[U-Boot-Users] [patch] MPC8266 March-C memory test
Patch to do full March-C- memory test on the MPC8266ADS board.
Later on I will probably include it as a command, right now it runs as test_dram() which is called first thing after SDRAM is initializes, and before anything uses the SDRAM.
Note: Fast version uses floating point assembly and is therefore PPC specific. Generic version takes more than twice the amount of time to complete. (Fast version: ~50 sec on 256MB RAM, slow version ~1:50 on 256MB RAM) ADS Rune Torgersen Software Developer Innovative Systems 1000 Innovative Drive Mitchell, SD 57301 Phone: 605-995-6120 Fax: 605-995-0084 email: runet@innovsys.com www.innovsys.com
Should this also be portable to the 8xx[F]ADS platforms?
I am currently debugging UPM/memory problems, and a real memory test would be invaluable :)
On Mon, May 19, 2003 at 04:26:47PM -0500, Rune Torgersen wrote:
Patch to do full March-C- memory test on the MPC8266ADS board.
Later on I will probably include it as a command, right now it runs as test_dram() which is called first thing after SDRAM is initializes, and before anything uses the SDRAM.
Note: Fast version uses floating point assembly and is therefore PPC specific. Generic version takes more than twice the amount of time to complete. (Fast version: ~50 sec on 256MB RAM, slow version ~1:50 on 256MB RAM) ADS Rune Torgersen Software Developer Innovative Systems 1000 Innovative Drive Mitchell, SD 57301 Phone: 605-995-6120 Fax: 605-995-0084 email: runet@innovsys.com www.innovsys.com
In message 20030519220854.GA15969@zumanetworks.com you wrote:
Should this also be portable to the 8xx[F]ADS platforms?
Please note that there is also extensive POST code for MPC8xx systems, which includes an IMHO excellent memory test.
I am currently debugging UPM/memory problems, and a real memory test would be invaluable :)
It won't help much. Usually memory tests test just the memory _contents_ using read and write accesses, while 99% of all RAM problems happen in burst mode, i. e. while fetching instructions or while doing DMA. I have yet to see a RAM test that catches this reliably.
Best regards,
Wolfgang Denk
Stupid question:
Will load/store string/multiple cause a burst if i implement a test in asm?
On Tue, May 20, 2003 at 09:55:40PM +0200, Wolfgang Denk wrote:
In message 20030519220854.GA15969@zumanetworks.com you wrote:
Should this also be portable to the 8xx[F]ADS platforms?
Please note that there is also extensive POST code for MPC8xx systems, which includes an IMHO excellent memory test.
I am currently debugging UPM/memory problems, and a real memory test would be invaluable :)
It won't help much. Usually memory tests test just the memory _contents_ using read and write accesses, while 99% of all RAM problems happen in burst mode, i. e. while fetching instructions or while doing DMA. I have yet to see a RAM test that catches this reliably.
Best regards,
Wolfgang Denk
-- Software Engineering: Embedded and Realtime Systems, Embedded Linux Phone: (+49)-8142-4596-87 Fax: (+49)-8142-4596-88 Email: wd@denx.de Do not underestimate the value of print statements for debugging. Don't have aesthetic convulsions when using them, either.
Dear Rune,
in message 9F5558593173D21190EA0008C7B1769873138B@innsys1 you wrote:
Patch to do full March-C- memory test on the MPC8266ADS board.
I hesitate to include this.
First, as implemented now the code, while being generic for all boards using 860 class processors, is limited to the MPC8266ADS board. If we implement it, we should make it optional for all these boards.
Second, the code duplicates existing functionality (see the memory test code in post/memory.c). Instead of adding new versions of the same stuff again and again (and having to clean it all up later) we rather should get the post/ code working on your board, too.
Finally, there are some minor issues with your patch: - there is no README documentation for the new CFG_ variables - there are errors in the code; for example:
+int testdram (void) +{
- DECLARE_GLOBAL_DATA_PTR;
- MEMDATATYPE *pstart = (MEMDATATYPE *) CFG_SDRAM_BASE;
- MEMDATATYPE *pend;
- MEMDATATYPE *p;
- uint size;
- uint erraddr;
- pend = (MEMDATATYPE*)gd->ram_size;
This should read:
pend = pstart + (gd->ram_size / sizeof(MEMDATATYPE));
What do you suggest how to proceed?
Best regards,
Wolfgang Denk
participants (3)
-
Nye Liu -
Rune Torgersen -
Wolfgang Denk