This is around 3x or so speedup, since we would read wide rows at a time, and
clflush each tile 8 times as a result. We'll want code related to this anyway
when we do fault-based per-page clflushing for sw fallbacks.
Various chips have exciting interactions between the CPU and the GPU's
different ways of accessing interleaved memory, so we need some kernel
assistance in determining how it works.
Only fully tested on GM965 so far.
With the interrupt enable/disable using only the mask register, it was wrong
to use the enable register to detect which pipes had vblank detection
turned on. Also, as we keep a local copy of the mask register around, and
MSI machines smack the hardware during the interrupt handler, it is more
efficient and more correct to use the local copy.
Clean up queues, free objects. On the next entervt, unmark the hardware to
let the user try again (presumably after resetting the chip). Someday we'll
automatically recover...
Noting that the interrupt mask register was more reliable than the interrupt
enable register for managing interrupts in user_irq_on/user_irq_off, this
patch replaces the remaining IER frobbing with IMR instead.
The test which exposes IER related failures is:
$ glxgears & glxgears & glxgears
(reposition the glxgears windows away from the upper left corner)
$ while :; do x11perf -rect100 -reps 800 -repeat 1; sleep 1; done &
$ while :; do runoa; runet; done &
find_or_create_page doesn't quite set up pages correctly; any newly created
pages aren't hooked into the shmem object quite right; user space mmaps of
those pages end up mapping pages full of zeros which then get written to the
real pages inappropriately. This patch requires that the kernel export
shmem_getpage.
When a software fallback has completed, usermode must notify the kernel so
that any scanout buffers can be synchronized. This ioctl should be called
whenever a fallback completes to flush CPU and chipset caches.
This fixes registration when MSI is set up after the stub function fills in
dev->irq. Otherwise /proc/interrupts would report attachment to the fasteoi
interrupt. dev->irq is still exposed (and updated at IRQ setup)
for the drivers that use it for whatever reason.
Record the last execbuffer sequence for each client.
Record that sequence in the throttle ioctl as the 'throttle sequence'.
Wait for the last throttle sequence in the throttle ioctl.
The interrupt identity register must be writen before any work occurs lest
we drop an interrupt on the floor. This patch just shuffles code around to
make sure that IIR is written as early as possible.
We want request retirement to occur about once a second when the request
queue is non-empty. This was done with a timer that queued a work_struct,
using a delayed_work instead makes a lot more sense.
In the short-circuit code for the breadcrumb already being new enough, we
need to update the sarea_priv copy of the breadcrumb just as if we had
waited. Otherwise userland error checking will notice that we returned
too early based on its wrong information, and call wait_irq again (leading
to spinning until someone else comes along and updates the sarea_priv).
This bug was hidden when we had interrupt masking disabled, such as in
master, since the interrupt handler would update sarea_priv.
This was insufficient once we started masking interrupts to only when someone
was waiting for them (and would thus retire requests themselves). It was
replaced by the retire_timer.
This is the create (may want location flags), pread/pwrite/mmap
(performance tuning hints), and set_domain (will 32 bits be enough for
everyone?) ioctls. Left in the generic set are just flink/open/close.
The 2D driver must be updated for this change, and API but not ABI is broken
for 3D. The driver version is bumped to mark this.
This requires that the X Server use the execbuf interface for buffer
submission, as it no longer has direct access to the ring. This is
therefore a flag day for the gem interface.
This also adds enter/leavevt ioctls for use by the X Server. These would
get stubbed out in a modesetting implementation, but are required while
in an environment where the device's state is only managed by the DRM while
X has the VT.
Without the user IRQ running constantly, there's no wakeup when the ring
empties to go retire requests and free buffers. Use a 1 second timer to make
that happen more often.
Instead of throttling and execbuffer time, have the application ask to
throttle explicitly. This allows the throttle to happen less often, and
without holding the DRM lock.
The interrupt enable register cannot be used to temporarily disable
interrupts, instead use the interrupt mask register.
Note that this change means that a pile of buffers will be left stuck on the
chip as the final interrupts will not be recognized to come and drain things.
Recording the tail pointer in a local variable improves performance, but if
someone messes up and fails to reload at the right time, the driver will
write commands to the wrong part of the ring and scramble execution badly.
This change (available by setting I915_RING_VALIDATE to 1) checks to make
sure the cached tail pointer matches the hardware tail pointer at each ring
buffer addition, calling BUG_ON when that's not true.
There are now 3 lists. Active is buffers currently in the ringbuffer.
Flushing is not in the ringbuffer, but needs a flush before unbinding.
Inactive is as before. This prevents object_free → unbind →
wait_rendering → object_reference and a kernel oops about weird refcounting.
This also avoids an synchronous extra flush and wait when freeing a buffer
which had a write_domain set (such as a temporary rendered to and then from
using the 2d engine). It will sit around on the flushing list until the
appropriate flush gets emitted, or we need the GTT space for another
operation.
This lets us get some qualities we desire, such as using the full 32-bit
range (except zero), avoiding DRM_WAIT_ON, and a 1:1 mapping of active
sequence numbers to request structs, which will be used soon for throttling
and interrupt-driven list cleanup.
Additionally, a boolean active field is added to indicate which list an
object is on, rather than smashing last_rendering_cookie to 0 to show
inactive. This will help with flush-reduction later on, and makes the code
clearer.
Eric Anholt
Keith Packard
Dave Airlie
Dave Airlie
Alex Deucher
Jie Luo
Robert Noland