OS/2, eCS & ArcaOS - Technical > Storage

DISKIO - updated version in need of TEST

(1/4) > >>

Dariusz Piatkowski:
Hi Everyone,

I figured I'd take a stab at resolving the annoying issue with DISKIO not reporting correct throughput rates for the modern hardware, i.e. stuff like the fast SSD devices, etc.

Alright, so to that end, here is a DEBUG run for my SSD (Samsung 850 Evo - OK, I said "modern", yes, it's a few years old  ;), but it's what pushed DISKIO over the edge on my box here):


--- Code: ---Dhrystone 2.1 C benchmark routines (C) 1988 Reinhold P. Weicker
Dhrystone benchmark for this CPU: 2283670 runs/sec

Hard disk 2: 255 sides, 30401 cylinders, 63 sectors per track = 238472 MB
Drive cache/bus transfer rate:
 DEBUG => nData is: 1189762560
 DEBUG => nTime is: 10260
115961 k/sec
Data transfer rate on cylinder 0   :
 DEBUG => nData is: 3087028224
 DEBUG => nTime is: 10264
300762 k/sec
Data transfer rate on cylinder 30399:
 DEBUG => nData is: 2904620544
 DEBUG => nTime is: 10244
283543 k/sec
CPU usage by full speed disk transfers: 23%
Average latency time:
 DEBUG => nTime is: 10036.599360
 DEBUG => nCnt is: 85352.196980
0.1 ms
Average data access time: Disk read error.
Multithreaded disk I/O (4 threads): 121669 k/sec, 19% CPU usage

--- End code ---

The stuff marked as "DEBUG => " is the raw data, if you're going to run this on your machine please report back the full output.

And here is the RETAIL output:


--- Code: ---Dhrystone 2.1 C benchmark routines (C) 1988 Reinhold P. Weicker
Dhrystone benchmark for this CPU: 2088687 runs/sec

Hard disk 2: 255 sides, 30401 cylinders, 63 sectors per track = 238472 MB
Drive cache/bus transfer rate: 118449 k/sec
Data transfer rate on cylinder 0   : 332803 k/sec
Data transfer rate on cylinder 30399: 302953 k/sec
CPU usage by full speed disk transfers: 23%
Average latency time: 0.1 ms
Average data access time: Disk read error.
Multithreaded disk I/O (4 threads): 122717 k/sec, 23% CPU usage

--- End code ---

As you can tell, the numbers are much better (no more overflow) and they for the most part jive with what I see in SysBench.

There are a few outstanding issues that I'm aware of:

1) this build is optimized for Pentium2, that's old stuff, I need to either remove or optimize for Pentium4, I just haven't looked up all the ICC flags & options, may be the reason why the Dhrystone value reported by this build is about 50% of what the official build has, not sure yet

2) the "Average data access time: Disk read error" is still here, this only fails on my SSD drive, I need to dig into the DosDevIOCtl API

3) "Drive cache/bus transfer rate:" shows a smaller rate than SysBench, not sure why that is, need to review further

4) RAMDISK numbers are still bad, or just flat out error out

ZIP attachment has two builds :

1) DEBUG - has the extra raw data dumps
2) RETAIL - if you just want an updated (hopefully correct) result and that's all you want

Feedback is welcome and appreciated!

-Dariusz

Martin Iturbide:
Hi Dariusz

I give it a quick try on a ArcaOS VM. The VM is running on a Win10 host with an SDD drive.

- Can it be possible to release it as 1.19? so it can not get confused with 1.18z
- Is it possible to share the source code to update the github repository? - once you produce the final release.

It seems to me that tthe program runs fine (diskio and debug)
But when I try to create an output file this way, I get this error?

--- Quote ---[C:\HOME\DOWNLOADS\DISKIO_FIX_1]diskio-debug >debug.txt

SYS1808:
The process has stopped.  The software diagnostic
code (exception code) is  009B.

--- End quote ---

Is there any other way you recommend you get the output file?

Regards

David McKenna:
 Hi Dariusz,

 I gave the retail diskio a shot (diskio > diskio.txt) and got this:

DISKIO - Fixed Disk Benchmark, Version 1.18z
(C) 1994-1998 Kai Uwe Rommel
(C) 2004 madded2

Number of fixed disks: 5
Number of CD-ROM drives: 2

Dhrystone 2.1 C benchmark routines (C) 1988 Reinhold P. Weicker
Dhrystone benchmark for this CPU: 4009371 runs/sec

Hard disk 1: 255 sides, 60801 cylinders, 63 sectors per track = 476937 MB
Drive cache/bus transfer rate: 86400 k/sec
Data transfer rate on cylinder 0   : 277997 k/sec
Data transfer rate on cylinder 60799: 260108 k/sec
CPU usage by full speed disk transfers: 48%
Average latency time: 0.0 ms
Average data access time: Disk read error.
Multithreaded disk I/O (4 threads): 138280 k/sec, 41% CPU usage

Hard disk 2: 255 sides, 60321 cylinders, 127 sectors per track = 953855 MB
Drive cache/bus transfer rate: -107190 k/sec
Data transfer rate on cylinder 0   : 185943 k/sec
Data transfer rate on cylinder 60319: 87011 k/sec
CPU usage by full speed disk transfers: 35%
Average latency time: 0.0 ms
Average data access time: 15.3 ms
Multithreaded disk I/O (4 threads): 45298 k/sec, 28% CPU usage

