The version that Roger uploaded earlier
(http://tech.groups.yahoo.com/group/hercules-390/message/65634) was not
properly customized, and as a result did not perform a complete suite of
tests of the hardware. Many of the tests were completely missing and several
did not really perform a proper test (which for some requires customized
assembler routines to directly access the floating-point hardware).
This new version addresses those issues and thus produces more
*** I WOULD LIKE TO REQUEST... ***
...that someone with access to a real z10 to please try running it and then
send me the results.
A simple level-1 test only takes 2:19 (two minutes, 19 seconds) to complete
on Hercules (and runs 100% clean, no errors whatsoever), so it should only
run for a few *seconds* on a real z I would think.
I am very anxious to know whether it runs clean on a real z10 as well. (I
would think it damn well better!!(*))
Whoever decides to do this for me should probably first verify that
softfloat itself (and thus testfloat too) is working correctly (i.e. was
built correctly), by first performing a "testsoftfloat" test: