Iterated a bit on the bitpop:count design, hopefully sped it up a bit.

src/bitpop.erl

@@ -25,10 +25,48 @@
 
 -export([count/1]).
 
+-ifdef(WORKING_METHOD_1). % Slow, but effective.
 count(0) -> 0;
 count(X) when is_integer(X), X > 0 -> count(X, 0).
 count(0, Acc) -> Acc;
 count(X, Acc) -> count((X band (X - 1)), (Acc + 1)).
+-endif.
+
+%-ifdef(WORKING_METHOD_2). % Works well enough
+count(0) -> 0;
+count(X)
+  when is_integer(X), X > 0, X < 16#FFFFFFFF ->
+    ((c4(X) bsr 16) + c4(X)) band 16#0000FFFF.
+c1(V) -> V - ((V bsr 1) band 16#55555555).
+c2(V) -> ((c1(V) bsr 2) band 16#33333333) + (c1(V) band 16#33333333).
+c3(V) -> ((c2(V) bsr 4) + c2(V)) band 16#0F0F0F0F.
+c4(V) -> ((c3(V) bsr 8) + c3(V)) band 16#00FF00FF.
+%-endif.
+
+-ifdef(NOT_YET_WORKING_METHOD_3). % Might be faster than method 2
+count(0) -> 0;
+count(X) when is_integer(X), X > 0, X =< 16#FFFFFFFF ->
+    ((c2(X) + (c2(X) bsr 4) band 16#0F0F0F0F) * 16#01010101) bsr 24.
+c1(V) -> V - ((V bsr 1) band 16#55555555).
+c2(V) -> (c1(V) band 16#33333333) + ((c1(V) bsr 2) band 16#33333333).
+-endif.
+
+-ifdef(NOT_YET_WORKING_METHOD_3_64bit).
+count(X) when is_integer(X) -> %, X > 16#FFFFFFFF, X =< 16#FFFFFFFFFFFFFFFF ->
+    M1  = 16#5555555555555555, %% 0101...
+    M2  = 16#3333333333333333, %% 00110011..
+    M4  = 16#0f0f0f0f0f0f0f0f, %% 4 zeros,  4 ones ...
+    H01 = 16#0101010101010101, %% the sum of 256 to the power of 0,1,2,3...
+
+    %% 1. put count of each 2 bits into those 2 bits
+    X1 = X - (X bsr 1) band M1,
+    %% 2. put count of each 4 bits into those 4 bits
+    X2 = (X1 band M2) + ((X1 bsr 2) band M2),
+    %% 3. put count of each 8 bits into those 8 bits
+    X3 = (X2 + (X2 bsr 4)) band M4,
+    %% 4. returns left 8 bits of x + (x<<8) + (x<<16) + (x<<24) + ...
+    (X3 * H01) bsr 56.
+-endif.
 
 -ifdef(TEST).