From 012c61878fb8c36990af6a34cadea01acf3b8880 Mon Sep 17 00:00:00 2001
From: Nathan Daly <nathan.daly@relational.ai>
Date: Fri, 31 Jan 2025 10:13:54 -0700
Subject: [PATCH] FINALLY add a benchmark script. We should've had this from
 the beginning. (#39)

* FINALLY add a benchmark script. We should've had this from the beginning.

This is the status as of now:
```julia
julia> include("bench/runbench.jl")
  11.000 ns (0 allocations: 0 bytes)
  61.641 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  15.489 ns (0 allocations: 0 bytes)
  4.000 ns (0 allocations: 0 bytes)
  26.104 ns (0 allocations: 0 bytes)
  2.250 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
  2.250 ns (0 allocations: 0 bytes)
  2.250 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
0
```

(And here is the above benchmark run on v0.5.1 - when everything
used `@generated` functions. You can see that we have recovered all of
that performance.):
```julia
julia> include("bench/runbench.jl")
  10.844 ns (0 allocations: 0 bytes)
  62.855 ns (0 allocations: 0 bytes)
  3.041 ns (0 allocations: 0 bytes)
  16.074 ns (0 allocations: 0 bytes)
  4.416 ns (0 allocations: 0 bytes)
  26.230 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.834 ns (0 allocations: 0 bytes)
  2.916 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
0
```

* Add compile time benchmarks
---
 bench/runbench.jl | 69 +++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 69 insertions(+)
 create mode 100644 bench/runbench.jl

diff --git a/bench/runbench.jl b/bench/runbench.jl
new file mode 100644
index 0000000..1ebc164
--- /dev/null
+++ b/bench/runbench.jl
@@ -0,0 +1,69 @@
+using Blobs, BenchmarkTools
+
+@info "------- Benchmark Runtimes ---------"
+
+@eval struct MixedStruct10
+    $((:($(Symbol("f$i"))::Float64) for i in 1:5)...)
+    $((:($(Symbol("x$i"))::Int64) for i in 1:5)...)
+end
+m10 = Blobs.malloc(MixedStruct10)
+
+@eval struct MixedStruct100
+    $((:($(Symbol("f$i"))::Float64) for i in 1:50)...)
+    $((:($(Symbol("x$i"))::Int64) for i in 1:50)...)
+end
+m100 = Blobs.malloc(MixedStruct100)
+
+@btime Blobs.malloc(MixedStruct10)
+@btime Blobs.malloc(MixedStruct100)
+
+@btime $(m10)[]
+@btime $(m100)[]
+
+@btime $(m10)[] = $(m10[])
+@btime $(m100)[] = $(m100[])
+
+@btime $(m10).x5[]
+@btime $(m100).x5[]
+@btime $(m100).x50[]
+
+@btime $(m10).f5[] = 0
+@btime $(m10).x5[] = 0
+@btime $(m100).f5[] = 0
+@btime $(m100).x5[] = 0
+@btime $(m100).x50[] = 0
+
+
+@info "------- Benchmark Compile Times ---------"
+
+# Benchmark Compile Times:
+function eval_type(N)
+    S = gensym("MixedType$N")
+    @eval struct $S
+        $((:($(Symbol("f$i"))::Float64) for i in 1:(N÷2) )...)
+        $((:($(Symbol("x$i"))::Int64) for i in 1:(N÷2) )...)
+    end
+    return @eval($S)
+end
+
+for N in (10,100)
+    @info " --  N = $N -- "
+
+    @btime Blobs.malloc(S) setup=(S=eval_type($N))
+
+    @btime (Blobs.malloc(S))[] setup=(S=eval_type($N))
+
+    @btime (Blobs.malloc(S))[] = (Blobs.malloc(S)[]) setup=(S=eval_type($N))
+
+    @btime (Blobs.malloc(S)).x5[] setup=(S=eval_type($N))
+    if N >= 100
+        @btime (Blobs.malloc(S)).x50[] setup=(S=eval_type($N))
+    end
+
+    @btime (Blobs.malloc(S)).f5[] = 0 setup=(S=eval_type($N))
+    @btime (Blobs.malloc(S)).x5[] = 0 setup=(S=eval_type($N))
+    if N >= 100
+        @btime (Blobs.malloc(S)).x50[] = 0 setup=(S=eval_type($N))
+    end
+
+end