::vecuType
   linsolve::F
   jac_config::JCType
 end
@@ -1009,11 +1011,12 @@ function alg_cache(alg::Rosenbrock23,u,rate_prototype,uEltypeNoUnits,tTypeNoUnit
   vfr = VectorFReturn(f,size(u))
   tf = TimeGradientWrapper(vf,uprev)
   uf = UJacobianWrapper(vfr,t)
-  linsolve_tmp = vec(similar(u,indices(u)))
+  linsolve_tmp = similar(u,indices(u))
+  linsolve_tmp_vec = vec(linsolve_tmp)
   jac_config = ForwardDiff.JacobianConfig(uf,vec(du1),vec(uprev),ForwardDiff.Chunk{determine_chunksize(u,alg)}())
   Rosenbrock23Cache(u,uprev,k₁,k₂,k₃,du1,du2,f₁,vectmp,vectmp2,vectmp3,fsalfirst,
-                    fsallast,dT,J,W,tmp,tab,tf,uf,linsolve_tmp,alg.linsolve,
-                    jac_config)
+                    fsallast,dT,J,W,tmp,tab,tf,uf,linsolve_tmp,linsolve_tmp_vec,
+                    alg.linsolve,jac_config)
 end
 
 function alg_cache(alg::Rosenbrock32,u,rate_prototype,uEltypeNoUnits,tTypeNoUnits,uprev,uprev2,f,t,::Type{Val{true}})
@@ -1037,9 +1040,10 @@ function alg_cache(alg::Rosenbrock32,u,rate_prototype,uEltypeNoUnits,tTypeNoUnit
   vfr = VectorFReturn(f,size(u))
   tf = TimeGradientWrapper(vf,uprev)
   uf = UJacobianWrapper(vfr,t)
-  linsolve_tmp = vec(similar(u,indices(u)))
+  linsolve_tmp = similar(u,indices(u))
+  linsolve_tmp_vec = vec(linsolve_tmp)
   jac_config = ForwardDiff.JacobianConfig(uf,vec(du1),vec(uprev),ForwardDiff.Chunk{determine_chunksize(u,alg)}())
-  Rosenbrock32Cache(u,uprev,k₁,k₂,k₃,du1,du2,f₁,vectmp,vectmp2,vectmp3,fsalfirst,fsallast,dT,J,W,tmp,tab,tf,uf,linsolve_tmp,alg.linsolve,jac_config)
+  Rosenbrock32Cache(u,uprev,k₁,k₂,k₃,du1,du2,f₁,vectmp,vectmp2,vectmp3,fsalfirst,fsallast,dT,J,W,tmp,tab,tf,uf,linsolve_tmp,linsolve_tmp_vec,alg.linsolve,jac_config)
 end
 
 immutable Rosenbrock23ConstantCache{T,TF,UF} <: OrdinaryDiffEqConstantCache
@@ -1100,15 +1104,16 @@ vecu_cache(c::ImplicitEulerCache) = (c.uhold,)
 dual_cache(c::ImplicitEulerCache) = (c.dual_cache,)
 
 function alg_cache(alg::ImplicitEuler,u,rate_prototype,uEltypeNoUnits,tTypeNoUnits,uprev,uprev2,f,t,::Type{Val{true}})
-  u_old = similar(u,indices(u)); k = zeros(rate_prototype)
+  tmp = similar(u)
+  u_old = vec(tmp); k = zeros(rate_prototype)
   dual_cache = DiffCache(u,Val{determine_chunksize(u,get_chunksize(alg.nlsolve))})
   uhold = vec(u) # this makes uhold the same values as integrator.u
   rhs = RHS_IE(f,u_old,t,t,dual_cache,size(u),eachindex(u))
   fsalfirst = zeros(rate_prototype)
   nl_rhs = alg.nlsolve(Val{:init},rhs,uhold)
-  tmp = u_old
-  ImplicitEulerCache{typeof(u),typeof(u_old),typeof(uhold),typeof(dual_cache),typeof(k),
-                     typeof(rhs),typeof(nl_rhs)}(
+
+  ImplicitEulerCache{typeof(u),typeof(u_old),typeof(uhold),typeof(dual_cache),
+                     typeof(k),typeof(rhs),typeof(nl_rhs)}(
                      u,uprev,uprev2,uhold,dual_cache,u_old,tmp,k,fsalfirst,rhs,nl_rhs)
 end
 
@@ -1147,12 +1152,13 @@ vecu_cache(c::TrapezoidCache) = (c.uhold,)
 dual_cache(c::TrapezoidCache) = (c.dual_cache,)
 
 function alg_cache(alg::Trapezoid,u,rate_prototype,uEltypeNoUnits,tTypeNoUnits,uprev,uprev2,f,t,::Type{Val{true}})
-  u_old = similar(u,indices(u)); k = zeros(rate_prototype)
+  tmp = similar(u)
+  u_old = vec(tmp); k = zeros(rate_prototype)
   uhold = vec(u); fsalfirst = zeros(rate_prototype)
+  f_old = vec(fsalfirst)
   dual_cache = DiffCache(u,Val{determine_chunksize(u,get_chunksize(alg.nlsolve))})
-  rhs = RHS_Trap(f,u_old,fsalfirst,t,t,size(u),dual_cache,eachindex(u))
+  rhs = RHS_Trap(f,u_old,f_old,t,t,size(u),dual_cache,eachindex(u))
   nl_rhs = alg.nlsolve(Val{:init},rhs,uhold)
-  tmp = u_old
   TrapezoidCache{typeof(u),typeof(u_old),typeof(uhold),typeof(dual_cache),typeof(k),
     typeof(rhs),typeof(nl_rhs)}(u,uprev,uprev2,uhold,u_old,fsalfirst,dual_cache,tmp,k,rhs,nl_rhs)
 end
@@ -1206,7 +1212,6 @@ function alg_cache(alg::IIF1,u,rate_prototype,uEltypeNoUnits,tTypeNoUnits,uprev,
 end
 
 function alg_cache(alg::IIF1,u,rate_prototype,uEltypeNoUnits,tTypeNoUnits,uprev,uprev2,f,t,::Type{Val{true}})
-
   tmp = similar(u,indices(u)); rtmp1 = zeros(rate_prototype)
   dual_cache = DiffCache(u,Val{determine_chunksize(u,get_chunksize(alg.nlsolve))})
   uhold = vec(u) # this makes uhold the same values as integrator.u