platform/upstream/intel-gpu-tools.git
9 years agoMediafill/skl: Remove redundant field to fix GPU hang
Zhao Yakui [Thu, 6 Nov 2014 08:02:31 +0000 (16:02 +0800)]
Mediafill/skl: Remove redundant field to fix GPU hang

After applying the commit(982f7eb238a0898c456e0574dee7c4507738d75f), the
OUT_RELOC is updated on Broadwell and later, which is to handle the
64-bit field of gfx address internally. In such case some commands
should be fixed, otherwise GPU hang will be triggered when running
gem_media_fill.  (It is already fixed on Broadwell)

Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
9 years agoRendercopy/skl: Remove redundant field to fix GPU hang
Zhao Yakui [Thu, 6 Nov 2014 08:02:30 +0000 (16:02 +0800)]
Rendercopy/skl: Remove redundant field to fix GPU hang

After applying the commit(982f7eb238a0898c456e0574dee7c4507738d75f), the
OUT_RELOC is updated on Broadwell and later, which is to handle the
64-bit field of gfx address internally. In such case some commands
should be fixed, otherwise GPU hang will be triggered when running
rendercopy.  (It is already fixed on Broadwell)

Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
9 years agotests/drv_hangman: skip a few asserts when using the cmd parser
Brad Volkin [Tue, 4 Nov 2014 22:00:43 +0000 (14:00 -0800)]
tests/drv_hangman: skip a few asserts when using the cmd parser

This test has a few checks that batch buffer addresses in the error
state match the expected  address for the userspace supplied batch.
But the batch buffer copy piece of the command  parser means that
the logged addresses are actually _supposed_ to be different. So
skip just those checks.

Signed-off-by: Brad Volkin <bradley.d.volkin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agotests: Remove spurious binaries from gem_tiled_wb commit
Chris Wilson [Tue, 4 Nov 2014 16:34:40 +0000 (16:34 +0000)]
tests: Remove spurious binaries from gem_tiled_wb commit

commit c45216b26d223ffd7f68c5794870214445a64b5c
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Nov 4 07:30:57 2014 +0000

    igt/gem_tiled_wb: Exercise CPU mmaps with swizzling

brought along a few interlopers.

9 years agotests/prime_nv_test: Skip cpu mmap testcase
Daniel Vetter [Tue, 4 Nov 2014 16:24:55 +0000 (17:24 +0100)]
tests/prime_nv_test: Skip cpu mmap testcase

It's not implemented yet, so no point in running the test really.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68638
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
9 years agoigt/gem_tiled_wb: Remove extraneous mmap(wc) requirement
Chris Wilson [Tue, 4 Nov 2014 12:13:11 +0000 (12:13 +0000)]
igt/gem_tiled_wb: Remove extraneous mmap(wc) requirement

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
9 years agoigt/gem_tiled_wb: Exercise CPU mmaps with swizzling
Chris Wilson [Tue, 4 Nov 2014 07:30:57 +0000 (07:30 +0000)]
igt/gem_tiled_wb: Exercise CPU mmaps with swizzling

This exercises the the extended get_tiling ioctl in order to determine
proper swizzling for direct access to objects through WB.
Userspace cannot handle bit17 swizzling through wc mmaps (because bit17
requires swizzling based on the actual physical address of the page -
which is unknown to userspace) and so we need an extended get_tiling
ioctl to report the actual as well as the logical swizzling on an
object.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
9 years agoioctl_wrappers: Pass in offset to CPU mmaps
Chris Wilson [Tue, 4 Nov 2014 12:06:17 +0000 (12:06 +0000)]
ioctl_wrappers: Pass in offset to CPU mmaps

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
9 years agoigt/gem_linear_blits: Add sufficient RAM check
Chris Wilson [Tue, 4 Nov 2014 07:40:11 +0000 (07:40 +0000)]
igt/gem_linear_blits: Add sufficient RAM check

References: https://bugs.freedesktop.org/show_bug.cgi?id=85834
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
9 years agolib/igt_core.h: add debug messages for test requirements
Thomas Wood [Wed, 29 Oct 2014 12:26:09 +0000 (12:26 +0000)]
lib/igt_core.h: add debug messages for test requirements

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
9 years agolib/igt_core.h: fix igt_skip_on_f requirement message
Thomas Wood [Wed, 29 Oct 2014 12:03:22 +0000 (12:03 +0000)]
lib/igt_core.h: fix igt_skip_on_f requirement message

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
9 years agolib/tests: use the "check_" prefix for tests
Thomas Wood [Wed, 29 Oct 2014 17:20:52 +0000 (17:20 +0000)]
lib/tests: use the "check_" prefix for tests

The "check_" prefix ensures the test programs are not installed and are
only built when "make check" is run.

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
9 years agobuild: fix distcheck
Thomas Wood [Wed, 29 Oct 2014 17:18:06 +0000 (17:18 +0000)]
build: fix distcheck

Fix distcheck issues introduced by commit 685e577 (Move library
selftests to lib/tests).

