DPDK patches and discussions
 help / color / mirror / Atom feed
* [PATCH 0/2] *** Memory Allocation: Fixes ignore_msk during find_next_n() in fb_array library***
@ 2023-01-13 13:14 Vipin P R
  2023-01-13 13:14 ` [PATCH 1/2] Memory Allocation: Fixes ignore_msk during find_next_n() in fb_array library Vipin P R
                   ` (2 more replies)
  0 siblings, 3 replies; 7+ messages in thread
From: Vipin P R @ 2023-01-13 13:14 UTC (permalink / raw)
  To: anatoly.burakov; +Cc: dev, Vipin P R

*** 
In the lookahead logic, let's say after the Right-Shift-And operation to check for contiguity, we hit case http://code.dpdk.org/dpdk/latest/source/lib/eal/common/eal_common_fbarray.c#L235

            /* if first bit is not set, we've lost the run */
            if ((lookahead_msk & 1) == 0) {
                /*
                 * we've scanned this far, so we know there are
                 * no runs in the space we've lookahead-scanned
                 * as well, so skip that on next iteration.
                 */
                ignore_msk = ~((1ULL << need) - 1);
                msk_idx = lookahead_idx;
                break;
            }
        lets say for mask size of 64 bits : in msk_idx 4 we need 4 consecutive bits.
        let need = 4. 
        lets say some of the bits starting from LSB are xx11011. 
        Operating on the inverted mask for better clarity. RSA - RightShiftAnd, xx -- don't-care bits before

        This condition could hit if there aren't "need" number of contiguous bits starting from LSB.
        But, that doesn't necessarily mean there aren't "need" number of such bits elsewhere in the same lookahead_idx.
        We seem to be ignoring "need" number of bits starting from the LSB for the next iteration.

        Due to ignore_mask we might end losing some bits.
        /* if we have an ignore mask, ignore once */
        if (ignore_msk) {
            cur_msk &= ignore_msk;
            ignore_msk = 0;
        }
e.g.
lookahead_msk before RSA logic : xx11100 , need = 4, 2 bits lost
lookahead_msk before RSA logic : xx11011, need = 4, 1 bit lost
lookahead_msk before RSA logic : xx11110, need = 4, 3 bits lost

NB : To understand the number of bits lost, look at need; that's the number of bits (starting from LSB) that's cleared to zero before the next iteration.
***

Vipin P R (2):
  Memory Allocation: Fixes ignore_msk during find_next_n() in fb_array
    library
  Memory Allocation: Alternative fix for ignore_msk during find_next_n()
    in fb_array library

 lib/eal/common/eal_common_fbarray.c | 26 +++++++++++++++++++++++---
 1 file changed, 23 insertions(+), 3 deletions(-)

-- 
2.7.4


^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2023-05-22  8:14 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-01-13 13:14 [PATCH 0/2] *** Memory Allocation: Fixes ignore_msk during find_next_n() in fb_array library*** Vipin P R
2023-01-13 13:14 ` [PATCH 1/2] Memory Allocation: Fixes ignore_msk during find_next_n() in fb_array library Vipin P R
2023-05-19 16:23   ` Burakov, Anatoly
2023-01-13 13:14 ` [PATCH 2/2] Memory Allocation: Alternative fix for " Vipin P R
2023-05-19 16:45   ` Burakov, Anatoly
2023-05-22  5:12     ` Vipin P R
2023-02-20 11:03 ` [PATCH 0/2] *** Memory Allocation: Fixes ignore_msk during find_next_n() in fb_array library*** Thomas Monjalon

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).