benchmarking: added gpu evaluators to performance testing. getting execution errors still

package/src/ExpressionExecutorCuda.jl

@@ -2,6 +2,7 @@ module ExpressionExecutorCuda
 include("Utils.jl")
 include("ExpressionProcessing.jl")
 include("Interpreter.jl")
+include("Transpiler.jl")
 
 module CpuInterpreter
 include("Code.jl")
@@ -20,20 +21,31 @@ export test
 #
 
 # Evaluate Expressions on the GPU
-function interpret_gpu(exprs::Vector{Expr}, X::Matrix{Float32}, p::Vector{Vector{Float32}})::Matrix{Float32}
+function interpret_gpu(exprs::Vector{Expr}, X::Matrix{Float32}, p::Vector{Vector{Float32}}; repetitions=1)::Matrix{Float32}
 	@assert axes(exprs) == axes(p)
 	ncols = size(X, 2)
 
-	result = Matrix{Float32}(undef, ncols, length(exprs))
-	# interpret
+	results = Matrix{Float32}(undef, ncols, length(exprs))
+
+	for i in 1:repetitions # Simulate parameter tuning
+		results = Interpreter.interpret(exprs, X, p)
+	end
+
+	return results
 end
 
 # Convert Expressions to PTX Code and execute that instead
-function evaluate_gpu(exprs::Vector{Expr}, X::Matrix{Float32}, p::Vector{Vector{Float32}})::Matrix{Float32}
+function evaluate_gpu(exprs::Vector{Expr}, X::Matrix{Float32}, p::Vector{Vector{Float32}}; repetitions=1)::Matrix{Float32}
 	@assert axes(exprs) == axes(p)
 	ncols = size(X, 2)
 
-	result = Matrix{Float32}(undef, ncols, length(exprs))
+	results = Matrix{Float32}(undef, ncols, length(exprs))
+
+	for i in 1:repetitions # Simulate parameter tuning
+		results = Transpiler.evaluate(exprs, X, p)
+	end
+
+	return results
 end