Eric wrote:> Ulf Samuelsson wrote: > >If you run in Thumb mode, you run full speed > > I've heard this before but I don't understand it. Are you doing fetches > 32 bits at a time in Thumb mode, and doing one 32 bit fetch for every > 2, 16 bit opcodes? If so, then this is something like what the LPC MAM > is doing, right?Yes, but with much less complexity. The LPC fetches 4 words at a time, but as you say, more slowly, so a jump is more costly. Also, the LPC speeds up ARM code to single cycle as well as Thumb code. But there are more things that determine CPU speed than just Flash speed. The clock speed is a factor even if you are running from Flash with wait states. IO accesses can affect your speed depending on the IO throughput requirements of your app. So look at the chip in the total context of your app rather than considering individual pieces of the MCU.> Is the LPC slower than the SAM7 in thumb mode? Is this related to the > faster underlying speed of the SAM flash (I understand your flash can > run at 30 Mhz, but the LPC flash can only run at 20 Mhz)?On the other hand the Luminary Micro Cortex M3 MCUs have 50 MHz flash and can run at full speed with no wait states from Flash without any "special" tricks. Also, most of the instructions are 16 bit Thumb 2. Everything I hear about the LM Cortex M3 parts is good. They don't fit our apps so well because we mostly are looking for minimal power and the ARM7 cores from Atmel and Philips are still lower power for a given clock speed. I have not tried bench marking these parts to see if the higher performance per clock offsets the higher power consumption at all. BTW, the higher power consumption of the LM CM3 parts is most likely from the slightly older technology, 250 nm vs 180 nm, rather than an intrinsic property of the CM3 core. They tell us that the CM3 core should be lower power if all other factors are equal. But then when has anything been "equal"?> >and since the SAM7 only has a single waitstate it should be faster than the LPC > >at any given frequency. > > I understand that you're talking about thumb mode, but how can this be > faster than the LPC in thumb mode? I guess the LPC in inserting more > than 1 wait state in thumb mode? But aren' t they reading from flash in > chunks of 128 bits (using MAM) - that speedup should apply in both Arm > and Thumb modes, right? > > You guys have brought up some new things that I didn't know about the > SAM7 devices and this is good information. I started looking at the > SAM7X last year, but it got delayed and my interest moved to the > Philips devices..The SAM7 parts also have DMA for the peripherals which can be a real boon when you have a lot of IO going on.> Regarding support issues, I contacted Atmel AT91 support several times > last year, and I got great answers. I also like at91.com (it has an > ugly look, but excellent content with knowledgeable users). And I > didn't have any luck with Philips tech support when I contacted them > this year. With the recent change to NXP this might be a good time to > "go back to the basics" and revisit my strategy.I personally am not a fan of the AT91.com web site. This is mainly because I don't like having to go to two web sites to try to find the info and not knowing which site it is likely to be hiding. It is more than once that I have looked for info and not found it, then asked the local support for it and told there is none, only to find it later at one of the two sites. If it were at a single site, I would at least know to look a bit harder at that site until I find it. Two poor sites does not equal one good site! Perhaps they sould just give up on providing ARM info at the main site and just put it all at AT91.com?> The only support I can get for LPC is from the Yahoo forum. That's > pretty good, but its not something that's funded or endorsed by > Philips, and sometimes I wondor who's right when I see conflicting > message posts. At91.com is funded and endorsed by Atmel, if I > understand correctly.I believe I will be using Philips parts on my current project. We need two SPI ports and Atmel only supports that on the higher end SAM7X. Now that it is available in the smaller sized 100 pin BGA, it might still be a contender, but the Philips parts look very good for this socket. They have so many options in terms of memory and peripheral combinations. I believe the SAM7 parts all have the same peripherals except for the SAM7S32 and SAM7S16 which is due out.> The only minor gripe I have with Atmel is that they're very quiet when > it comes to their future plans. Philips used to "wow" us with early > news on upcoming devices. I'd like to see more info about what Atmel > has in the pipeline. We've got a long design and testing cycle so we > need to keep our eyes on the future.We get an occasional glimpse into their future offerings. But they seem to be struggling to get their latest SAM9 chips out the door and have not fixed many of the errata on the SAM7 chips. There are some that are significant to us in some apps. Power and size are both very crucial commodities to us. I don't recall hearing about anything other than the SAM9 parts which still very preliminary last fall when they briefed us. Maybe we are due for another peek into the world of tomorrow at Atmel?
Favorite Philips/NXP LPC21xx websites/forums?
