add benchmark (test)

benchmark/benchmarks.jl

@@ -0,0 +1,43 @@
+using YaoArrayRegister, YaoBlocks, YaoBlocks.ConstGate, Random
+using BenchmarkTools, PkgBenchmark
+
+function bench(block)
+    N = nqubits(block)
+    r = rand_state(N)
+    return @benchmarkable apply!($r, $block)
+end
+
+const SUITE = BenchmarkGroup()
+
+SUITE["primitive"] = BenchmarkGroup()
+SUITE["composite"] = BenchmarkGroup()
+
+for G in [:X, :Y, :Z, :S, :T, :Sdag, :Tdag]
+    GT = Expr(:(.), :ConstGate, QuoteNode(Symbol(G, :Gate)))
+    GN = Expr(:(.), :ConstGate, QuoteNode(G))
+
+    SUITE["primitive"][string(G)] = BenchmarkGroup()
+    for n in 1:4:25
+        SUITE["primitive"][string(G)][n] = @eval bench(repeat($n, $GN))
+    end
+end
+
+for n in 5:5:25
+    SUITE["composite"]["kron(rand_const)"] = bench(kron(rand([X, Y, Z, H]) for _ in 1:n))
+    SUITE["composite"]["kron(sparse_const)"] = bench(kron(n, k=>rand([X, Y, Z, H]) for k in randperm(n)[1:n÷5]))
+end
+
+function heisenberg(n::Int; periodic::Bool=true)
+    Sx(i) = put(n, i=>X)
+    Sy(i) = put(n, i=>Y)
+    Sz(i) = put(n, i=>Z)
+
+    return sum(1:(periodic ? n : n-1)) do i
+        j = mod1(i, n)
+        Sx(i) * Sx(j) + Sy(i) * Sy(j) + Sz(i) * Sz(j)
+    end
+end
+
+for n in 5:5:25
+    SUITE["primitive"]["TimeEvolution(heisenberg(n), 0.2)"] = bench(TimeEvolution(heisenberg(n), 0.2))
+end

benchmark/tune.json

@@ -0,0 +1 @@
+[{"Julia":"1.1.0","BenchmarkTools":"0.2.2"},[["BenchmarkGroup",{"data":{"composite":["BenchmarkGroup",{"data":{"kron(sparse_const)":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"kron(rand_const)":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}]},"tags":[]}],"primitive":["BenchmarkGroup",{"data":{"Y":["BenchmarkGroup",{"data":{"9":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"13":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"25":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"17":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"5":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":10,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"21":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"1":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":274,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}]},"tags":[]}],"Z":["BenchmarkGroup",{"data":{"9":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":7,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"13":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"25":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"17":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"5":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":10,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"21":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"1":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":262,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}]},"tags":[]}],"S":["BenchmarkGroup",{"data":{"9":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":6,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"13":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"25":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"17":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"5":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":10,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"21":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"1":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":259,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}]},"tags":[]}],"Sdag":["BenchmarkGroup",{"data":{"9":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":7,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"13":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"25":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"17":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"5":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":10,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"21":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"1":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":261,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}]},"tags":[]}],"X":["BenchmarkGroup",{"data":{"9":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":5,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"13":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"25":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"17":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"5":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":10,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"21":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"1":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":282,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}]},"tags":[]}],"T":["BenchmarkGroup",{"data":{"9":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":7,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"13":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"25":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"17":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"5":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":10,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"21":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"1":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":254,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}]},"tags":[]}],"TimeEvolution(heisenberg(n), 0.2)":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"Tdag":["BenchmarkGroup",{"data":{"9":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":6,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"13":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"25":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"17":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"5":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":10,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"21":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":1,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}],"1":["BenchmarkTools.Parameters",{"gctrial":true,"time_tolerance":0.05,"samples":10000,"evals":276,"gcsample":false,"seconds":5.0,"overhead":0.0,"memory_tolerance":0.01}]},"tags":[]}]},"tags":[]}]},"tags":[]}]]]

src/algebra.jl

@@ -132,9 +132,9 @@ export eliminate_nested
 eliminate_nested(ex::AbstractBlock) = ex
 
