linux/drivers/gpu/drm
Alexandre Courbot 69c4938249 drm/nouveau/instmem/gk20a: use direct CPU access
The Great Nouveau Refactoring Take II brought us a lot of goodness,
including acquire/release methods that are called before and after an
instobj is modified. These functions can be used as synchronization
points to manage CPU/GPU coherency if we modify an instobj using the
CPU.

This patch replaces the legacy and slow PRAMIN access for gk20a instmem
with CPU mappings and writes. A LRU list is used to unmap unused
mappings after a certain threshold (currently 1MB) of mapped instobjs is
reached. This allows mappings to be reused most of the time.

Accessing instobjs using the CPU requires to maintain the GPU L2 cache,
which we do in the acquire/release functions. This triggers a lot of L2
flushes/invalidates, but most of them are performed on an empty cache
(and thus return immediately), and overall context setup performance
greatly benefits from this (from 250ms to 160ms on Jetson TK1 for a
simple libdrm program).

Making L2 management more explicit should allow us to grab some more
performance in the future.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-11-03 15:02:18 +10:00
..
amd Merge tag 'topic/drm-misc-2015-10-22' of git://anongit.freedesktop.org/drm-intel into drm-next 2015-10-30 09:49:06 +10:00
armada drm/armada: Convert the probe function to the generic drm_of_component_probe() 2015-10-20 12:01:34 +02:00
ast
atmel-hlcdc drm: Stop using drm_vblank_count() as the hw frame counter 2015-10-07 16:13:52 +02:00
bochs
bridge drm: bridge/dw_hdmi: replace CTS calculation for the ACR 2015-10-09 17:21:21 +01:00
cirrus
exynos Merge tag 'topic/drm-misc-2015-10-19' of git://anongit.freedesktop.org/drm-intel into drm-next 2015-10-20 09:01:49 +10:00
fsl-dcu drm: Stop using drm_vblank_count() as the hw frame counter 2015-10-07 16:13:52 +02:00
gma500 drm/irq: Use unsigned int pipe in public API 2015-10-06 12:57:47 +02:00
i2c Merge branch 'drm-tda998x-devel' of git://ftp.arm.linux.org.uk/~rmk/linux-arm into drm-next 2015-10-15 14:17:13 +10:00
i810
i915 Merge tag 'drm-intel-next-fixes-2015-10-22' of git://anongit.freedesktop.org/drm-intel into drm-next 2015-10-30 09:45:33 +10:00
imx drm/imx: Convert the probe function to the generic drm_of_component_probe() 2015-10-20 12:01:30 +02:00
mga drm/irq: Use unsigned int pipe in public API 2015-10-06 12:57:47 +02:00
mgag200 drm/mgag200: Fix driver_load error handling 2015-09-24 08:10:44 +10:00
msm drm/msm: Remove local fbdev emulation Kconfig option 2015-10-22 15:46:36 -04:00
nouveau drm/nouveau/instmem/gk20a: use direct CPU access 2015-11-03 15:02:18 +10:00
omapdrm drm: Check crtc viewport correctly with rotated primary plane on atomic drivers 2015-10-19 11:00:47 +02:00
panel drm/panel: Add support for LG LG4573 480x800 4.3" panel 2015-08-14 21:35:35 +02:00
qxl Merge tag 'topic/drm-misc-2015-10-19' of git://anongit.freedesktop.org/drm-intel into drm-next 2015-10-20 09:01:49 +10:00
r128 drm/irq: Use unsigned int pipe in public API 2015-10-06 12:57:47 +02:00
radeon Merge tag 'topic/drm-misc-2015-10-22' of git://anongit.freedesktop.org/drm-intel into drm-next 2015-10-30 09:49:06 +10:00
rcar-du drm: rcar-du: Add support for the R8A7794 DU 2015-10-21 18:52:38 +03:00
rockchip drm/rockchip: Convert the probe function to the generic drm_of_component_probe() 2015-10-20 12:01:30 +02:00
savage
shmobile drm: Stop using drm_vblank_count() as the hw frame counter 2015-10-07 16:13:52 +02:00
sis include/uapi/drm/sis_drm.h: move sis_file_private to drivers/gpu/drm/sis/sis_drv.h 2015-10-16 11:27:46 +10:00
sti drm: Stop using drm_vblank_count() as the hw frame counter 2015-10-07 16:13:52 +02:00
tdfx
tegra drm/<drivers>: Drop DRM_UNLOCKED from modeset drivers 2015-10-16 15:50:54 +02:00
tilcdc drm: Stop using drm_vblank_count() as the hw frame counter 2015-10-07 16:13:52 +02:00
