Re: [microblaze-uclinux] xenet_FifoSend(struct sk_buff *orig_skb,structnet_device *dev)

["Jim Law" <jlaw@xxxxxxxxxxxxx>]

Hi again,

I guess it depends on the size of the buffers being checksummed: i.e. 
overhead vs. time in the inner loop.  Any idea of the average size of the 
buffers being worked on?

For the last version of the code i sent, my inner loop is 8 cycles long and 
the gcc-produced version is 12 cycles.  On big buffers, the time should 
approach about 2/3 of the gcc version.  Inline here is a version that does 
it in 7 cycles - yeah, I'm playing around <G>.

I 'm not sure how such things are viewed around the linux world: whether or 
not the maintainability aspects of C trumps putting such (un-necessary) 
assembler stuff in the code.  I'm playing with an assembler memcpy at the 
moment, on the theory that it may be asked to move larger blocks more often 
than do_csum is asked to work on large blocks - let me know if there's any 
interest.

Regards,

Jim Law

########## new optimized inner csum loop ##########
2: # when get here, then more than one word to be summed
addi r11,r11,-1 # word count = word count - 1
addi r0,r0,0  # clear carry for add with carry in loop

beqi r11,4f  # if no more words to do don't enter loop
5: lwi r4,r7,0  # temp = *word address
addc r3,r3,r4 # csum = csum + temp + carry
addik r11,r11,-1 # word count = word count - 1, don't disturb carry
bneid r11,5b  # loop back if still some words to do
addik r7,r7,4  # word address++, don't disturb carry - IN DELAY SLOT

4: # deal with last (possibly partial) word
lwi r4,r7,0  # temp = *word address
and r4,r4,r10 # temp = temp & endmask
addc r3,r3,r4 # csum = csum + temp + carry
#############################################


----- Original Message ----- 
From: "Brettschneider Falk" <fbrettschneider@xxxxxxxxxxxxxxx>
To: <microblaze-uclinux@xxxxxxxxxxxxxx>
Sent: Monday, April 28, 2008 12:01 PM
Subject: RE: [microblaze-uclinux] xenet_FifoSend(struct sk_buff 
*orig_skb,structnet_device *dev)


> Hi Jim,
>
> Jim Law wrote:
> > I had a look at the do_csum() routine in the
> > arch/microblaze/lib/checksum.c file and produced an assembler
> > version in the
> > hopes of providing some speedup.
> Thanks very much. Though it seems to me mb-gcc produces almost perfect 
> assembler instructions since the time is still by about 100us for 
> computing the checksum. Anyway at first sight... I'm going to compare more 
> in detail more later...
>
> Cheers, Falk
>
> ___________________________
> microblaze-uclinux mailing list
> microblaze-uclinux@xxxxxxxxxxxxxx
> Project Home Page : http://www.itee.uq.edu.au/~jwilliams/mblaze-uclinux
> Mailing List Archive : 
> http://www.itee.uq.edu.au/~listarch/microblaze-uclinux/

___________________________
microblaze-uclinux mailing list
microblaze-uclinux@xxxxxxxxxxxxxx
Project Home Page : http://www.itee.uq.edu.au/~jwilliams/mblaze-uclinux
Mailing List Archive : http://www.itee.uq.edu.au/~listarch/microblaze-uclinux/