Some optimization

src/ADerrorsCF.jl

@@ -196,8 +196,9 @@ function uwerror(a::uwreal, ws::wspace, wpm::Dict{Int64,Vector{Float64}})
     nid = unique_ids!(a, ws)
 
     if (length(a.cfd) == 0)
+        a.cfd = Vector{cfdata}(undef, nid)
         for j in 1:nid
-            new = cfdata(0.0, 0.0, 0.0, 0, Vector{Float64}(), Vector{Float64}())
+            a.cfd[j] = cfdata()
             idx = ws.map_ids[a.ids[j]]
             nd  = ws.fluc[idx].nd
             if (nd != 1)
@@ -215,7 +216,7 @@ function uwerror(a::uwreal, ws::wspace, wpm::Dict{Int64,Vector{Float64}})
             for i in 1:length(a.prop)
                 if (a.prop[i] && (ws.map_nob[i] == a.ids[j]))
                     if (nd == 1)
-                        new.var = new.var + a.der[i]*ws.fluc[i].delta[1]
+                        a.cfd[j].var = a.cfd[j].var + a.der[i]*ws.fluc[i].delta[1]
                     else
                         for k in 1:nrep
                             ftemp[k] = ftemp[k] + a.der[i]*ws.fluc[i].fourier[k]
@@ -225,56 +226,60 @@ function uwerror(a::uwreal, ws::wspace, wpm::Dict{Int64,Vector{Float64}})
             end
 
             if (nd == 1)
-                new.var = new.var^2
+                a.cfd[j].var = a.cfd[j].var^2
             else
-                new.gamm = zeros(nt)
+                a.cfd[j].gamm = zeros(nt)
                 for k in 1:nrep
                     ftemp[k] .= ftemp[k].*conj(ftemp[k])
                     FFTW.ifft!(ftemp[k])
 
                     for ig in 1:min(nt,length(ftemp[k]))
-                        new.gamm[ig] = new.gamm[ig] + real(ftemp[k][ig])
+                        a.cfd[j].gamm[ig] = a.cfd[j].gamm[ig] + real(ftemp[k][ig])
                     end
                 end
                 for ig in 1:nt
-                    nrcnt = count(map(x -> div(x, 2*ws.fluc[idx].ibn), ws.fluc[idx].ivrep) .> ig-1)
-                    new.gamm[ig] = new.gamm[ig] / (nd_eff - nrcnt*(ig-1))
+                    nrcnt = 0
+                    for k in 1:nrep
+                        if (div(ws.fluc[idx].ivrep[k], 2*ws.fluc[idx].ibn) > ig-1 )
+                            nrcnt = nrcnt + 1
+                        end
+                    end
+                    a.cfd[j].gamm[ig] = a.cfd[j].gamm[ig] / (nd_eff - nrcnt*(ig-1))
                 end
 
-                iw = wopt_ulli(nd_eff, DEFAULT_STAU, new.gamm)
-                new.iw = iw
-                
-                dbias = new.gamm[1] + 2.0*sum(new.gamm[2:iw])
-                new.gamm .= new.gamm .+ dbias/nd_eff
+                iw = wopt_ulli(nd_eff, DEFAULT_STAU, a.cfd[j].gamm)
+                a.cfd[j].iw = iw
+
+                gwin  = view(a.cfd[j].gamm, 2:iw)
+                dbias = a.cfd[j].gamm[1] + 2.0*sum(gwin)
+                a.cfd[j].gamm .= a.cfd[j].gamm .+ dbias/nd_eff
 
-                new.drho = zeros(nt)
-                if (new.gamm[1] != 0.0)
+                a.cfd[j].drho = zeros(nt)
+                if (a.cfd[j].gamm[1] != 0.0)
                     for i in 1:nt
                         is = max(1, i-iw-2) + 1
                         ie = i + iw-1
                         for k in is:ie
                             if (k < nt+1)
-                                cont = -2.0*new.gamm[k]*new.gamm[i]/new.gamm[1]^2
+                                cont = -2.0*a.cfd[j].gamm[k]*a.cfd[j].gamm[i]/a.cfd[j].gamm[1]^2
                             else
                                 cont = 0.0
                             end
 
                             if ((i+k-2) < nt)
-                                cont = cont + new.gamm[i+k-1]/new.gamm[1]
+                                cont = cont + a.cfd[j].gamm[i+k-1]/a.cfd[j].gamm[1]
                             end
                             if (abs(i-k) < nt)
-                                cont = cont + new.gamm[abs(i-k)+1]/new.gamm[1]
+                                cont = cont + a.cfd[j].gamm[abs(i-k)+1]/a.cfd[j].gamm[1]
                             end
-                            new.drho[i] = new.drho[i] + cont^2
+                            a.cfd[j].drho[i] = a.cfd[j].drho[i] + cont^2
                         end
-                        new.drho[i] = sqrt(new.drho[i]/nd_eff)
+                        a.cfd[j].drho[i] = sqrt(a.cfd[j].drho[i]/nd_eff)
                     end
                 else
-                    new.var = new.var^2
+                    a.cfd[j].var = a.cfd[j].var^2
                 end
             end
-
-            push!(a.cfd, new)
         end
     end
 
@@ -323,7 +328,8 @@ function uwerror(a::uwreal, ws::wspace, wpm::Dict{Int64,Vector{Float64}})
                 end
                 iw = a.cfd[j].iw
                 
