add more regression benchmark tests (#233)

* add more regression benchmark tests

* repeat char does not exist in 0.6

src/broadcast/BroadcastBenchmarks.jl

@@ -82,4 +82,22 @@ for t in (t1, t2, t3)
     g[length(t), "scal_tup_x3"] = @benchmarkable broadcast(+, 1, $t, 1, $t, 1, $t)
 end
 
+###########################################################################
+
+xset= 10:15
+yset= 12:14
+f(x,y) = sqrt( x^2 + y^2 )
+
+function perf_loop3(xset, yset)
+    m = Matrix{Float64}(undef, 3, length(xset)*length(yset))
+    i = 1
+    @inbounds for x in xset, y in yset
+        m[:,i] .= (Float64(x), Float64(y), f(x,y))
+        i += 1
+    end
+    return m
+end
+
+SUITE["26942"] = @benchmarkable perf_loop3($xset, $yset)
+
 end # module

src/linalg/LinAlgBenchmarks.jl

@@ -262,4 +262,11 @@ for b in values(g)
     b.params.time_tolerance = 0.40
 end
 
+#############
+# small exp #
+#############
+
+SUITE["small exp #29116"] = @benchmarkable exp([1. 0; 2 0])
+
+
 end # module

src/misc/MiscellaneousBenchmarks.jl

@@ -305,4 +305,89 @@ for N in (1,1000), M in 1:4
     g["zip($(join(fill("1:$N", M), ", ")))"] = @benchmarkable collect($X)
 end
 
+####################################################
+# Allocation elision stumped by conditional #28226 #
+# Note, not fixed when this benchmark was written  #
+####################################################
+
+function perf_colwise_alloc!(r, a, b)
+    @inbounds for j = 1:size(a,2)
+        r[j] = evaluate_cond(view(a, :, j), view(b, :, j))
+    end
+    r
+end
+
+@inline function evaluate_cond(a, b)
+    length(a) == 0 && return 0.0 # comment out and 0.7 is super fast
+    @inbounds begin
+        s = 0.0
+        @simd for I in eachindex(a, b)
+            ai = a[I]
+            bi = b[I]
+            s += abs2(ai - bi)
+        end
+        return s
+    end
+end
+
+function perf_colwise_noalloc!(r, a, b)
+    @inbounds for j = 1:size(a,2)
+        r[j] = evaluate_nocond(view(a, :, j), view(b, :, j))
+    end
+    r
+end
+
+@inline function evaluate_nocond(a, b)
+    @inbounds begin
+        s = 0.0
+        @simd for I in eachindex(a, b)
+            ai = a[I]
+            bi = b[I]
+            s += abs2(ai - bi)
+        end
+        return s
+    end
+end
+
+z = zeros(41); A = rand(2, 41); B = rand(2, 41);
+g = addgroup!(SUITE, "allocation elision view")
+g["conditional"] = @benchmarkable perf_colwise_alloc!($z, $A, $B)
+g["no conditional"] = @benchmarkable perf_colwise_noalloc!($z, $A, $B)
+
+
+####################################################
+# Fastmath infererence large number of args #22275 #
+####################################################
+
+function f2(a,b,c,d,e,f,g,h,j,k,l,m,n,o,p)
+    aidx = eachindex(a)
+    @fastmath for i in aidx
+        @inbounds a[i] = b[i]+c*(d*e[i]+f*g[i]+h*j[i]+k*l[i]+m*n[i]+o*p[i])
+    end
+end
+a = rand(10)
+b = rand(10)
+c = 0.1
+d = 0.1
+e = rand(10)
+f = 0.1
+g = rand(10)
+h = 0.1
+j = rand(10)
+k = 0.1
+l = rand(10)
+m = 0.1
+n = rand(10)
+o = 0.1
+p = rand(10)
+
+SUITE["fastmath many args"] = @benchmarkable f2($a,$b,$c,$d,$e,$f,$g,$h,$j,$k,$l,$m,$n,$o,$p)
+
+##############################################################
+# Performance and typing of 6+ dimensional generators #21058 #
+##############################################################
+perf_g6() = sum([+(a,b,c,d,e,f) for a in 1:4, b in 1:4, c in 1:4, d in 1:4, e in 1:4, f in 1:4])
+
+SUITE["perf highdim generator"] = @benchmarkable perf_g6()
+
 end