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
9 years agolib: ensure the library is built before the tests
Thomas Wood [Wed, 29 Oct 2014 16:17:41 +0000 (16:17 +0000)]
lib: ensure the library is built before the tests

This fixes the build problems introduced by commit 685e577 (Move library
selftests to lib/tests).

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
9 years agodemos/intel_sprite_on: Added support to display all sprites.
Gagandeep S Arora [Fri, 12 Sep 2014 19:56:04 +0000 (01:26 +0530)]
demos/intel_sprite_on: Added support to display all sprites.

Extended intel_sprite_on functionality to display all the
available sprite planes on a particular connector.

Signed-off-by: Gagandeep S Arora <gagandeep.s.arora@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
9 years agotests: Reenable testdisplay
Daniel Vetter [Wed, 29 Oct 2014 12:31:15 +0000 (13:31 +0100)]
tests: Reenable testdisplay

This seems to have been accidentally disabled in

commit 982f7eb238a0898c456e0574dee7c4507738d75f
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Aug 29 15:19:57 2014 +0100

    Prepare for 64bit relocation addresses

Apparently no one noticed.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
9 years agotests/testdisplay: Fix test status check fumble
Daniel Vetter [Wed, 29 Oct 2014 12:28:55 +0000 (13:28 +0100)]
tests/testdisplay: Fix test status check fumble

This is a regression from

4306538d1d3f60877866c39c9ca953cc5e541dae is the first bad commit
commit 4306538d1d3f60877866c39c9ca953cc5e541dae
Author:     Daniel Vetter <daniel.vetter@ffwll.ch>
AuthorDate: Thu Oct 2 11:18:20 2014 +0200
Commit:     Daniel Vetter <daniel.vetter@ffwll.ch>
CommitDate: Thu Oct 2 11:34:55 2014 +0200

    tests: Sprinkle missing igt_exit() where needed.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85582
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
9 years agoAdd CONTRIBUTING file
Daniel Vetter [Tue, 28 Oct 2014 09:54:32 +0000 (10:54 +0100)]
Add CONTRIBUTING file

i-g-t has become a fairly big project with lots of people involved, so
lets document the basics and formalize the current process a bit.

Also use this opportunity to announce Thomas Wood as igt maintainer
once more.

v2: Recommend --subject-prefix="PATCH i-g-t" as suggested by Damien.

v3: Clean out contributing-related information from README.

Cc: Thomas Wood <thomas.wood@intel.com>
Cc: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
9 years agoMove library selftests to lib/tests
Daniel Vetter [Mon, 27 Oct 2014 14:08:50 +0000 (15:08 +0100)]
Move library selftests to lib/tests

Again they're not really igt testcases so are in the way of
running spatch unconditionally. Move them someplace else.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agoMove watermark code from tests to tools
Daniel Vetter [Mon, 27 Oct 2014 13:51:55 +0000 (14:51 +0100)]
Move watermark code from tests to tools

They're now igt tests, and so if you blindly run lib/igt.cocci with
spatch on tests/*c they get mangled. Move them away, but still keep
them as noinst targets.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
9 years agotests/pm_rpm: add system-suspend-{execbuf,modeset} subtests
Paulo Zanoni [Fri, 24 Oct 2014 19:40:04 +0000 (17:40 -0200)]
tests/pm_rpm: add system-suspend-{execbuf,modeset} subtests

We were having a problem where the system-suspend test was passing,
but then the next test - gem-execbuf-stress - was failing because of
bugs caused by the suspend subtest. So add a single test that emulates
the same problem, and another test that just sets a mode after resume.
This way, we should be reproducing the problem even if you don't run
the tests in the order they are defined.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
9 years agoskl_ddb_allocation: Respect the minimum number of blocks
Damien Lespiau [Wed, 15 Oct 2014 22:13:30 +0000 (23:13 +0100)]
skl_ddb_allocation: Respect the minimum number of blocks

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
9 years agotests/kms_flip: Make flip-vs-panning-vs-hang change DSPSURF
Ville Syrjälä [Tue, 27 May 2014 18:33:10 +0000 (21:33 +0300)]
tests/kms_flip: Make flip-vs-panning-vs-hang change DSPSURF

Make sure DSPSURF will change during the panning operation
in flip-vs-panning-vs-hang.

This will now test agains bugs between the kernel's mmio vs.
CS flip race handling and GPU resets. If the kernel is buggy
if will fail to notice that the panning operation changed the
base address before the GPU reset had a chance to deal with the
pending page flips, and so the flip would never complete due to
DSPSURFLIVE not matching the expected value.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agoigt/gem_userptr_blits/dmabuf: Map the right pointer for !llc
Chris Wilson [Fri, 24 Oct 2014 06:40:04 +0000 (07:40 +0100)]
igt/gem_userptr_blits/dmabuf: Map the right pointer for !llc

References: https://bugs.freedesktop.org/show_bug.cgi?id=85354#c2
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
9 years agotests/gem_workarounds: adapt to constant wa list from driver
Mika Kuoppala [Mon, 15 Sep 2014 10:01:31 +0000 (13:01 +0300)]
tests/gem_workarounds: adapt to constant wa list from driver

Driver now returns the constant values that were set once
on ring initialization time. Extend the simple interface
check to all gens (i915_wa_registers needs to be present).
Due to delay when render context is visible on cpu side,
wait gpu to execute something on default context to get rid
of undeterministic behaviour by sometimes getting all ones.

Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
9 years agotests/*: lib/igt.cocci found something!
Daniel Vetter [Thu, 23 Oct 2014 15:54:44 +0000 (17:54 +0200)]
tests/*: lib/igt.cocci found something!

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agokms_rotation_crc: Add test for cursor rotation (v2)
Sonika Jindal [Thu, 23 Oct 2014 15:48:50 +0000 (08:48 -0700)]
kms_rotation_crc: Add test for cursor rotation (v2)

v2: Slight rebasing onto latest i-g-t codebase (Matt).

Signed-off-by: Sonika Jindal <sonika.jindal@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agolib: fix #define max
Mika Kuoppala [Mon, 20 Oct 2014 10:27:55 +0000 (13:27 +0300)]
lib: fix #define max

Regression from:

commit be4710a541b517b5f8663448bffed5656d59b47b
Author: Thomas Wood <thomas.wood@intel.com>
Date:   Fri Oct 10 11:20:35 2014 +0100

    lib: add common min and max macros

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85218
Tested-by: Guo Jinxian <jinxianx.guo@intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
9 years agoigt/gem_userptr_blits/dmabuf: Provide partial coverage on !llc platforms
Chris Wilson [Thu, 23 Oct 2014 06:55:54 +0000 (07:55 +0100)]
igt/gem_userptr_blits/dmabuf: Provide partial coverage on !llc platforms

Without LLC, we cannot assume coherency between system RAM and the GTT.
If we try to use the GTT with snooped memory, the machine may lock up,
so the kernel prevents us from doing so, causing the test to fail.
However, we still do want to verify that we can access the userptr
through a dmabuf, so just check that we can make a copy from it using
the GPU and ignore the GTT cross-checking.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85354
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
10 years agolib/igt_aux: move audio RPM code to igt_setup_runtime_pm()
Paulo Zanoni [Tue, 21 Oct 2014 13:45:59 +0000 (11:45 -0200)]
lib/igt_aux: move audio RPM code to igt_setup_runtime_pm()

If we don't enable audio runtime PM, the audio driver won't release
its reference, the refcount won't ever become zero, so we will never
actually runtime suspend. So move this code from pm_rpm.c to
igt_aux.c, so kms_flip - and any other IGT test case using RPM - can
benefit from it.

Previously, if you ran pm_rpm before running the other tests - or if
you just didn't have snd_hda_intel loaded - you wouldn't notice this
bug.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=78893
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
10 years agotests/gem_exec_parse: test for chained batch buffers
Brad Volkin [Wed, 15 Oct 2014 21:52:42 +0000 (14:52 -0700)]
tests/gem_exec_parse: test for chained batch buffers

libva makes extensive use of chained batch buffers. The batch
buffer copy portion of the command parser has the potential to
break chained batches, so add a simple test to make sure that
doesn't happen.

Signed-off-by: Brad Volkin <bradley.d.volkin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agotests/kms_cursor_crc: HSW/BDW only have square cursors
Paulo Zanoni [Tue, 14 Oct 2014 16:52:58 +0000 (13:52 -0300)]
tests/kms_cursor_crc: HSW/BDW only have square cursors

When I look at cursor_size_ok() (from the Kernel's intel_display.c), I
see that only 845g and i865g support non-square cursors, so SKIP the
tests on HSW/BDW instead of failing them.

This problem happened because support for non-square cursors was being
developed for HSW/BDW, but it ended up not being merged due to a small
problem and priority changes. Let's make those tests SKIP until we
have proper support on the Kernel: after that, we can revert this
patch.

v2: Improve commit message.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84268
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
10 years agolib/igt_aux: make igt_wait_for_pm_status() resist the signal helper
Paulo Zanoni [Tue, 14 Oct 2014 19:01:10 +0000 (16:01 -0300)]
lib/igt_aux: make igt_wait_for_pm_status() resist the signal helper

If the signal helper is active, the usleep() calls return earlier, and
we may end up returning false way before the 10s timeout, failing the
subtests. This currently happens on the kms_flip RPM interruptible
subtests.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=78893
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
10 years agotests/kms_force_connector: fix assertion when VGA is already connected
Thomas Wood [Wed, 15 Oct 2014 11:10:13 +0000 (12:10 +0100)]
tests/kms_force_connector: fix assertion when VGA is already connected

Compare the number of modes available when the edid is reset with the
number before the edid was overridden.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82230
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agotests: add a check for command line handling
Thomas Wood [Thu, 16 Oct 2014 16:29:55 +0000 (17:29 +0100)]
tests: add a check for command line handling

Check that command line handling works consistently across all tests.