-                vti = 0.5 + sum(a.cfd[j].gamm[2:iw])/a.cfd[j].gamm[1]
+                gwin  = view(a.cfd[j].gamm, 2:iw)
+                vti = 0.5 + sum(gwin)/a.cfd[j].gamm[1]
                 a.cfd[j].dtaui = sqrt(vti^2 * (4.0*iw-2.0*vti+2.0)/nd_eff)
                 a.cfd[j].taui  = vti + texp*a.cfd[j].gamm[iw+1]/a.cfd[j].gamm[1]
             end

src/ADerrorsIO.jl

@@ -201,7 +201,8 @@ function details(a::uwreal, ws::wspace, io::IO=stdout, names::Dict{Int64, String
 
         n = 0
         for i in 1:length(a.cfd)
-            if (length(a.cfd[i].gamm) > 0)
+            idx  = ws.map_ids[a.ids[i]]
+            if (ws.fluc[idx].nd > 1)
                 n = n + 1
             end
         end
@@ -217,10 +218,10 @@ function details(a::uwreal, ws::wspace, io::IO=stdout, names::Dict{Int64, String
         end
         ip = sortperm(v, rev=true)
         for i in 1:length(a.cfd)
+            idx  = ws.map_ids[a.ids[ip[i]]]
+            nd   = ws.fluc[idx].nd
             sid = truncate_ascii(get(names, a.ids[ip[i]], string(a.ids[ip[i]])), ntrunc)
-            if (length(a.cfd[ip[i]].gamm) > 0)
-                idx  = ws.map_ids[a.ids[i]]
-                nd   = ws.fluc[idx].nd
+            if (nd > 1)
                 Printf.@printf("  #  %45s %6.2f   %10d\n",
                         sid, 100.0 .* a.cfd[ip[i]].var ./ a.err^2, nd)
             else

src/ADerrorsTools.jl

@@ -14,13 +14,12 @@ function cobs(avgs::Vector{Float64}, cov::Array{Float64, 2}, ids::Vector{Int64})
     ch = LinearAlgebra.cholesky(cov)
 
     n = length(avgs)
-    p = Vector{uwreal}()
+    p = Vector{uwreal}(undef, n)
     for j in 1:n
-        new = uwreal([avgs[j], ch.L[j, 1]], ids[1])
+        p[j] = uwreal([avgs[j], ch.L[j, 1]], ids[1])
         for i in 2:n
-            new = new + uwreal([0.0, ch.L[j,i]], ids[i])
+            p[j] = p[j] + uwreal([0.0, ch.L[j,i]], ids[i])
         end
-        push!(p, new)
     end
     return p
 end
@@ -55,7 +54,7 @@ function addobs(a::Vector{uwreal}, der::Array{Float64, 2}, mvl::Vector{Float64})
         n = max(n,length(a[i].der))
     end
 
-    x = Vector{uwreal}()
+    x = Vector{uwreal}(undef, size(der, 1))
     for i in 1:size(der, 1)
         p = [false for i in 1:n]
         d = zeros(Float64, n)
@@ -69,7 +68,7 @@ function addobs(a::Vector{uwreal}, der::Array{Float64, 2}, mvl::Vector{Float64})
                 end
             end
         end
-        push!(x, uwreal(mvl[i], p, d))
+        x[i] = uwreal(mvl[i], p, d)
     end
     
     return x

src/ADerrorsTypes.jl

@@ -10,6 +10,13 @@
 ###                               
 
 
+mutable struct hyperd
+    v::Float64
+    d1::Float64
+    d2::Float64
+    d3::Float64
+end
+       
 mutable struct cfdata
     var::Float64
     taui::Float64
@@ -19,6 +26,15 @@ mutable struct cfdata
     
     gamm::Vector{Float64}
     drho::Vector{Float64}
+
+    function cfdata()
+        x = new()
+        x.var   = 0.0
+        x.taui  = 0.0
+        x.dtaui = 0.0
+        x.iw = 0
+        return x
+    end
 end
 
 mutable struct uwreal

src/ADerrorsUtils.jl

@@ -108,17 +108,24 @@ function fit_error(chisq::Function,
 
     n = length(xp)   # Number of fit parameters
     m = length(data) # Number of data
-    ccsq(x::Vector) = chisq(x[1:n], x[n+1:end])
 
-    xav = zeros(Float64, n+m)
+    xav = Vector{Float64}(undef, n+m)
     for i in 1:n
         xav[i] = xp[i]
     end
     for i in n+1:n+m
         xav[i] = data[i-n].mean
     end
-    hess = ForwardDiff.hessian(ccsq, xav)
 
+    function cls(x0)
+        x1 = view(x0, 1:n)
+        x2 = view(x0, n+1:n+m)
+        return chisq(x1, x2)
+    end
+
+    hess = Array{Float64}(undef, n+m, n+m)
+    ForwardDiff.hessian!(hess, cls, xav)
+    
     hinv = LinearAlgebra.pinv(hess[1:n,1:n])
     grad = - hinv[1:n,1:n] * hess[1:n,n+1:n+m]
     
@@ -144,7 +151,7 @@ function fit_error(chisq::Function,
         else
             Ww = W
         end
-
+        
         cse = chiexp(hess, data, Ww)        
     end