Hard disk 3: 255 sides, 650 cylinders, 63 sectors per track = 5098 MB
Drive cache/bus transfer rate: -142336 k/sec
Data transfer rate on cylinder 0   : Disk read error.
Data transfer rate on cylinder 648 : Disk read error.
CPU usage by full speed disk transfers: 93%
Average latency time: 0.0 ms
Average data access time: Disk read error.
Multithreaded disk I/O (4 threads): Disk read error.
-34014 k/sec, 64% CPU usage

Hard disk 4: 12 sides, 512 cylinders, 9 sectors per track = 27 MB
Drive cache/bus transfer rate: Disk read error.
Data transfer rate on cylinder 0   : Disk read error.
Data transfer rate on cylinder 510 :
____________________________________________________

 It gave an error on the command line, I'm guessing because it detected 5 drives, but there are really only 3 (Disk 1 is an SSD, Disk 2 is a rotating hard drive, Disk 3 is a RAM drive) but the other 2 are possible USB drives. CDROMs there is only 1 (empty) and the other is also a possible USB drive. Not sure what to make of the results otherwise....

Regards,

Dariusz Piatkowski:
Hi Martin,


--- Quote from: Martin Iturbide on December 20, 2020, 02:40:40 am ---...
- Can it be possible to release it as 1.19? so it can not get confused with 1.18z
- Is it possible to share the source code to update the github repository? - once you produce the final release.

--- End quote ---

Yes, absolutely. For now my focus is on getting as many samples of different hardware as possible. I suspect the really big drives (well, one should not think of a 1T that as being that big anymore...LOL, but David's example below still shows an issue there).

Once the little tweaks are behind us this is of course going out there. My changes are really tiny things, and quite honestly very simple changes from 'int' to 'long' or ULONG variables for the most part.


--- Quote from: Martin Iturbide on December 20, 2020, 02:40:40 am ---...
It seems to me that tthe program runs fine (diskio and debug)
But when I try to create an output file this way, I get this error?

--- Quote ---[C:\HOME\DOWNLOADS\DISKIO_FIX_1]diskio-debug >debug.txt

SYS1808:
The process has stopped.  The software diagnostic
code (exception code) is  009B.

--- End quote ---

--- End quote ---

Aha...so I wasn't getting this in my tests because I was individually testing just my SSD or HDD devices, but running it just like you did eventually produced the same result. What I found is the following in my POPUPLOG.OS2:


--- Code: ---12-19-2020  20:49:36  SYS3184  PID 0819  TID 0001  Slot 0125
Y:\TMP\DISKIO.EXE
c000009b
00022182
EAX=00000000  EBX=00000000  ECX=000402ec  EDX=00000000
ESI=00000090  EDI=00000090 
DS=0053  DSACC=f0f3  DSLIM=ffffffff 
ES=0053  ESACC=f0f3  ESLIM=ffffffff 
FS=150b  FSACC=00f3  FSLIM=00000030
GS=0000  GSACC=****  GSLIM=********
CS:EIP=005b:00022182  CSACC=f0df  CSLIM=ffffffff
SS:ESP=0053:00092440  SSACC=f0f3  SSLIM=ffffffff
EBP=000924fc  FLG=00210246

DISKIO.EXE 0002:00002182

--- End code ---

...and SYS3184 is a "Divide by Zero".

Looking at where my output from DISKIO stopped it looks like the un-mounted USB removable drive is what caused the problem here. I'm running the DEBUG version right now to see the raw data but I'm sure that pesky 0 (zero) will show up.

In the meantime: thank you for the test and feedback!

Dariusz Piatkowski:
Hi David,


--- Quote from: David McKenna on December 20, 2020, 03:01:29 am ---...
Hard disk 2: 255 sides, 60321 cylinders, 127 sectors per track = 953855 MB
Drive cache/bus transfer rate: -107190 k/sec
Data transfer rate on cylinder 0   : 185943 k/sec
Data transfer rate on cylinder 60319: 87011 k/sec
CPU usage by full speed disk transfers: 35%
Average latency time: 0.0 ms
Average data access time: 15.3 ms
Multithreaded disk I/O (4 threads): 45298 k/sec, 28% CPU usage

--- End quote ---

Can you run the DEBUG version for this drive?

The 'cache/bus transfer rate' shows overflow and I'm curious what the raw counters actually are.


--- Quote from: David McKenna on December 20, 2020, 03:01:29 am ---... It gave an error on the command line, I'm guessing because it detected 5 drives, but there are really only 3 (Drive C: is an SSD, Drive D: is a rotating hard drive, Drive Z: is a RAM drive) but the other 2 are possible USB drives. CDROMs there is only 1 (empty) and the other is also a possible USB drive. Not sure what to make of the results otherwise....

--- End quote ---

Yeah, I think this is the same issue Martin reported and which I was able to reproduce here. This happens when I run the full DISKIO and it kicks off a test for each storage device without me explicitly specifying which device it should be testing.

Truth be told changing the code to automatically exclude a USB drive (when it's not mounted) is probably beyond my OS/2 coding skillset...but I'll investigate it to learn a thing or two!

Much appreciate the feedback David! Thanks...

Navigation

[0] Message Index

[#] Next page

Go to full version