Communications of the ACM

Martin Sandberg

This article makes a common fundamental mistake - that language are really powerful. They are not. Let me use that analogy of modern machine tools to illustrate my point.

Assembler - a basic hand tool, a file, a chisel, a screwdriver
C - perhaps a drill press
C++ - A table saw with some attachments, such as a mortising jig

We're done. We haven't even reached the level of a multi-axis CNC machine.

Since languages aren't powerful, coders have to do the heavy lifting. Therefor, good and bad code can be written in ANY language. Most GPUs are coded in C with some special operators or library calls, after all. I've written C programs that spawn off threads by the thousands as required by the work load.

In C++ it is even easier. For example it's possible to make thread-safe blocks completely trivial. Here's a complete header file:

// ===========================================================================
//
// A simple way to get locking and unlocking a pthread mutex correct. Just
// put:
// threadLock lockThis(pointer to your pthread mutex)
// at the beginning of the exclusive section of the code and when it goes out
// of scope at the end of the section, the mutex will be unlocked by the
// destructor.
//
// ===========================================================================
// Modification History:
//
// 06/13/2011 - File Created (MCS & WLH)
//
// ***************************************<< o >>***************************************

#include

class threadLock{
public:
threadLock(pthread_mutex_t *theLock):myLock(theLock){pthread_mutex_lock(myLock);};
threadLock(pthread_mutex_t &theLock):myLock(&theLock){pthread_mutex_lock(myLock);};
threadLock(const threadLock& source):myLock(source.myLock){};
threadLock& operator =(const threadLock &rhs){
if(this == &rhs)
return *this;
myLock = rhs.myLock;
return *this;
};

~threadLock(){pthread_mutex_unlock(myLock);};

pthread_mutex_t *myLock;
};