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agodrm_lib.sh: add standard command line options
Thomas Wood [Thu, 16 Oct 2014 15:35:44 +0000 (16:35 +0100)]
drm_lib.sh: add standard command line options

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agoconfigure: include the skylake quick_dump files in EXTRA_DIST
Thomas Wood [Thu, 16 Oct 2014 13:23:24 +0000 (14:23 +0100)]
configure: include the skylake quick_dump files in EXTRA_DIST

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agoquick_dump: fix distcheck
Thomas Wood [Thu, 16 Oct 2014 12:59:37 +0000 (13:59 +0100)]
quick_dump: fix distcheck

Commit 711a17a (quick_dump: Move base_display.txt to indivual platforms)
renamed base_display.txt to common_display.txt, but didn't update
Makefile.am accordingly. Since common_display.txt is now included in the
platform files, it is automatically included in $(QUICK_DUMP_EXTRA_DIST)
and therefore just the reference to base_display.txt needs removing.

Cc: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agolib: add common min and max macros
Thomas Wood [Fri, 10 Oct 2014 10:20:35 +0000 (11:20 +0100)]
lib: add common min and max macros

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agolib: various documentation fixes
Thomas Wood [Wed, 8 Oct 2014 16:06:58 +0000 (17:06 +0100)]
lib: various documentation fixes

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agoskl_ddb_allocation: Make 'end' exclusive in the DDB allocation entry
Damien Lespiau [Mon, 13 Oct 2014 14:14:15 +0000 (15:14 +0100)]
skl_ddb_allocation: Make 'end' exclusive in the DDB allocation entry

Ville suggested that we should use the same semantics as C arrays to
reduce the number of those pesky +1/-1 in the allocation code.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoskl_ddb_allocation: Add checks on the DDB entries
Damien Lespiau [Mon, 28 Apr 2014 06:11:04 +0000 (07:11 +0100)]
skl_ddb_allocation: Add checks on the DDB entries

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoskl_ddb_allocation: Add a standalone version of the DDB allocator
Damien Lespiau [Mon, 28 Apr 2014 04:56:31 +0000 (05:56 +0100)]
skl_ddb_allocation: Add a standalone version of the DDB allocator

This is useful to debug the allocation algorithm.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agotests/kms_force_connect: skip if a VGA connector is not available
Thomas Wood [Thu, 9 Oct 2014 15:44:12 +0000 (16:44 +0100)]
tests/kms_force_connect: skip if a VGA connector is not available

v2: rename "connector" to "vga_connector" to improve the skip message
    (Chris Wilson)

Suggested-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agotests/kms_force_connector: ensure igt_exit is called at exit
Thomas Wood [Thu, 9 Oct 2014 15:29:20 +0000 (16:29 +0100)]
tests/kms_force_connector: ensure igt_exit is called at exit

Since commit 5782eca (lib/igt_core.c: disable lowmemorykiller during
tests), igt_exit needs to be called before the test exits.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84771
Cc: Tim Gore <tim.gore@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agotests/pm_rc6_residency: Split RC6p and RC6pp.
Rodrigo Vivi [Tue, 14 Oct 2014 17:19:13 +0000 (10:19 -0700)]
tests/pm_rc6_residency: Split RC6p and RC6pp.

sysfs interface for rc6 has been changed to only expose deep RC6 file to the platorms
that really support them, aka SNB and IVB.

So let's split this test and only test RC6p and RC6pp on these platforms.

Reference: https://bugs.freedesktop.org/show_bug.cgi?id=84886
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
10 years agotests/pm_rpm: use igt_assert_eq() on COMPARE()
Paulo Zanoni [Tue, 14 Oct 2014 18:09:22 +0000 (15:09 -0300)]
tests/pm_rpm: use igt_assert_eq() on COMPARE()

A few weeks ago we were hitting this assertion, and the _eq macro is
more useful when you're debugging.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
10 years agotests/pm_rpm: fix some compilation warnings
Paulo Zanoni [Tue, 14 Oct 2014 17:59:36 +0000 (14:59 -0300)]
tests/pm_rpm: fix some compilation warnings

 warning: ignoring return value of ‘write’, declared with attribute
 warn_unused_result [-Wunused-result]

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
10 years agotests/pm_rpm: reduce the number of rounds on the stress tests
Paulo Zanoni [Tue, 14 Oct 2014 17:42:23 +0000 (14:42 -0300)]
tests/pm_rpm: reduce the number of rounds on the stress tests

With the current Kernel, it takes more than 10 minutes to run
modeset-stress-extra-wait on BYT with eDP, and we don't really want
IGT subtests that take more than 10 minutes to run: QA reports them as
bugs.

Since, as far as I remember, any of these RPM race condition bugs we
had could always be triggered with just 4-5 rounds of testing, let's
tune down the number of rounds we test:

Before:
rounds = (--quick option used) ? 10 : 50;

After:
rounds = (--quick option used) ? 10 : 40;
rounds = (wait_flags & WAIT_EXTRA) ? rounds / 2 : rounds;

This should make all subtests be way above the 10m limit.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84897
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
10 years agoigt/gem_bad_reloc: Handle real offset being 0
Chris Wilson [Tue, 14 Oct 2014 09:10:42 +0000 (10:10 +0100)]
igt/gem_bad_reloc: Handle real offset being 0

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
10 years agoigt/gem_negative_reloc: Execute a BLT operation with a negative reloc
Chris Wilson [Tue, 14 Oct 2014 08:27:40 +0000 (09:27 +0100)]
igt/gem_negative_reloc: Execute a BLT operation with a negative reloc

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
10 years agoRevert "tests/kms_sink_crc_basic: Wait 2 vblanks before grabing the new crc."
Rodrigo Vivi [Fri, 10 Oct 2014 15:18:28 +0000 (08:18 -0700)]
Revert "tests/kms_sink_crc_basic: Wait 2 vblanks before grabing the new crc."

