using (32bit) latest snapshot version on w2k, running a program doing a lot of 32 bit address arithmetics, there seems to be an initialization error, maybe coupled with sign-extension in 64 bit register.
when I run the same program multiple times, with identical input, I get different address values for access violations, although the backtrace shows program termination in some cases with identical PC positions, e.g.:
eventually, after a couple of re-starts, the program completes without error.
in 64 bit addressing mode,
the same application either stops with self-test assertion (seems reproducible, so I might be able find the instruction going wrong in the emu) or it hangs in endless loop (also not reproducible), depending on input data.
I don't really need an alpha emulator but it'd be a real cute tool if it worked on a pc.
if the es40 developper(s) are interested in help tracking down this emu error, please let me know what kind of info you'd need.
thank you!
I also have a few more or less sophisticated test programs for s,t,g,f,d-floating point number conversions and arithmetics; I wouldn't call it FP test suite, but the program output log could be compared to a log produced on a real alpha.
(I did not yet try to run these and other test programs on the emu. but will now copy the programs into the emu to be able to run them before I attack my larger apps.)
p.s.: On a real alpha (same opsys version vms v8.3), the application works fine since many years. 32 and 64 bit mode.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
using (32bit) latest snapshot version on w2k, running a program doing a lot of 32 bit address arithmetics, there seems to be an initialization error, maybe coupled with sign-extension in 64 bit register.
when I run the same program multiple times, with identical input, I get different address values for access violations, although the backtrace shows program termination in some cases with identical PC positions, e.g.:
%SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual address=FFFFFFFFFFFF
FFB8, PC=000000007AFA1E00, PS=0000001B
%SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual address=FFFFFFFFF800
0004, PC=000000000074D36C, PS=0000001B
%SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual address=000000002D00
0004, PC=000000000074D36C, PS=0000001B
eventually, after a couple of re-starts, the program completes without error.
in 64 bit addressing mode,
the same application either stops with self-test assertion (seems reproducible, so I might be able find the instruction going wrong in the emu) or it hangs in endless loop (also not reproducible), depending on input data.
I don't really need an alpha emulator but it'd be a real cute tool if it worked on a pc.
if the es40 developper(s) are interested in help tracking down this emu error, please let me know what kind of info you'd need.
thank you!
I also have a few more or less sophisticated test programs for s,t,g,f,d-floating point number conversions and arithmetics; I wouldn't call it FP test suite, but the program output log could be compared to a log produced on a real alpha.
(I did not yet try to run these and other test programs on the emu. but will now copy the programs into the emu to be able to run them before I attack my larger apps.)
p.s.: On a real alpha (same opsys version vms v8.3), the application works fine since many years. 32 and 64 bit mode.