Junio C Hamano writes:
> I know that the 512MiB default for the bitFileThreashold (aka
> "forget about delta compression") came out of thin air. It was just
> "1GB is always too huge for anybody, so let's cut it in half and
> declare that value the initial version of a sane threashold",
> nothin
Jeff King writes:
> If you have before-and-after numbers for just this patch on some
> repository, that would be an interesting thing to put in the commit
> message.
It's a hen-and-egg problem regarding the benchmarks. The most
impressive benchmarks arise with the git-blame performance work in
On Fri, Mar 21, 2014 at 1:04 PM, David Kastrup wrote:
>> Hmm, doesn't packing need to read existing data?
>
> Judging from the frequent out-of-memory conditions of git gc
> --aggressive, packing is not restrained by deltaBaseCacheLimit.
pack-objects memory usage is more controlled by pack.windowm
Junio C Hamano writes:
> David Kastrup writes:
>
>> Junio C Hamano writes:
>>
>>> Duy Nguyen writes:
>>>
The only
downside I see is large blobs will be packed undeltified, which could
increase pack size if you have lots of them.
>>>
>>> I think that is something that can be twe
Jeff King writes:
> On Wed, Mar 19, 2014 at 01:38:32PM +0100, David Kastrup wrote:
>
>> The default of 16MiB causes serious thrashing for large delta chains
>> combined with large files.
>
> Does it make much sense to bump this without also bumping
> MAX_DELTA_CACHE in sha1_file.c? In my measurem
On Wed, Mar 19, 2014 at 01:38:32PM +0100, David Kastrup wrote:
> The default of 16MiB causes serious thrashing for large delta chains
> combined with large files.
Does it make much sense to bump this without also bumping
MAX_DELTA_CACHE in sha1_file.c? In my measurements of linux.git, bumping
the
David Kastrup writes:
> Junio C Hamano writes:
>
>> Duy Nguyen writes:
>>
>>> The only
>>> downside I see is large blobs will be packed undeltified, which could
>>> increase pack size if you have lots of them.
>>
>> I think that is something that can be tweaked, unless the user tells
>> us oth
Junio C Hamano writes:
> Duy Nguyen writes:
>
>> The only
>> downside I see is large blobs will be packed undeltified, which could
>> increase pack size if you have lots of them.
>
> I think that is something that can be tweaked, unless the user tells
> us otherwise via command line override, w
Duy Nguyen writes:
> On Thu, Mar 20, 2014 at 5:11 AM, Junio C Hamano wrote:
> ...
>> I know that the 512MiB default for the bitFileThreashold (aka
>> "forget about delta compression") came out of thin air. It was just
>> "1GB is always too huge for anybody, so let's cut it in half and
>> declar
On Thu, Mar 20, 2014 at 5:11 AM, Junio C Hamano wrote:
> David Kastrup writes:
>
>> Junio C Hamano writes:
>>
>>> David Kastrup writes:
>>>
The default of 16MiB causes serious thrashing for large delta chains
combined with large files.
Signed-off-by: David Kastrup
---
David Kastrup writes:
> Junio C Hamano writes:
>
>> David Kastrup writes:
>>
>>> The default of 16MiB causes serious thrashing for large delta chains
>>> combined with large files.
>>>
>>> Signed-off-by: David Kastrup
>>> ---
>>
>> Is that a good argument? Wouldn't the default of 128MiB burde
Junio C Hamano writes:
> David Kastrup writes:
>
>> The default of 16MiB causes serious thrashing for large delta chains
>> combined with large files.
>>
>> Signed-off-by: David Kastrup
>> ---
>
> Is that a good argument? Wouldn't the default of 128MiB burden
> smaller machines with bloated pr
David Kastrup writes:
> The default of 16MiB causes serious thrashing for large delta chains
> combined with large files.
>
> Signed-off-by: David Kastrup
> ---
Is that a good argument? Wouldn't the default of 128MiB burden
smaller machines with bloated processes?
> Forgot the signoff. For t
The default of 16MiB causes serious thrashing for large delta chains
combined with large files.
Signed-off-by: David Kastrup
---
Forgot the signoff. For the rationale of this patch and the 128MiB
choice, see the original patch.
Documentation/config.txt | 2 +-
environment.c| 2 +-
14 matches
Mail list logo
