// Tests measure speed // FIB measures the inner interpreters call overhead really well. // it is a benchmark of call speed.
: FIB ( n -- n1 ) [ FLAT FIB ] DUP 1> IF 1- DUP 1- FIB SWAP FIB + THEN ;
: FIB34 [ FLAT FIB34 ] 34 FIB DROP ;
: t1 25 TIMESDO FIB34 ;
TIMEIT t1
// 27th November 2021 TIMEIT t1 9781 : ms to run ( 97813 ) ns
// 29th November 2021 - more reasonable. FASTER FIB // not traceable TIMEIT t1 3864 : ms to run ( 38647 ) ns
// 23rd December TIMEIT t1 3574 : ms to run ( 35741 ) ns
// 26th December TIMEIT t1 3397 : ms to run ( 33978 ) ns
// Jan 16th FLOAD FIB.forth 3413 : ms to run ( 34136 ) ns
// PRIME SIEVE benchmark // tests variety of features.
CREATE FLAGS 8190 ALLOT 0 VARIABLE EFLAG FLAGS 8190 + EFLAG !
: PRIMES ( -- n ) [ FLAT PRIMES ] FLAGS 8190 1 FILL 0 3 EFLAG @ FLAGS DO I C@ IF DUP I + DUP EFLAG @ < IF EFLAG @ SWAP DO 0 I C! DUP +LOOP ELSE DROP THEN SWAP 1+ SWAP THEN 2+ LOOP DROP ;
: BENCHMARK FLAGS 8190 + EFLAG ! 0 1000 0 DO PRIMES NIP LOOP ;
: t2 25 TIMESDO BENCHMARK ;
// 432 ms Sat 4th December // t2 10357 : ms to run ( 103577 ) ns // t2 8043 ms // after adding FILL.
// 26th December // t2 7851 : ms to run ( 78510 ) ns // t2 (FLAT) 6978 : ms to run ( 69781 ) ns
// 16th January FLOAD PRIME.forth 7953 : ms to run ( 79535 ) ns
8295 : ms to run ( 82956 ) ns 8299 : ms to run ( 82998 ) ns
4463 : ms to run ( 44630 ) ns 4344 : ms to run ( 43442 ) ns