This reverts
     commit d92fbc23138b1014e8574daf29dbb06b8c81aa7a.
     Author: Rodrigo Vivi <rodrigo.vivi@intel.com>
     Date:   Tue Sep 9 12:25:47 2014 -0400

         tests/kms_sink_crc_basic: Wait 2 vblanks before grabing the new crc.

Sink CRC is fixed on kernel to wait as many vblanks as needed. It was fixed b
     commit ad9dc91b6e21266bfc6f466db4b95e10211f31ee
     Author: Rodrigo Vivi <rodrigo.vivi@gmail.com>
     Date:   Tue Sep 16 19:18:12 2014 -0400

         drm/i915: Fix Sink CRC

This this fix in place we don't need this extra 2 vblanks on test case itself

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
10 years agoRevert "tests/kms_psr_sink_crc: Wait 2 vblanks before grabing the new crc."
Rodrigo Vivi [Fri, 10 Oct 2014 15:14:32 +0000 (08:14 -0700)]
Revert "tests/kms_psr_sink_crc: Wait 2 vblanks before grabing the new crc."

This reverts
     commit d190a1f9668680c7a429110c91f8faa27949b220.
     Author: Rodrigo Vivi <rodrigo.vivi@intel.com>
     Date:   Wed Sep 3 21:26:29 2014 -0400

           tests/kms_psr_sink_crc: Wait 2 vblanks before grabing the new crc.

Sink CRC is fixed on kernel to wait as many vblanks as needed. It was fixed by:
     commit ad9dc91b6e21266bfc6f466db4b95e10211f31ee
     Author: Rodrigo Vivi <rodrigo.vivi@gmail.com>
     Date:   Tue Sep 16 19:18:12 2014 -0400

         drm/i915: Fix Sink CRC

This this fix in place we don't need this extra 2 vblanks on test case itself.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
10 years agotools/null_state_gen: Add copyright notice to state output
Mika Kuoppala [Fri, 10 Oct 2014 13:59:00 +0000 (16:59 +0300)]
tools/null_state_gen: Add copyright notice to state output

along with info about what generated it.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
10 years agotools/null_state_gen: Add GEN9 golden context batch buffer creation
Armin Reese [Fri, 5 Sep 2014 20:58:11 +0000 (13:58 -0700)]
tools/null_state_gen: Add GEN9 golden context batch buffer creation

Modifications to 'null_state_gen' so it can generate GEN9
golden context batch buffer source for SKL.

v2: - rebased on top of gen8 changes (Mika)
    - fixed state base address command size (Mika)
    - base address size macro as pages (Mika)

v3: - rebased on top of current master (Mika)
    - removed obsolete #includes (Mika)
    - added copyright (Mika)
    - render and component packing added (Mika)

Cc: Damien Lespiau <damien.lespiau@intel.com>
Cc: Armin Reese <armin.c.reese@intel.com>
Cc: Volkin, Bradley D <bradley.d.volkin@intel.com>
Reviewed-by: Volkin, Bradley D <bradley.d.volkin@intel.com> (v2)
Signed-off-by: Armin Reese <armin.c.reese@intel.com> (v1)
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
10 years agotools/null_state_gen: Add Gen8 golden state
Mika Kuoppala [Mon, 8 Sep 2014 07:49:59 +0000 (10:49 +0300)]
tools/null_state_gen: Add Gen8 golden state

Previously we didn't have a clear understanding what is necessary
for a pipeline state to be properly initialized. So we had to improvise
and use a stripped out render copy.

Now we have a more clear understanding so switch out render copy based
frankenstate to state we can call golden state.

v2: - export intel_batch_state_offset
    - add 3DSTATE_RASTER (Bradley Volkin)

Cc: Volkin, Bradley D <bradley.d.volkin@intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
10 years agotools/null_state_gen: Add macro to emit commands with null state
Mika Kuoppala [Mon, 8 Sep 2014 07:41:38 +0000 (10:41 +0300)]
tools/null_state_gen: Add macro to emit commands with null state

In null/golden context there are multiple state commands where
the actual state is always zero. For more compact batch representation
add a macro which just emits command and the rest of the state as zero.

v2: - Be more verbose about length bias (Bradley Volkin)
    - strip out unrelated state_offset declaration (Bradley Volkin)

Cc: Volkin, Bradley D <bradley.d.volkin@intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
10 years agotools/null_state_gen: Limit the total state len to 4096 bytes
Mika Kuoppala [Thu, 9 Oct 2014 15:39:05 +0000 (18:39 +0300)]
tools/null_state_gen: Limit the total state len to 4096 bytes

