ECC word not necessarily full cache line

By: David Kanter (dkanter.delete@this.realworldtech.com), September 18, 2022 11:29 am
Room: Moderated Discussions
Paul A. Clayton (paaronclayton.delete@this.gmail.com) on September 17, 2022 10:59 am wrote:
> David Kanter (dkanter.delete@this.realworldtech.com) on September 16, 2022 9:44 pm wrote:
> [snip]
> > Most L1Ds are writeback. They also receive many partial writes (e.g., byte writes, 2B, 4B,
> > etc.). ECC forces every write to trigger a full cache line read and write.
>
> For L1 caches using ECC, the granularity is probably less than a cache line. The SRAM subarray word size
> is one reasonable ECC word size, but if 64-bit writes are common using a 64-bit ECC word might avoid
> enough reads to justify the area/power overhead. (Side thought: byte granular write enable is likely
> not to be supported with ECC-protected L1 caches, which might save a tiny bit of area and power.)

That's true, but sub-line ECC has a much higher area overhead. So maybe I'd state it as 'power or area, take your pick'...

> > In the L2, most
> > writes are at full line granularity. In the L1, most writes are less than a full line.
> >
> > Also, ECC read+calc+write adds latency (as others noted above).
>
> The latency does not seem that important. One would use a buffer anyway to
> provide recently written data, though greater latency would imply a larger buffer.
> Also a buffer alongside L1 could impact access latency of L1 contents.

On older AMD parts I know the ECC on the L1 was a challenge for timing and required some careful work. That was probably 45nm or older, so not sure about 7nm and beyond.

David
< Previous Post in ThreadNext Post in Thread >
TopicPosted ByDate
Data integrity of L1 cachesanon22022/09/15 06:04 PM
  Data integrity of L1 cachesGroo2022/09/15 10:46 PM
    Data integrity of L1 cachesanon22022/09/16 08:00 AM
      Data integrity of L1 cachesgroo2022/09/16 10:06 AM
        ECC outside critical path?hobold2022/09/16 12:03 PM
          ECC outside critical path?Mr. Camel2022/09/16 02:39 PM
            ECC outside critical path?anonymou52022/09/16 04:01 PM
          ECC outside critical path?anonymou52022/09/16 03:50 PM
            ECC outside critical path?hobold2022/09/17 05:57 AM
        Data integrity of L1 cachesanon22022/09/16 04:45 PM
  Data integrity of L1 cachesanon.12022/09/16 05:51 AM
    Data integrity of L1 cachesanon22022/09/16 08:04 AM
      Data integrity of L1 cachesBrett2022/09/16 11:12 AM
  Data integrity of L1 caches---2022/09/16 10:28 AM
    Data integrity of L1 cachesdmcq2022/09/16 12:41 PM
      Data integrity of L1 caches---2022/09/16 01:42 PM
    Data integrity of L1 cachesanon22022/09/16 04:49 PM
      Data integrity of L1 caches---2022/09/16 05:25 PM
        Read the thread (NT)anon22022/09/16 05:55 PM
        Data integrity of L1 cachesanon22022/09/16 05:57 PM
    Data integrity of L1 cachesMichael S2022/09/17 04:02 PM
  Data integrity of L1 cachesDavid Kanter2022/09/16 08:44 PM
    ECC word not necessarily full cache linePaul A. Clayton2022/09/17 09:59 AM
      ECC word not necessarily full cache lineDavid Kanter2022/09/18 11:29 AM
        ECC word not necessarily full cache lineAnon2022/09/18 11:54 AM
          ECC word not necessarily full cache linehobold2022/09/18 05:32 PM
            ECC word not necessarily full cache lineMichael S2022/09/19 07:47 AM
              ECC word not necessarily full cache linehobold2022/09/20 05:38 AM
                ECC word not necessarily full cache linedmcq2022/09/21 04:10 AM
                ECC word not necessarily full cache lineMichael S2022/09/21 05:55 AM
                  ECC word not necessarily full cache linehobold2022/09/21 12:59 PM
  Data integrity of L1 cachesDavid Hess2022/09/17 09:03 AM
  Data integrity of L1 cachesMichael S2022/09/17 04:12 PM
Reply to this Topic
Name:
Email:
Topic:
Body: No Text
How do you spell tangerine? ūüćä