ttm drm/ttm: Fix memory space allocation v2 2015-09-15 00:57:01 -07:00
udl
vc4 Merge tag 'drm-vc4-next-2015-10-21' of http://github.com/anholt/linux into drm-next 2015-10-22 10:31:17 +10:00
vgem drm/vgem: Drop vgem_drm_gem_mmap 2015-10-19 11:00:44 +02:00
via via_drm.h: move struct via_file_private definition to drivers/gpu/drm/via/via_drv.h 2015-10-16 11:27:49 +10:00
virtio virtio-gpu: add page flip support 2015-10-16 10:44:03 +02:00
vmwgfx Merge tag 'topic/drm-misc-2015-10-19' of git://anongit.freedesktop.org/drm-intel into drm-next 2015-10-20 09:01:49 +10:00
ati_pcigart.c
drm_agpsupport.c drm: Remove __OS_HAS_AGP 2015-09-30 10:44:48 +02:00
drm_atomic_helper.c drm: Swap w/h when converting the mode to src coordidates for a rotated primary plane 2015-10-19 11:00:46 +02:00
drm_atomic.c drm: drm_atomic_crtc_get_property should be static 2015-09-24 20:14:26 +02:00
drm_auth.c
drm_bridge.c
drm_bufs.c drm: Remove __OS_HAS_AGP 2015-09-30 10:44:48 +02:00
drm_cache.c
drm_context.c
drm_crtc_helper.c
drm_crtc_internal.h
drm_crtc.c drm: correctly check failed allocation 2015-10-21 09:27:30 +02:00
drm_debugfs.c
drm_dma.c
drm_dp_helper.c Linux 4.3-rc2 2015-09-24 17:18:41 +02:00
drm_dp_mst_topology.c Merge tag 'topic/drm-misc-2015-10-19' of git://anongit.freedesktop.org/drm-intel into drm-next 2015-10-20 09:01:49 +10:00
drm_drv.c drm: misc cleanup 2015-10-16 15:50:57 +02:00
drm_edid_load.c drm: Constify generic_edid_names[] 2015-09-08 13:45:52 +02:00
drm_edid.c drm/edid: Round to closest when computing the CEA/HDMI alternate clock 2015-10-20 12:01:29 +02:00
drm_encoder_slave.c
drm_fb_cma_helper.c
drm_fb_helper.c Merge tag 'topic/drm-misc-2015-10-19' of git://anongit.freedesktop.org/drm-intel into drm-next 2015-10-20 09:01:49 +10:00
drm_flip_work.c
drm_fops.c
drm_gem_cma_helper.c drm/gem: Drop struct_mutex requirement from drm_gem_mmap_obj 2015-10-16 15:51:00 +02:00
drm_gem.c drm/gem: Use kref_get_unless_zero for the weak mmap references 2015-10-19 11:00:44 +02:00
drm_global.c
drm_hashtab.c
drm_info.c
drm_internal.h drm: move drm_class into drm_sysfs.c 2015-09-09 15:05:17 +02:00
drm_ioc32.c drm: Remove __OS_HAS_AGP 2015-09-30 10:44:48 +02:00
drm_ioctl.c drm: Enforce unlocked ioctl operation for kms driver ioctls 2015-10-16 15:50:54 +02:00
drm_irq.c drm: Add DRM_DEBUG_VBL() 2015-10-16 15:50:55 +02:00
drm_legacy.h
drm_lock.c
drm_memory.c drm: Remove __OS_HAS_AGP 2015-09-30 10:44:48 +02:00
drm_mipi_dsi.c
drm_mm.c drm/mm: Do DRM_MM_CREATE_TOP adj_start calculation after color_adjust 2015-09-08 13:45:47 +02:00
drm_modes.c
drm_modeset_lock.c drm/atomic: Make sure lock is held in trylock contexts. 2015-09-08 13:45:53 +02:00
drm_of.c drm: Introduce generic probe function for component based masters. 2015-10-20 12:01:29 +02:00
drm_panel.c
drm_pci.c drm/doc: Update docs about device instance setup 2015-09-30 10:06:06 +02:00
drm_plane_helper.c drm/atomic: Make prepare_fb/cleanup_fb only take state, v3. 2015-09-08 13:45:54 +02:00
drm_platform.c drm/doc: Update docs about device instance setup 2015-09-30 10:06:06 +02:00
drm_prime.c
drm_probe_helper.c drm: Add a non-locking version of drm_kms_helper_poll_enable(), v2 2015-09-30 16:04:08 +03:00
drm_rect.c drm: Use DRM_ROTATE_MASK and DRM_REFLECT_MASK 2015-10-05 15:32:51 +02:00
drm_scatter.c
drm_sysfs.c Merge commit '06d1ee32a4d25356a710b49d5e95dbdd68bdf505' of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux into drm-next 2015-10-16 10:25:28 +10:00
drm_trace_points.c
drm_trace.h
drm_vm.c drm: Remove __OS_HAS_AGP 2015-09-30 10:44:48 +02:00
drm_vma_manager.c drm/gem: Use kref_get_unless_zero for the weak mmap references 2015-10-19 11:00:44 +02:00
Kconfig drm/vc4: Add KMS support for Raspberry Pi. 2015-10-21 10:33:12 +01:00
Makefile Merge tag 'drm-vc4-next-2015-10-21' of http://github.com/anholt/linux into drm-next 2015-10-22 10:31:17 +10:00