Bitboard comparison test: really confused where the time savings is

[Chessnut1071]

I'm using pre-defined byte arrays, not bitboards, for move generation. It looks like the machine cycles for a simple memory lookup are no larger than a shift operation used in bitboards. So, the FEN below lists my performance using time to solution, nodes per second, and calls to the engine. I'm using a single core, I7, 2.6 GHz Windows 10 machine.

FEN[182] = "1r3n2/2pB1pq1/1bR3p1/N2NR2b/KP1kpp1r/1p2p3/1PP2Q1n/8 w - - 0 1 "; // Philip Klett 5-move mate [Bg4,QxR,Rc5] - (28)

time: .7204 seconds
Nodes visited: 4,045,680
nps = 5,615,880.

Anybody with similar hardware, what does bitboards do with the FEN above for comparative purposes?

I'm thinking of converting to bitboards if anybody can show improvement in time. So far, I can't find any.

[Chessnut1071]

I'm using pre-defined byte arrays, not bitboards, for move generation. It looks like the machine cycles for a simple memory lookup are no larger than a shift operation used in bitboards. So, the FEN below lists my performance using time to solution, nodes per second, and calls to the engine. I'm using a single core, I7, 2.6 GHz Windows 10 machine.

FEN[182] = "1r3n2/2pB1pq1/1bR3p1/N2NR2b/KP1kpp1r/1p2p3/1PP2Q1n/8 w - - 0 1 "; // Philip Klett 5-move mate [Bg4,QxR,Rc5] - (28)

time: .7204 seconds
Nodes visited: 4,045,680
nps = 5,615,880.

Anybody with similar hardware, what does bitboards do with the FEN above for comparative purposes?

I'm thinking of converting to bitboards if anybody can show improvement in time. So far, I can't find any.

Whoops, missed it by one, here is correct FEN

FEN[181] = "n1rb4/1p3p1p/1p6/1R5K/8/p3p1PN/1PP1R3/N6k w - - 0 1 "; // Sam Loyd (1867) 5 move mate 10 [Pb4] - (10)

sorry about confusion

[hgm]

Time to solution gives no information on the speed of an engine. It is completely dominated by how you shape the search tree (reductions and extensions).