Currently our kernel side buffer object is only one page.
Limit the amount of dwords to 1024 to enforce this.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
10 years agotools/null_state_gen: Add more debug output
Mika Kuoppala [Thu, 9 Oct 2014 15:35:49 +0000 (18:35 +0300)]
tools/null_state_gen: Add more debug output

Be more verbose about the state size we generate.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
10 years agotools/null_state_gen: Add copyrights
Mika Kuoppala [Thu, 9 Oct 2014 16:10:48 +0000 (19:10 +0300)]
tools/null_state_gen: Add copyrights

to files where they were missing.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
10 years agoDon't use += to append to a shell variable.
Adam Sampson [Tue, 7 Oct 2014 16:41:18 +0000 (17:41 +0100)]
Don't use += to append to a shell variable.

POSIX only requires "=" to be supported; "+=" works in bash but not in
dash.

Signed-off-by: Adam Sampson <ats@offog.org>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agoUse = rather than == in test.
Adam Sampson [Tue, 7 Oct 2014 16:41:17 +0000 (17:41 +0100)]
Use = rather than == in test.

The POSIX standard only requires test to support "="; "==" works in bash
but not in dash. Other comparisons in configure.ac use "=" already.

Signed-off-by: Adam Sampson <ats@offog.org>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agoigt/kms_flip/nonblocking_read: Demonstrate that O_NONBLOCK is a myth
Chris Wilson [Wed, 8 Oct 2014 09:44:06 +0000 (10:44 +0100)]
igt/kms_flip/nonblocking_read: Demonstrate that O_NONBLOCK is a myth

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
10 years agoquick_dump: Drop common_display.txt from VLV/CHV
Damien Lespiau [Wed, 8 Oct 2014 09:25:48 +0000 (10:25 +0100)]
quick_dump: Drop common_display.txt from VLV/CHV

Those registers are big core registers and weren't really relevant for
VLV/CHV. Drop them from the dump then.

Suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoquick_dump/skl: Make quick_dump SKL aware
Damien Lespiau [Mon, 6 Oct 2014 16:42:33 +0000 (17:42 +0100)]
quick_dump/skl: Make quick_dump SKL aware

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoquick_dump/skl: Add some display registers
Damien Lespiau [Mon, 6 Oct 2014 14:10:35 +0000 (15:10 +0100)]
quick_dump/skl: Add some display registers

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoquick_dump: Move base_display.txt to indivual platforms
Damien Lespiau [Mon, 6 Oct 2014 17:32:51 +0000 (18:32 +0100)]
quick_dump: Move base_display.txt to indivual platforms

SKL will have a whole separate display regs file, so merge
base_display.txt into each platform file.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoquick-dump: Make quick dump link against libintel_tools
Damien Lespiau [Mon, 6 Oct 2014 15:30:39 +0000 (16:30 +0100)]
quick-dump: Make quick dump link against libintel_tools

Because quick-dump was only selecting a few files in lib/ and we move
stuff around and/or add new dependencies we were failing to provide the
necessary symbols to the shim library providing python bindings. And so
we had a run-time error:

Traceback (most recent call last):
  File "./tools/quick_dump/quick_dump.py", line 17, in <module>
    import chipset
  File "/home/damien/gfx/sources/intel-gpu-tools/tools/quick_dump/chipset.py", line 28, in <module>
    _chipset = swig_import_helper()
  File "/home/damien/gfx/sources/intel-gpu-tools/tools/quick_dump/chipset.py", line 24, in swig_import_helper
    _mod = imp.load_module('_chipset', fp, pathname, description)
  File "/usr/lib64/python3.3/imp.py", line 183, in load_module
    return load_dynamic(name, filename, file)
ImportError: /home/damien/gfx/sources/intel-gpu-tools/tools/quick_dump/_chipset.so: undefined symbol: kmstest_pipe_name

So, let's simplify maintainance and just link against the library we're
building and using elsewhere.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agogem_wait: Use PRIu64 in format string
Damien Lespiau [Mon, 6 Oct 2014 15:49:05 +0000 (16:49 +0100)]
gem_wait: Use PRIu64 in format string

../../tests/gem_wait.c: In function ‘render_timeout’:
../../tests/gem_wait.c:182:3: warning: format ‘%llu’ expects argument of type
‘long long unsigned int’, but argument 3 has type ‘uint64_t’ [-Wformat=]

   igt_info("Finished with %llu time remaining\n", timeout);

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agooverlay: Fix compilation warning when not having xrandr
Damien Lespiau [Mon, 6 Oct 2014 17:23:33 +0000 (18:23 +0100)]
overlay: Fix compilation warning when not having xrandr

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agogem_seqno_wrap: Remove unused variable
Damien Lespiau [Mon, 6 Oct 2014 17:19:58 +0000 (18:19 +0100)]
gem_seqno_wrap: Remove unused variable

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agotests/gem_wait: Don't close drmfd in subtest
Daniel Vetter [Mon, 6 Oct 2014 06:32:49 +0000 (08:32 +0200)]
tests/gem_wait: Don't close drmfd in subtest

I didn't notice this on the machine I developed it since the original
wait testcase fails there. Oops.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agolib/Android.mk: add define of IGT_DATADIR for igt_fb.c
Tim Gore [Thu, 2 Oct 2014 15:26:32 +0000 (16:26 +0100)]
lib/Android.mk: add define of IGT_DATADIR for igt_fb.c

Since the igt_create_stereo_fb function moved into igt_fb.c
we need to add the definition of IGT_DATADIR in the Android
makefile in order to build the cairo dependent tests.

