Hi everyone,

I'm writing a driver for a PCIe device. My goal is to achieve the highest possible writing speed into the PCIe memory region of my device. I'm using write combining on x86 under Linux for this purpose. I would like to know if QNX 6.5.0 supports write combining on x86. I've tried to enable it through shm_ctl() and SHMCTL_LAZYWRITE flag, but it did not yield any results (I got the same writing speeds). That's what I've tried to do:

/**
 * First version
 */
buffer = mmap64(NULL, size, PROT_READ | PROT_WRITE, MAP_PHYS, NOFD, phys_addr);

fd = shm_open(name, O_RDWR | O_CREAT | O_EXCL, 0666);

if (fd >= 0) {

    shm_ctl(fd, SHMCTL_PHYS | SHMCTL_LAZYWRITE, phys_addr, size);
    shm_unlink(name);
    close(fd);
}

/**
 * Second version
 */
if ((fd = shm_open(name, O_RDWR | O_CREAT | O_EXCL, 0666)) == -1)
    return -1;

if (shm_ctl(fd, SHMCTL_PHYS | SHMCTL_LAZYWRITE, phys_addr, size) == -1)
    return -1;

buffer = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);

Could you give me some suggestions?