diff --git a/Makefile b/Makefile
index b1acca61..3e63a195 100644
--- a/Makefile
+++ b/Makefile
@@ -22,22 +22,24 @@ watch:
 DUNE_OPTS_BENCH?=--profile=release
 
 N?=40
-NITER?=3
+NITER?=2
 BENCH_PSIZE?=1,4,8,20
+BENCH_KIND?=simple,pool
 BENCH_CUTOFF?=20
 bench-fib:
 	@echo running for N=$(N)
 	dune build $(DUNE_OPTS_BENCH) benchs/fib_rec.exe
-	hyperfine -L psize $(BENCH_PSIZE) --warmup=1 \
-		'./_build/default/benchs/fib_rec.exe -cutoff $(BENCH_CUTOFF) -niter $(NITER) -psize={psize} -n $(N)'
+	hyperfine -L psize $(BENCH_PSIZE) -L kind $(BENCH_KIND) --warmup=1 \
+		'./_build/default/benchs/fib_rec.exe -cutoff $(BENCH_CUTOFF) -niter $(NITER) -psize={psize} -kind={kind} -n $(N)'
 
 PI_NSTEPS?=100_000_000
 PI_MODES?=seq,par1,forkjoin
+PI_KIND?=simple,pool
 bench-pi:
 	@echo running for N=$(PI_NSTEPS)
 	dune build $(DUNE_OPTS_BENCH) benchs/pi.exe
-	hyperfine -L mode $(PI_MODES) --warmup=1 \
-		'./_build/default/benchs/pi.exe -mode={mode} -n $(PI_NSTEPS)'
+	hyperfine -L mode $(PI_MODES) -L kind $(PI_KIND) --warmup=1 \
+		'./_build/default/benchs/pi.exe -mode={mode} -kind={kind} -n $(PI_NSTEPS)'
 
 .PHONY: test clean bench-fib bench-pi
 
diff --git a/benchs/dune b/benchs/dune
index 2c798176..0ae20bf3 100644
--- a/benchs/dune
+++ b/benchs/dune
@@ -3,4 +3,4 @@
  (names fib_rec pi)
  (preprocess (action
    (run %{project_root}/src/cpp/cpp.exe %{input-file})))
- (libraries moonpool unix))
+ (libraries moonpool unix trace trace-tef))
diff --git a/benchs/fib_rec.ml b/benchs/fib_rec.ml
index 1a3d3288..57a444f0 100644
--- a/benchs/fib_rec.ml
+++ b/benchs/fib_rec.ml
@@ -18,8 +18,14 @@ let rec fib ~on x : int Fut.t =
 
 let () = assert (List.init 10 fib_direct = [ 1; 1; 2; 3; 5; 8; 13; 21; 34; 55 ])
 
-let run ~psize ~n ~seq ~niter () : unit =
-  let pool = lazy (Pool.create ~min:psize ()) in
+let create_pool ~psize ~kind () =
+  match kind with
+  | "simple" -> Simple_pool.create ~min:psize ()
+  | "pool" -> Pool.create ~min:psize ()
+  | _ -> assert false
+
+let run ~psize ~n ~seq ~niter ~kind () : unit =
+  let pool = lazy (create_pool ~kind ~psize ()) in
   for _i = 1 to niter do
     let res =
       if seq then (
@@ -39,6 +45,7 @@ let () =
   let psize = ref 16 in
   let seq = ref false in
   let niter = ref 3 in
+  let kind = ref "pool" in
   let opts =
     [
       "-psize", Arg.Set_int psize, " pool size";
@@ -46,9 +53,12 @@ let () =
       "-seq", Arg.Set seq, " sequential";
       "-niter", Arg.Set_int niter, " number of iterations";
       "-cutoff", Arg.Set_int cutoff, " cutoff for sequential computation";
+      ( "-kind",
+        Arg.Symbol ([ "pool"; "simple" ], ( := ) kind),
+        " pick pool implementation" );
     ]
     |> Arg.align
   in
 
   Arg.parse opts ignore "";
-  run ~psize:!psize ~n:!n ~seq:!seq ~niter:!niter ()
+  run ~psize:!psize ~n:!n ~seq:!seq ~niter:!niter ~kind:!kind ()
diff --git a/benchs/pi.ml b/benchs/pi.ml
index 1dd55fb9..36b4e92a 100644
--- a/benchs/pi.ml
+++ b/benchs/pi.ml
@@ -17,15 +17,23 @@ let run_sequential (num_steps : int) : float =
   pi
 
 (** Create a pool *)
-let with_pool f =
-  if !j = 0 then
-    Pool.with_ ~per_domain:1 f
-  else
-    Pool.with_ ~min:!j f
+let with_pool ~kind f =
+  match kind with
+  | "pool" ->
+    if !j = 0 then
+      Pool.with_ ~per_domain:1 f
+    else
+      Pool.with_ ~min:!j f
+  | "simple" ->
+    if !j = 0 then
+      Simple_pool.with_ ~per_domain:1 f
+    else
+      Simple_pool.with_ ~min:!j f
+  | _ -> assert false
 
 (** Run in parallel using {!Fut.for_} *)
-let run_par1 (num_steps : int) : float =
-  let@ pool = with_pool () in
+let run_par1 ~kind (num_steps : int) : float =
+  let@ pool = with_pool ~kind () in
 
   let num_tasks = Pool.size pool in
 
@@ -53,8 +61,8 @@ let run_par1 (num_steps : int) : float =
 
 [@@@ifge 5.0]
 
-let run_fork_join num_steps : float =
-  let@ pool = with_pool () in
+let run_fork_join ~kind num_steps : float =
+  let@ pool = with_pool ~kind () in
 
   let num_tasks = Pool.size pool in
 
@@ -90,9 +98,11 @@ type mode =
   | Fork_join
 
 let () =
+  let@ () = Trace_tef.with_setup () in
   let mode = ref Sequential in
   let n = ref 1000 in
   let time = ref false in
+  let kind = ref "pool" in
 
   let set_mode = function
     | "seq" -> mode := Sequential
@@ -109,6 +119,9 @@ let () =
         " mode of execution" );
       "-j", Arg.Set_int j, " number of threads";
       "-t", Arg.Set time, " printing timing";
+      ( "-kind",
+        Arg.Symbol ([ "pool"; "simple" ], ( := ) kind),
+        " pick pool implementation" );
     ]
     |> Arg.align
   in
@@ -118,8 +131,8 @@ let () =
   let res =
     match !mode with
     | Sequential -> run_sequential !n
-    | Par1 -> run_par1 !n
-    | Fork_join -> run_fork_join !n
+    | Par1 -> run_par1 ~kind:!kind !n
+    | Fork_join -> run_fork_join ~kind:!kind !n
   in
   let elapsed : float = Unix.gettimeofday () -. t_start in