The kmalloc was taking up about 1.5% of the CPU on an ioctl-heavy workload (x11perf -aa10text on 965). Initial results look like they have a corresponding improvement in performance for aa10text, but more numbers might not hurt.
Thanks to ajax for pointing out this performance regression I'd introduced back in 2007. --- drivers/gpu/drm/drm_drv.c | 15 ++++++++++----- 1 files changed, 10 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c index 14c7a23..822d092 100644 --- a/drivers/gpu/drm/drm_drv.c +++ b/drivers/gpu/drm/drm_drv.c @@ -458,6 +458,7 @@ int drm_ioctl(struct inode *inode, struct file *filp, drm_ioctl_t *func; unsigned int nr = DRM_IOCTL_NR(cmd); int retcode = -EINVAL; + char stack_kdata[128]; char *kdata = NULL; atomic_inc(&dev->ioctl_count); @@ -496,10 +497,14 @@ int drm_ioctl(struct inode *inode, struct file *filp, retcode = -EACCES; } else { if (cmd & (IOC_IN | IOC_OUT)) { - kdata = kmalloc(_IOC_SIZE(cmd), GFP_KERNEL); - if (!kdata) { - retcode = -ENOMEM; - goto err_i1; + if (_IOC_SIZE(cmd) <= sizeof(stack_kdata)) { + kdata = stack_kdata; + } else { + kdata = kmalloc(_IOC_SIZE(cmd), GFP_KERNEL); + if (!kdata) { + retcode = -ENOMEM; + goto err_i1; + } } } @@ -520,7 +525,7 @@ int drm_ioctl(struct inode *inode, struct file *filp, } err_i1: - if (kdata) + if (kdata != stack_kdata) kfree(kdata); atomic_dec(&dev->ioctl_count); if (retcode) -- 1.6.2.1 ------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com -- _______________________________________________ Dri-devel mailing list Dri-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dri-devel