Re: [boost] [lock-free] CDS -yet another lock-free library

Subject: Re: [boost] [lock-free] CDS -yet another lock-free library
From: Khiszinsky, Maxim (Maxim.Khiszinsky_at_[hidden])
Date: 2010-03-30 02:57:03

Steven Watanabe wrote:

>> The problem is that the compiler (in some cases) generates case-based code when 'order' parameter is constant for caller:
>>    store( &myAtomic, 10, memory_order_relaxed) ;
>> in this case instead of ONE assembler store instruction the compiler may generate many branch instruction. It is not optimal :-(. And 99% of code with atomic primitives has *constant* memory_order parameter.
>>   

> Have you actually observed this?  On which compilers?

Sorry, my previous post is incorrect.
The problem has been observed on MS VC++ 2008 x86 build with full optimization.
GCC 4.3.3 is my compiler for Unix systems. I did not study the problem on GCC.

-----Original Message-----
From: boost-bounces_at_[hidden] [mailto:boost-bounces_at_[hidden]] On Behalf Of Steven Watanabe
Sent: Monday, March 29, 2010 6:49 PM
To: boost_at_[hidden]
Subject: Re: [boost] [lock-free] CDS -yet another lock-free library

AMDG

Khiszinsky, Maxim wrote:
> 2. Function-based implementation of atomics produces non-optimal code in some cases. Consider the usual implementation of atomic with explicit memory ordering:
> Static inline void store( atomic_t * pDest, atomic_t nVal, memory_order order ) 
> {
>    switch ( order ) {
>       case memory_order_relaxed: *pDest = nVal; break ;
>       case ... 
>       case ... 
>    }
> }
> The problem is that the compiler (in some cases) generates case-based code when 'order' parameter is constant for caller:
>    store( &myAtomic, 10, memory_order_relaxed) ;
> in this case instead of ONE assembler store instruction the compiler may generate many branch instruction. It is not optimal :-(. And 99% of code with atomic primitives has *constant* memory_order parameter.
>   

Have you actually observed this?  On which compilers?

In Christ,
Steven Watanabe

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost