rtl tests

[Stefan Israelsson Tampe <stefan.itampe@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1610 bytes --]

Hi,

I'm just finished debugging the rtl generator to actually compile files to
the rtl VM and today I managed to get srfi-1.scm to pass all tests.

Maybe more interesting is the performance test in the repo. And here are
the vote:
--------------------------------------------------------------------------------------
RTL VM
=======:
("srfi-1.bm: fold: big" 30
real 1.11

("srfi-1.bm: fold: small" 2000000
real 1.18

("srfi-1.bm: drop-while: big" 30
real 1.24

("srfi-1.bm: drop-while: small" 2000000
real 1.27

("srfi-1.bm: map: big" 30
real 2.95

("srfi-1.bm: map: small" 2000000
real 2.35

("srfi-1.bm: for-each: big" 30 real
1.88

("srfi-1.bm: for-each: small" 2000000
real 1.72
-------------------------------------------------------------------------------------
GUILE-2.0 VM in master?
===================
("srfi-1.bm: fold: big" 30
real 1.47

("srfi-1.bm: fold: small" 2000000 real
1.29 real/iteration

("srfi-1.bm: drop-while: big" 30 real
1.43 real/iteration

("srfi-1.bm: drop-while: small" 2000000 real
1.23 real/iteration

("srfi-1.bm: map: big" 30 real
3.42 real/iteration

("srfi-1.bm: map: small" 2000000 real
2.55 real/iteration

("srfi-1.bm: for-each: big" 30 real
2.25

("srfi-1.bm: for-each: small" 2000000 real
1.80
----------------------------------------------------------------------------------------

To note here are that I executed the functions from guile-2.0 VM
e.g. there is an expensive call to set upp the RTL VM for each iteration of
the test. Therefore I would not judge the small cases. On the other hand
the big cases shows quite a significant speedup!

Have fun!
/Stefan

[-- Attachment #2: Type: text/html, Size: 3184 bytes --]