Java performance in different firmwares?

[ehudshapira]

Is there anywhere a Java performance comparison between different S60.5 firmware or JRT versions, particularly for 5800 and N97?

[r2j7]

Hello ehudshapira,

what kind of Java performance do you mean?

Regards,
r2j7

[ehudshapira]

API (e.g., getRGB/drawRGB which is quite abysmal; seemingly slower than lower-end S40 devices), native language (calculations, array access), and anything else.

[r2j7]

This has been at least discussed on these discussion boards, here are some picks:

http://discussion.forum.nokia.com/fo...d.php?t=192027

http://www.forum.nokia.com/forum/sho...d.php?t=167551

Esp. this post by grahamhughes could provide some answers to you questions:
http://discussion.forum.nokia.com/fo...58&postcount=2

Regards,
r2j7

[ehudshapira]

Thanks. These threads also indicate 5800 performs poorly, but they're not about relative performance of different firmwares/JRTs (also 3D is less interesting to me than "core" JVM or APIs).

JBenchmark lumps data per device, so not much help with performance differences due to different firmware/software.

[r2j7]

Hello ehudshapira,

.g., getRGB/drawRGB which is quite abysmal; seemingly slower than lower-end S40 devices

What also makes a difference is how much of the display area is used for the purpose; in addition to lack of certain hw, the screen size (resolution) and therefore the area for showing the application is considerably larger (than in Series 40 for example) when utilized in full, therefore affecting the performance esp. when the full screen area for the application is used.

Regards,
r2j7

[ehudshapira]

What also makes a difference is how much of the display area is used for the purpose; in addition to lack of certain hw, the screen size (resolution) and therefore the area for showing the application is considerably larger (than in Series 40 for example) when utilized in full, therefore affecting the performance esp. when the full screen area for the application is used.

I think a new hi-res phone would be expected to be able to cope with the increased resolution at least as well as older models did with their lower-res. In this specific case though it's slower even per pixel (compared to a random low-endish Nokia).

[r2j7]

I think a new hi-res phone would be expected to be able to cope with the increased resolution at least as well as older models did with their lower-res. In this specific case though it's slower even per pixel (compared to a random low-endish Nokia).

Yes I agree with you that this would be indeed the preferred choice. However and as said previously, unfortunately it seems to be the case that when a device lacks the hw which is necessary for improved graphics acceleration performance, increased resolution alone cannot take the place of the previously available hw enabler.

Regards,
r2j7

[petrib]

Yes I agree with you that this would be indeed the preferred choice. However and as said previously, unfortunately it seems to be the case that when a device lacks the hw which is necessary for improved graphics acceleration performance, increased resolution alone cannot take the place of the previously available hw enabler.

Regards,
r2j7

Increased resolution without increased hardware perfomance (faster CPU, faster CPU clock speed, faster memory bus, faster display memory) and without 2D or 3D hardware graphics acceleration can only lead to slower performance.

That's exactly what happened with the Nokia 5800 or N97, N97 mini and other Nokia S60 5th Edition devices based more or less on the same hardware; slow processor, several times more pixels than QVGA, no GPU -> lousy graphics performance (compared to similar hardware with only QVGA).

[ektasrv]

Hope all these Issues will be taken care with Nokia N8 launch it has good CPU and a H/W GPU.
CPU Clock Rate 680 MHz
Graphics Processor 2D/3D Graphics HW Accelerator with OpenVG1.1 and OpenGL ES 2.0 support

[petrib]

Hope all these Issues will be taken care with Nokia N8 launch it has good CPU and a H/W GPU.
CPU Clock Rate 680 MHz
Graphics Processor 2D/3D Graphics HW Accelerator with OpenVG1.1 and OpenGL ES 2.0 support

For that one device model, it'll be OK, I presume (I haven't seen/touched an N8 yet). Of course, there's still 1.3-1.5B (billion) Nokia phone model users out there, which you need to get to switch/upgrade to this model. Or if not this model, then to some other, unannounced as of yet, models with N8 class or better hardware. Of course, most of the current Nokia phone users will not be switching to this or other high-end models, but will instead keep on using lower spec models, and will be also buying new lower spec models (as Nokia will also be manufacturing lower spec models many times more than high-end models). In other words, it'll take a while before there are significant numbers of well performing Nokia devices in use.

[Utumno]

drawRGB on a N97 mini (and I suspect every other S60 5th) is indeed terrible. One call to a fullscreen ( 640x360 ) drawRGB takes 30ms. This makes development of any more advanced, faster game an exercise in frustration.

[ehudshapira]

It's not only drawRGB, and I'm not convinced it's the CPU or lack of 2D acceleration. Blitting, for example, isn't that taxing anyway. Maybe it's slow memory access or inefficiencies in the Java to native API glue. Can anyone doing native Symbian code share impressions on performance?

If Nokia can't provide detailed info on performance differences between firmware versions, can we at least get more general info such as which firmware updates affect Java performance (if at all) and in what aspects?

Do firmware updates in general affect Java? Do they include performance-related API or JVM updates (seemingly JRT2 isn't included)?

[petrib]

Firmware updates are in general just bugfix releases. A bugfix could also be a serious performance related issue.

Of course, fixing one thing may affect another thing too (positively or negatively).

Sometimes the firmware updates do also bring new features/functionality, too depending on the case.

Nokia has never published any particular benchmarking/performance data for their devices. For Java performance, you could of course use tests like http://www.jbenchmark.com/ on different firmware versions of the same model to see if there is a difference (unless the site already shows/reports firmware version differences).

[ehudshapira]

Even if Nokia doesn't publish performance data I hope the least they could do is to have developer-oriented release notes for firmware updates including performance data when relevant.

If I had all devices and firmware versions of course I could test it, but I don't. And it's impractical when downgrading is not sanction by Nokia both by making it difficult to do and by not providing historic firmware versions.

JBenchmark doesn't have performance data per firmware version. But it does seem to indicate S60.5 is a poor performer in general.