Final performance optimisations for bhmie

test/Project.toml

@@ -1,4 +0,0 @@
-[deps]
-AirspeedVelocity = "1c8270ee-6884-45cc-9545-60fa71ec23e4"
-BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
-Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"

test/benchmarks.jl

@@ -3,15 +3,16 @@ module Benchmarks
 include("../anchors.jl")
 include("ffi_wraps.jl")
 
-import .Anchors.ROOT_DIR
+import .Anchors.SRC_DIR
 import .FFIWraps: bhmie_c, bhmie_fortran, bhmie_fortran77
 using BenchmarkTools
+using InteractiveUtils #! DEBUG
 
-include("$ROOT_DIR/src/miemfp.jl")
+include("$SRC_DIR/miemfp.jl")
 
 function bench_vs_ffi()
     # Fixed testing values
-    nang = UInt32(2)  # Example number of angles
+    nang = 2  # Example number of angles
 
     c_result = @benchmark bhmie_c(x, cxref, nang, cxs1, cxs2) setup=(
         x = rand(Float32);
@@ -36,11 +37,11 @@ function bench_vs_ffi()
         cxs1 = rand(ComplexF32, $nang);
         cxs2 = rand(ComplexF32, $nang);
     )
-
-    j_result = @benchmark miemfp.bhmie(Float64(x), ComplexF64(cxref), nang) setup=(
+    
+    j_result = @benchmark miemfp.bhmie(Float64(x), ComplexF64(cxref), Int64(nang)) setup=(
         x = rand(Float32);
         cxref = rand(ComplexF32);
-        nang = UInt32($nang);
+        nang = $nang;
     )
 
     println("\nC Implementation")

test/miemfp_tests.jl

@@ -1,7 +1,6 @@
 using Test
 using Random
 using PropCheck
-using Debugger
 using PyCall
 
 if !@isdefined TestUtils
@@ -25,7 +24,7 @@ miemfp.bhmie(
     nang::Int64,
     s1::Vector{ComplexF64},
     s2::Vector{ComplexF64},
-) = miemfp.bhmie(x, cxref, UInt32(nang))
+) = miemfp.bhmie(x, cxref, nang)
 
 function miemfp.bhmie(
     x::Float64,

test/miemfp_tests.py

@@ -22,9 +22,9 @@ def compare_bhmie_functions(
         # This is to ensure that only one instance of each function is running at a time
         # to avoid memory issues in the FFI code
         await event1.wait()
-        f1_result = f1(x, cxref, 2, cxs1, cxs2)[:2]
+        f1_result = f1(x, cxref, 2, cxs1, cxs2)[:2]  # Only testing at nang = 2 to avoid memory issues
         await event2.wait()
-        f2_result = f2(x, cxref, 2, cxs1, cxs2)[:2]
+        f2_result = f2(x, cxref, 2, cxs1, cxs2)[:2]  # Only testing at nang = 2 to avoid memory issues
 
         return np.all(np.isclose(f1_result, f2_result))