src/sparse/SparseBenchmarks.jl

@@ -258,7 +258,7 @@ for (om, ok, on) in (# order of matmul dimensions m, k, and n
         g["A_mul_Bt",  "dense $(m)x$(k), sparse $(n)x$(k) -> dense $(m)x$(n)"] = @benchmarkable A_mul_Bt($densemat, $tsparsemat)
         g["At_mul_B",  "dense $(k)x$(m), sparse $(k)x$(n) -> dense $(m)x$(n)"] = @benchmarkable At_mul_B($tdensemat, $sparsemat)
         g["At_mul_Bt", "dense $(k)x$(m), sparse $(n)x$(k) -> dense $(m)x$(n)"] = @benchmarkable At_mul_Bt($tdensemat, $tsparsemat)
-    end    
+    end
     # in-place dense-sparse -> dense ops, transpose variants, i.e. A[t]_mul[t]!(dense, dense, sparse)
     m, k, n, destmat, densemat, sparsemat, tdensemat, tsparsemat = allocmats_ds(om, ok, on, 4, 12, Float64)
     if VERSION >= v"0.7.0-DEV.3204"
@@ -282,7 +282,7 @@ for (om, ok, on) in (# order of matmul dimensions m, k, and n
         g["A_mul_Bc",  "dense $(m)x$(k), sparse $(n)x$(k) -> dense $(m)x$(n)"] = @benchmarkable A_mul_Bc($densemat, $tsparsemat)
         g["Ac_mul_B",  "dense $(k)x$(m), sparse $(k)x$(n) -> dense $(m)x$(n)"] = @benchmarkable Ac_mul_B($tdensemat, $sparsemat)
         g["Ac_mul_Bc", "dense $(k)x$(m), sparse $(n)x$(k) -> dense $(m)x$(n)"] = @benchmarkable Ac_mul_Bc($tdensemat, $tsparsemat)
-    end        
+    end
     # in-place dense-sparse -> dense ops, adjoint variants, i.e. A[c]_mul[c]!(dense, dense, sparse)
     m, k, n, destmat, densemat, sparsemat, tdensemat, tsparsemat = allocmats_ds(om, ok, on, 2, 8, Complex{Float64})
     if VERSION >= v"0.7.0-DEV.3204"
@@ -355,4 +355,19 @@ for b in values(g)
     b.params.time_tolerance = 0.3
 end
 
+
+#################
+# sparse matvec #
+#################
+g = addgroup!(SUITE, "sparse matvec")
+B = randn(100000, 100)
+A = sprand(100000, 100000, 0.00001)
+if VERSION >= v"0.7.0-DEV.3204"
+    g["non-adjoint"] = @benchmarkable A * B
+    g["adjoint"] = @benchmarkable A' * B
+else
+    g["non-adjoint"] = @benchmarkable A * B
+    g["adjoint"] = @benchmarkable At_mul_B(A, B)
+end
+
 end # module

src/string/StringBenchmarks.jl

@@ -62,4 +62,16 @@ buffer = IOBuffer(("bar" ^ 20000) * "ians")
 target = ("bar" ^ 300) * "ian"
 g["barbarian backtrack"] = @benchmarkable readuntil(seekstart($buffer), $target)
 
+#################
+# repeat #22462 #
+#################
+
+g = addgroup!(SUITE, "repeat")
+g["repeat str len 1"] = @benchmarkable repeat(" ", 500)
+g["repeat str len 16"] = @benchmarkable repeat("repeatmerepeatme", 500)
+if VERSION >= v"v0.7"
+    g["repeat char 1"] = @benchmarkable repeat(' ', 500)
+    g["repeat char 2"] = @benchmarkable repeat('α', 500)
+end
+
 end # module