Talk:TMS320

New pages added for series

I've started pages for the TMS320C5000 , TMS320C6000, TMS320C6200, TMS320C6700 and TMS320C6400. They need a lot more information to be added.

May need to revise the rest of this page, especially the bit "outside of the main series". Is it also worth listing all the available chips in each family ?

Also need to add typical application for each family. alxx 13:39, 20 November 2005 (UTC)

Pictures

It would be nice having some pictures here. If no one else has any, I've got a evaluation board (TMS320C5x DSK) that I could probably get some half-decent pictures of. C50 is old, but what the heck.
--magetoo 23:24, 23 May 2006 (UTC)

Thanks for the photographs. I'd like to point to pages 38 and 39 of the datasheet (TI, 1987, "TMS320 Second-Generation Signal Processors", unknown document ID, found by Google search for "TMS32020" and "datasheet", SHA1:06155fd6f09b23548da76084a50368c9efb3ebe2), which shows the package options. That chip on the top-right of page is obviously PGA but looking at the package options on page 38 proves it. GB part at end of name means grid array. FN at end would mean PLCC. They used a general pattern in their ordering codes so there is probably a better document that lists what the entire 3-letter code means. L is probably a speed or temperature grade but I can't remember and don't have the document handy. I've edited the title of the photo to say "PGA" with a link to that article. 73.181.82.26 (talk) 20:20, 21 August 2015 (UTC)

Actually, I don't know how to add https://en.wikipedia.org/wiki/Pin_grid_array inside an image link, like you can in the main article and references seem to not indicate it's possible.  :( 73.181.82.26 (talk) 20:41, 21 August 2015 (UTC)

Legacy Software

I do not konw, where to dump it, but the following file contains a dd-based floppy image and scanner images of the floppy, including its label. It came with a starter kit and says that it is a version from 1997. The log-in requirement in Wikipedia, including the stupidity of blocking tor network is certainly idiotic. Cheers! — Preceding unsigned comment added by 89.235.217.81 (talk) 20:50, 7 June 2016 (UTC)

needs more, don't you think? :-)

I was surprised there was no link to the page about the Harvard Architecture. It's possible that the more recent literature produced by TI no longer mentions this as an architectural feature, but it was definitely how the original TMS32010 was described. There isn't nearly enough about what distinguishes the TMS320 family from other CPU designs: the fact that there are C compilers for this chip family is hardly significant. 129.55.27.4 18:16, 16 August 2006 (UTC)

It does seem like an important distinction, but the lines are more blurred today with CPUs that wouldn't generally be considered DSPs including Harvard architectures. Harvard_architecture doesn't even mention the TMS320 devices, whereas Modified_Harvard_architecture does. JadonK (talk) 17:19, 21 March 2008 (UTC)

I'm not familiar with the other TMS320 families, but at least the C6x *cores* are true Harvard. The surrounding infrastructure ("GEM" for C64x+/C674x, renamed "CorePac" for C66x) may turn this into Modified Harvard depending on how it's configured at runtime: if the L1 program (L1P) and data (L1D) memories are entirely configured as caches then you get a typical split-cache processor. If however you configure them partially or entirely as memories, then the Harvard nature shows up again: the core cannot fetch code from L1D, nor can load/store instructions access L1P. You need to use the internal DMA controller or an external agent to load code into L1P in this case. L2 cache/memory and beyond is unified, but of course much slower than L1. xmath (talk) 02:12, 22 October 2015 (UTC)

DA25x

I didn't realize there was info on the DA25x here. Unfortunately, the architecture information is incorrect. If there is real interest in the correct information, I can consider providing. I'm not trying to be elusive, I just didn't think there was much demand for this information. JadonK (talk) 17:16, 21 March 2008 (UTC)

Apollo Computer videocard

Seems, the sentece about Apollo Computer is incorrect. AFAIK, their videocard was based on the TMS34010 GPU, not TMS320 DSP Alecv (talk) 18:13, 24 March 2008 (UTC)

Not that anybody really cares, but the DN570 definitely used a TMS32010. I wrote all the code. :-) Thomas144 (talk) — Preceding undated comment added 18:08, 24 September 2014 (UTC)

Ambiguous Acronym

In the "C2000 Series" section there is a link to SPI, which goes to a disambiguation page. I was reading to learn stuff when I found this, so clearly I have no idea what it is supposed to mean. Would someone who does know please change the link or spell out the acronym? 68.35.191.118 (talk) 21:13, 19 January 2013 (UTC)

For future reference for next reader, that is probably "serial peripheral interface", a fancy name for a CMOS or TTL (as opposed to RS232, these often have roughly 3 volts for logic true, and roughly 0 volts for logic false) port that normally features all of a data line, a clock line, and a strobe/latch. Yes, this is exactly the same as what SNES/NES game system controllers, some SD Card readers, and many other things use. It's so simple that everyone uses it somewhere.  ;) If you read that article, you'll notice that technically, the full SPI interface needs a separate input and output line for data, but many devices combine them like above. https://en.wikipedia.org/wiki/Serial_Peripheral_Interface_Bus 73.181.82.26 (talk) 20:34, 21 August 2015 (UTC)

Please mind the +

People, please don't conflate the C67x+ with the C67x and especially don't conflate the C64x+ with the C64x. The + may seem like a minor thing, but in addition to adding many instructions, the C64x+ is the single biggest architectural update of the entire C6x series: doubling the number of registers, adding support for privilege, exceptions, compact (16-bit) instruction encoding, and hardware-assisted loop pipelining (SPLOOP).

The C67x+ is a less significant update over the C67x: it also doubles the number of registers, but includes none of the other enhancements. It allows execute-packets to span fetch-packets, a feature which the C64x+ also has but I can't immediately find whether it introduced it or whether it was already present in the C64x.

The C674x is architectually identical to the C64x+ but merges in the floating-point instructions of the C67x+.

Note that TI's current compiler (v8) only supports C64x+ and later (C674x / C66x) cores, except for the mutant child Tesla which got killed off (along with the OMAP 4/5 itself).

xmath (talk) 01:04, 22 October 2015 (UTC)

prefefinition

hi guy, could you translate and import this [1] regards.Mvdiogoce (talk) 13:56, 7 December 2015 (UTC)

floating point

As far as I know, the floating point versions of these processors use a format not used with other processors. Seems to me that it should be documented somewhere in Wikipedia, if not in this article. Gah4 (talk) 10:03, 18 December 2020 (UTC)

Advertising

Someone threw an advert section under the C2000 series section, But I'm not sure that's required? I've removed it for now because it portrays the product in a somewhat positive light (...is well suited to many real time applications...), however the tag is for "sections that are directly trying to sell a product to our readers" and I don't believe this fits under that criteria. Feel free to revert though, as I'm not too familiar with the chips, I'm just copy-editing. OneTrueTreeTree (talk) 03:51, 1 March 2024 (UTC)