qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH v0 3/3] qcow2: add zstd cluster compression


From: Denis Plotnikov
Subject: Re: [Qemu-devel] [PATCH v0 3/3] qcow2: add zstd cluster compression
Date: Tue, 2 Jul 2019 14:48:36 +0000


On 02.07.2019 17:37, Kevin Wolf wrote:
> Am 02.07.2019 um 14:49 hat Denis Plotnikov geschrieben:
>> On 28.06.2019 14:57, Kevin Wolf wrote:
>>> Am 28.05.2019 um 16:37 hat Denis Plotnikov geschrieben:
>>>> diff --git a/configure b/configure
>>>> index 1c563a7027..c90716189c 100755
>>>> --- a/configure
>>>> +++ b/configure
>>>> @@ -433,6 +433,7 @@ opengl_dmabuf="no"
>>>>    cpuid_h="no"
>>>>    avx2_opt=""
>>>>    zlib="yes"
>>>> +zstd="yes"
>>>
>>> This should be zstd="" so that a missing library will automatically
>>> disable it instead of producing an error. (Building QEMU without zlib is
>>> impossible, but building it without ZSTD should work.)
>> But if we add zstd for clusters compression we have to build with zstd
>> each time. If we want to chose whether we want to build zstd we need to
>> enclose all zstd related code with "ifdef"-s.
>> I don't think it's good because we can end up with mess of version
>> supporting and not supporting zstd compression.
> 
> Yes, we'll need ifdefs. Or we could do it like the dmg compression
> formats, a spearate source file for zstd that is compiled conditionally.
> 
> Anyway, I don't think making zstd a hard dependency for qemu is
> acceptable.
> 
>> Another point is what the benefit of building qemu without zstd support
>> is since it's available and provides better performance than zlib (i.e.
>> the replacement for zlib) ?
> 
> Is it available in all relevant distros and even non-Linux platforms
> that we support?
> 
> Another reason for wanting to compile it out even when it is available
> is startup time. Each shared library to be loaded takes some time, and
> there are use cases where you want a minimal guest to boot up really
> fast.
> 
> Kevin

Sounds like important points. Thanks for the explanation!

> 

-- 
Best,
Denis

reply via email to

[Prev in Thread] Current Thread [Next in Thread]