Re: [PATCH v2 0/2] Refactor hash search with fanout table

[Jonathan Tan]

On Tue, 13 Feb 2018 11:57:16 -0800
Junio C Hamano  wrote:

> Jonathan Tan  writes:
> 
> > Updates from v1:
> >  - use uint32_t so that we can operate on packfiles of up to 4G objects
> >(this also means that I had to change the signature of the function)
> >  - don't hide types
> >
> > Derrick: you'll need to slightly change your patch to use the new API.
> > As for find_abbrev_len_for_pack(), that's a good idea - I didn't do it
> > in this set but it definitely should be done.
> >
> > Jonathan Tan (2):
> >   packfile: remove GIT_DEBUG_LOOKUP log statements
> >   packfile: refactor hash search with fanout table
> 
> Hmm, is this meant to replace the topic that was merged to 'next'
> last week?

Yes - René pointed out [1] that V1 of my patch series (which you merged
to 'next') does not handle packfiles of more than 2G pack entries, so I
sent out a new version. Yes, this replaces jt/binsearch-with-fanout.
Sorry for not being clearer.

[1] https://public-inbox.org/git/cfbde137-dbac-8796-f49f-2a543303d...@web.de/

Re: [PATCH v2 0/2] Refactor hash search with fanout table

[Junio C Hamano]

Jonathan Tan  writes:

> Updates from v1:
>  - use uint32_t so that we can operate on packfiles of up to 4G objects
>(this also means that I had to change the signature of the function)
>  - don't hide types
>
> Derrick: you'll need to slightly change your patch to use the new API.
> As for find_abbrev_len_for_pack(), that's a good idea - I didn't do it
> in this set but it definitely should be done.
>
> Jonathan Tan (2):
>   packfile: remove GIT_DEBUG_LOOKUP log statements
>   packfile: refactor hash search with fanout table

Hmm, is this meant to replace the topic that was merged to 'next'
last week?

>
>  packfile.c| 29 -
>  sha1-lookup.c | 28 
>  sha1-lookup.h | 22 ++
>  3 files changed, 54 insertions(+), 25 deletions(-)

Re: [PATCH v2 0/2] Refactor hash search with fanout table

[Derrick Stolee]

On 2/13/2018 1:39 PM, Jonathan Tan wrote:

Updates from v1:
  - use uint32_t so that we can operate on packfiles of up to 4G objects
(this also means that I had to change the signature of the function)


I take it "4G objects" means "4 billion objects". The distinction is 
only important that the limit is not a 4GB IDX file, but in fact a large 
enough object count that the IDX file can be 100+ GB.



  - don't hide types

Derrick: you'll need to slightly change your patch to use the new API.
As for find_abbrev_len_for_pack(), that's a good idea - I didn't do it
in this set but it definitely should be done.


Thanks. I'll try to remember to do that change when this lands in 
master. I'll bring the new prototype into my commit-graph patch.


Reviewed-By: Derrick Stolee 

[PATCH v2 0/2] Refactor hash search with fanout table

[Jonathan Tan]

Updates from v1:
 - use uint32_t so that we can operate on packfiles of up to 4G objects
   (this also means that I had to change the signature of the function)
 - don't hide types

Derrick: you'll need to slightly change your patch to use the new API.
As for find_abbrev_len_for_pack(), that's a good idea - I didn't do it
in this set but it definitely should be done.

Jonathan Tan (2):
  packfile: remove GIT_DEBUG_LOOKUP log statements
  packfile: refactor hash search with fanout table

 packfile.c| 29 -
 sha1-lookup.c | 28 
 sha1-lookup.h | 22 ++
 3 files changed, 54 insertions(+), 25 deletions(-)

-- 
2.16.0.rc1.238.g530d649a79-goog