Started by ●September 14, 2006
Reply by ●September 22, 20062006-09-22
Reply by ●September 24, 20062006-09-24
> I personally am not a fan of the AT91.com web site. This is mainly > because I don't like having to go to two web sites to try to find the > info and not knowing which site it is likely to be hiding. It is more > than once that I have looked for info and not found it, then asked the > local support for it and told there is none, only to find it later at > one of the two sites. If it were at a single site, I would at least > know to look a bit harder at that site until I find it. Two poor sites > does not equal one good site! Perhaps they sould just give up on > providing ARM info at the main site and just put it all at AT91.com? >I agree it would be a very good if this was implemented. The at91.com web site is under reconstruction, and the new version will be online in a month or two.> We get an occasional glimpse into their future offerings. But they > seem to be struggling to get their latest SAM9 chips out the door and > have not fixed many of the errata on the SAM7 chips. There are some > that are significant to us in some apps. Power and size are both very > crucial commodities to us. I don't recall hearing about anything other > than the SAM9 parts which still very preliminary last fall when they > briefed us. Maybe we are due for another peek into the world of > tomorrow at Atmel?The RM9200 (Ethernet ) Production since 2003 The SAM9260 (Ethernet ) is out of the door. First batch of development kits shipped to customers The SAM9260A(Ethernet ) in design (high speed SAM9260A) The SAM9261 ( LCD ) is in production The SAM9261S ( LCD ) in design (SAM9261 w 16 kB SRAM) The SAM9262 (Ethernet + LCD + GPS) is out of the door, but is moved to the GPS team and will be renamed to ATR6... The SAM9263 (Ethernet + LCD ) will be available around the end of the year. The SAM9xxx (-: ............................ :-) First silicon real soon. Everything points to the SAM9260 cleaning up the low cost embedded Linux market. -- Best Regards, Ulf Samuelsson This is intended to be my personal opinion which may, or may not be shared by my employer Atmel Nordic AB
Reply by ●September 24, 20062006-09-24
Ulf Samuelsson wrote:> The RM9200 (Ethernet ) Production since 2003 > The SAM9260 (Ethernet ) is out of the door. First > batch of development kits shipped to customers > The SAM9260A(Ethernet ) in design (high speed > SAM9260A) > The SAM9261 ( LCD ) is in production > The SAM9261S ( LCD ) in design (SAM9261 w 16 kB > SRAM) > The SAM9262 (Ethernet + LCD + GPS) is out of the door, but is moved to > the GPS team and will be renamed to ATR6... > The SAM9263 (Ethernet + LCD ) will be available around the end > of the year. > The SAM9xxx (-: ............................ :-) First silicon real > soon. > > Everything points to the SAM9260 cleaning up the low cost embedded Linux > market.Maybe I was out sick the day Atmel came to show us what they were working on and totally missed it. I just found a couple of new parts at Digikey that are not in your list. SAM7SE256 and SAM7SE512. When I pulled up what Digikey lists as a data sheet I got a selection guide (neither Atmel site had any info at all). The selection guide also showed a SAM7S512 part due out next month. The pair of SE parts seem to add an external memory interface including SDRAM, but otherwise are the same as the SAM7S (in bigger packages of course). I am adding what I found to the selection guide on www.gnuarm.com at the Resources page. If Digikey has prices on silicon, I can only expect that they parts are relatively real.
Reply by ●September 24, 20062006-09-24
"rickman" <gnuarm@gmail.com> skrev i meddelandet news:1159109923.921586.110350@b28g2000cwb.googlegroups.com...> Ulf Samuelsson wrote: >> The RM9200 (Ethernet ) Production since 2003 >> The SAM9260 (Ethernet ) is out of the door. >> First >> batch of development kits shipped to customers >> The SAM9260A(Ethernet ) in design (high speed >> SAM9260A) >> The SAM9261 ( LCD ) is in production >> The SAM9261S ( LCD ) in design (SAM9261 w 16 kB >> SRAM) >> The SAM9262 (Ethernet + LCD + GPS) is out of the door, but is moved to >> the GPS team and will be renamed to ATR6... >> The SAM9263 (Ethernet + LCD ) will be available around the >> end >> of the year. >> The SAM9xxx (-: ............................ :-) First silicon real >> soon. >> >> Everything points to the SAM9260 cleaning up the low cost embedded Linux >> market. > > Maybe I was out sick the day Atmel came to show us what they were > working on and totally missed it. I just found a couple of new parts > at Digikey that are not in your list. SAM7SE256 and SAM7SE512. When I > pulled up what Digikey lists as a data sheet I got a selection guide > (neither Atmel site had any info at all). The selection guide also > showed a SAM7S512 part due out next month. The pair of SE parts seem > to add an external memory interface including SDRAM, but otherwise are > the same as the SAM7S (in bigger packages of course). I am adding what > I found to the selection guide on www.gnuarm.com at the Resources page. > >Yes, the list is only for ARM9 based circuits. There is also a SAM7X512 and a SAM7XC512 due real soon.> If Digikey has prices on silicon, I can only expect that they parts are > relatively real.-- Best Regards, Ulf Samuelsson This is intended to be my personal opinion which may, or may not be shared by my employer Atmel Nordic AB