Signed-off-by: Tim Gore <tim.gore@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agotests: Sprinkle missing igt_exit() where needed.
Daniel Vetter [Thu, 2 Oct 2014 09:18:20 +0000 (11:18 +0200)]
tests: Sprinkle missing igt_exit() where needed.

Yay for breaking piles of tests.

This regression has been introduced with

commit 5782eca1e19e85a04ad402fa4094aa1b5f9c53ce
Author: Tim Gore <tim.gore@intel.com>
Date:   Wed Oct 1 13:25:20 2014 +0100

    lib/igt_core.c: disable lowmemorykiller during tests

Cc: Tim Gore <tim.gore@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
10 years agolib/igt_core.c: disable lowmemorykiller during tests
Tim Gore [Wed, 1 Oct 2014 12:25:20 +0000 (13:25 +0100)]
lib/igt_core.c: disable lowmemorykiller during tests

Several IGT tests cycle through a lot of GEM memory and
when running these tests on Android they tend to get
killed by the lowmemorykiller. The lowmemorykiller really
is not usefull in this context and is just preventing the
test from doing its job. This commit adds a function to
disable the lowmemorykiller by writing "9999" to its
oom adj parameter, which means it will never "select"
any process to kill. The normal linux oom killer is still
there to protect the kernel.
The low memory killer is disabled during the common
init function and then re-enabled by the exit handler.
To make this work for single tests the exit handler is now
installed in the common init function also so that it is
invoked for all tests.
This is just a hack to get round the fact that the i915
driver uses the concept of purgeable memory which is not
understood by the lowmemorykiller. If this ever gets
fixed then this patch can be removed.

Signed-off-by: Tim Gore <tim.gore@intel.com>
[danvet: A bit of polish (coding style, static and rename
check_igt_exit to common_exit_handler, drop the is_disabled check
since we only call this once now).]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agoNEWS: Updates
Daniel Vetter [Wed, 1 Oct 2014 10:12:43 +0000 (12:12 +0200)]
NEWS: Updates

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
10 years agolib/aux: Print progress output at INFO level
Daniel Vetter [Tue, 30 Sep 2014 16:44:15 +0000 (18:44 +0200)]
lib/aux: Print progress output at INFO level

With the structured logging it makes more sense to tune this down a
bit. Also, this way it is consistent with Thomas Wood's new activity
indicator helper.

Spotted while discussing Thomas' patch with him.

v2: Thomas noticed that I've forgotten the fflush. Extract
igt_interactive_info for both igt_progress and igt_print_activity.

v3: Interactive output should go to stderr. Also extract the "is this
a terminal" check.

Cc: Thomas Wood <thomas.wood@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
10 years agolib: add a function to indicate activity
Thomas Wood [Tue, 30 Sep 2014 16:05:39 +0000 (17:05 +0100)]
lib: add a function to indicate activity

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agotests/sysfs_l3_parity: fix warnings in test enumeration
Thomas Wood [Mon, 29 Sep 2014 11:08:31 +0000 (12:08 +0100)]
tests/sysfs_l3_parity: fix warnings in test enumeration

Source drm_lib.sh before skipping the test to ensure that subtest
enumeration is always handled correctly.

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agolib: ensure any buffers are flushed before fork
Thomas Wood [Fri, 26 Sep 2014 14:16:21 +0000 (15:16 +0100)]
lib: ensure any buffers are flushed before fork

Flush any buffers before forking to prevent duplicated output.

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agodoc: various spelling and typo fixes
Thomas Wood [Fri, 26 Sep 2014 13:24:52 +0000 (14:24 +0100)]
doc: various spelling and typo fixes

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
10 years agotest/kms, pm: use drm_open_any_master
Imre Deak [Thu, 18 Sep 2014 15:31:29 +0000 (18:31 +0300)]
test/kms, pm: use drm_open_any_master

These tests require DRM master right, so make sure they have it from the
beginning. This gives an early indication if another DRM master is running
and makes the given test skip (with a proper explanation of the reason)
instead of exiting with error.