 # TODO: eliminate nested expr e.g chain(X, chain(X, Y))
-function eliminate_nested(ex::Union{ChainBlock, Sum})
+function eliminate_nested(ex::T) where {T <: Union{ChainBlock, Sum}}
     _flatten(x) = (x, )
-    _flatten(x::ChainBlock) = subblocks(x)
+    _flatten(x::T) = subblocks(x)
 
     isone(length(ex)) && return first(subblocks(ex))
     return chsubblocks(ex, Iterators.flatten(map(_flatten, subblocks(ex))))

src/primitive/time_evolution.jl

@@ -10,16 +10,15 @@ TimeEvolution, where GT is block type. input matrix should be hermitian.
 !!!note:
     `TimeEvolution` contructor check hermicity of the input block by default, but sometimes it can be slow. Turn off the check manually by specifying optional parameter `check_hermicity = false`.
 """
-mutable struct TimeEvolution{N, T, Tt, Hamilton <: AbstractBlock{N}} <: PrimitiveBlock{N}
-    H::BlockMap{Complex{T}, Hamilton}
+mutable struct TimeEvolution{N, T, Tt, Hamilton <: AbstractMatrix} <: PrimitiveBlock{N}
+    H::Hamilton
     dt::Tt
     tol::T
 
     function TimeEvolution(
-        H::BlockMap{Complex{T}, TH},
-        dt::Tt, tol::T; check_hermicity::Bool=true) where {N, Tt, T, TH <: AbstractBlock{N}}
-        (check_hermicity && !ishermitian(H)) && error("Time evolution Hamiltonian has to be a Hermitian")
-        return new{N, T, Tt, TH}(H, dt, tol)
+        H::TH,
+        dt::Tt, tol::T) where {Tt, T, TH <: AbstractMatrix}
+        return new{log2dim1(H), T, Tt, TH}(H, dt, tol)
     end
 end
 
@@ -33,20 +32,14 @@ Optional keywords are tolerance `tol` (default is `1e-7`)
 `TimeEvolution` block can also be used for
 [imaginary time evolution](http://large.stanford.edu/courses/2008/ph372/behroozi2/) if dt is complex.
 """
-TimeEvolution(H::AbstractBlock, dt; tol::Real=1e-7, check_hermicity=true) =
-    TimeEvolution(BlockMap(H), dt, tol, check_hermicity=check_hermicity)
+TimeEvolution(H::AbstractBlock, dt; tol::Real=1e-7) = TimeEvolution(mat(H), dt, tol)
+TimeEvolution(M::AbstractMatrix, dt; tol::Real) = TimeEvolution(M, dt, tol)
 
-TimeEvolution(M::BlockMap, dt; tol::Real, check_hermicity=true) =
-    TimeEvolution(M, dt, tol, check_hermicity=check_hermicity)
-
-time_evolve(M::BlockMap, dt; kwargs...) = TimeEvolution(M, dt; kwargs...)
+time_evolve(M::AbstractMatrix, dt; kwargs...) = TimeEvolution(M, dt; kwargs...)
 time_evolve(M::AbstractBlock, dt; kwargs...) = TimeEvolution(M, dt; kwargs...)
 time_evolve(dt; kwargs...) = @λ(M->time_evolve(M, dt; kwargs...))
 
-function mat(::Type{T}, te::TimeEvolution{N}) where {T, N}
-    A = Matrix{T}(te.H.block)
-    return exp(-im*T(te.dt) * A)
-end
+mat(::Type{T}, te::TimeEvolution{N}) where {T, N} = exp(-im*T(te.dt) * Matrix{T}(te.H))
 
 function apply!(reg::ArrayReg, te::TimeEvolution)
     st = state(reg)