Signed-off-by: Imre Deak <imre.deak@intel.com>
10 years agotests/pm_rpm: factor out drm_open_any_master
Imre Deak [Thu, 18 Sep 2014 15:25:37 +0000 (18:25 +0300)]
tests/pm_rpm: factor out drm_open_any_master

Signed-off-by: Imre Deak <imre.deak@intel.com>
10 years agolib/skl: Return the render copy and media fill functions
Xiang, Haihao [Fri, 11 Apr 2014 13:57:21 +0000 (21:57 +0800)]
lib/skl: Return the render copy and media fill functions

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
[Ben: Reordered if tree]
Signed-off-by: Ben Widawsky <benjamin.widawsky@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agomediafill/skl: Follow spec to configure media sampler DOP clock gating in PIPELINE_SE...
Zhao Yakui [Tue, 11 Feb 2014 02:48:47 +0000 (10:48 +0800)]
mediafill/skl: Follow spec to configure media sampler DOP clock gating in PIPELINE_SELECTION

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agomediafill/skl: Follow spec to configure FORCE_MEDIA_AWAKE in PIPELINE_SELECTION
Zhao Yakui [Tue, 11 Feb 2014 02:48:46 +0000 (10:48 +0800)]
mediafill/skl: Follow spec to configure FORCE_MEDIA_AWAKE in PIPELINE_SELECTION

The FORCE_MEDIA_AWAKE bit is added for the PIPELINE_SELECTION command and
some instructions requires that the media enginee is awake.

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agomediafill/skl: Follow the spec to add pipeline_select mask
Zhao Yakui [Tue, 11 Feb 2014 02:48:45 +0000 (10:48 +0800)]
mediafill/skl: Follow the spec to add pipeline_select mask

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agomediafill/skl: follow the spec to update STATE_BASE_ADDRESS command
Zhao Yakui [Tue, 11 Feb 2014 02:48:44 +0000 (10:48 +0800)]
mediafill/skl: follow the spec to update STATE_BASE_ADDRESS command

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agomediafill/skl: Start the gen9 media_fill from the gen8 version
Zhao Yakui [Tue, 11 Feb 2014 02:48:43 +0000 (10:48 +0800)]
mediafill/skl: Start the gen9 media_fill from the gen8 version

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoassembler/skl: update the extdesc field for SEND instruction
Zhao Yakui [Thu, 23 Jan 2014 05:26:12 +0000 (13:26 +0800)]
assembler/skl: update the extdesc field for SEND instruction

The send instruction on gen9 uses the 32bit immediate instead of 6bit immediate
for the extended message descriptors. And some bits of SEND instruction are defined
as the extdesc field.

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoassembler/skl: Add more cache agent for write(...)
Zhao Yakui [Thu, 23 Jan 2014 05:26:11 +0000 (13:26 +0800)]
assembler/skl: Add more cache agent for write(...)

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoassembler/skl: update read(...)
Zhao Yakui [Thu, 23 Jan 2014 05:26:10 +0000 (13:26 +0800)]
assembler/skl: update read(...)

READ(...) is used for Render Target read and Media Block read. But there is no
sampler cache agent on gen9. At the same time two message types don't
share the same cache agent any more. So a parameter is needed for cache agent.
The 2th parameter of read(...) is not used for gen6/gen7/gen8. Hence it is
reused as cache agent for SKL as that on ILK.

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoassembler/skl: Redefine the cache agent type for some fixed functions
Zhao Yakui [Thu, 23 Jan 2014 05:26:09 +0000 (13:26 +0800)]
assembler/skl: Redefine the cache agent type for some fixed functions

The different cache agent type is defined for SKL although it still uses
the same function ID as the previous generations.

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agoassembler/skl: Add gen 9 to the -g option
Damien Lespiau [Tue, 5 Feb 2013 12:10:44 +0000 (12:10 +0000)]
assembler/skl: Add gen 9 to the -g option

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agorendercopy/skl: Set the URB VS start address to 4
Xiang, Haihao [Fri, 11 Apr 2014 13:57:20 +0000 (21:57 +0800)]
rendercopy/skl: Set the URB VS start address to 4

A value less than 4 might result in GPU hang on simulation

Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agorendercopy/skl: Follow the spec to add the Pipeline selection mask
Zhao Yakui [Wed, 22 Jan 2014 01:37:26 +0000 (09:37 +0800)]
rendercopy/skl: Follow the spec to add the Pipeline selection mask

Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agorendercopy/skl: Emit 3DSTATE_WM_HZ_OP
Zhao Yakui [Wed, 22 Jan 2014 01:37:25 +0000 (09:37 +0800)]
rendercopy/skl: Emit 3DSTATE_WM_HZ_OP

This is from that on BDW. Without it, the pixel pipeline can't work well.

Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agorendercopy/skl: Fix the 3DSTATE_DS instruction length
Zhao Yakui [Wed, 22 Jan 2014 01:37:24 +0000 (09:37 +0800)]
rendercopy/skl: Fix the 3DSTATE_DS instruction length

Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agorendercopy/skl: Fix the STATE_BASE_ADDRESS instruction length
Zhao Yakui [Wed, 22 Jan 2014 01:37:23 +0000 (09:37 +0800)]
rendercopy/skl: Fix the STATE_BASE_ADDRESS instruction length

Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
10 years agorendercopy/skl: Set Instruction Buffer size Modify Enable to 1
Zhao Yakui [Wed, 22 Jan 2014 01:37:22 +0000 (09:37 +0800)]
rendercopy/skl: Set Instruction Buffer size Modify Enable to 1

This is ported from that on BDW.

Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>