Index of /archives/NetBSD/NetBSD-release-9/xsrc/external/mit/xf86-video-intel-2014/dist/src/sna
Name Last modified Size Description
Parent Directory -
CVS/ 2020-03-20 07:41 -
brw/ 2020-03-20 07:40 -
fb/ 2020-03-20 07:40 -
sna_module.h 2019-05-10 15:22 54
gen8_vertex.h 2019-05-10 15:22 337
gen4_source.h 2019-05-10 15:22 424
gen4_vertex.h 2019-05-10 15:22 514
gen8_eu.h 2019-05-10 15:22 886
kgem_debug.h 2019-05-10 15:22 1.0K
sna_vertex.c 2019-05-10 15:22 1.4K
debug.h 2019-05-10 15:22 1.5K
xassert.h 2019-05-10 15:22 1.6K
sna_video_hwmc.h 2019-05-10 15:22 1.6K
README 2019-05-10 15:22 1.7K
gen4_common.h 2019-05-10 15:22 1.7K
gen4_common.c 2019-05-10 15:22 2.0K
sna_cpuid.h 2019-05-10 15:22 2.1K
gen6_common.c 2019-05-10 15:22 2.1K
sna_reg.h 2019-05-10 15:22 2.8K
sna_cpu.c 2019-05-10 15:22 2.9K
compiler.h 2019-05-10 15:22 3.0K
atomic.h 2019-05-10 15:22 3.1K
Makefile.am 2019-05-10 15:22 3.6K
sna_stream.c 2019-05-10 15:22 4.0K
gen6_common.h 2019-05-10 15:22 4.6K
sna_acpi.c 2019-05-10 15:22 5.1K
sna_transform.c 2019-05-10 15:22 5.4K
gen4_source.c 2019-05-10 15:22 5.6K
sna_video.h 2019-05-10 15:22 5.7K
rop.h 2019-05-10 15:22 6.3K
sna_video_hwmc.c 2019-05-10 15:22 6.4K
sna_threads.c 2019-05-10 15:22 8.2K
sna_display_fake.c 2019-05-10 15:22 8.5K
sna_render_inline.h 2019-05-10 15:22 8.8K
sna_damage.h 2019-05-10 15:22 8.9K
gen8_vertex.c 2019-05-10 15:22 9.2K
sna_dri3.c 2019-05-10 15:22 10K
sna_video_textured.c 2019-05-10 15:22 11K
sna_trapezoids.h 2019-05-10 15:22 11K
sna_gradient.c 2019-05-10 15:22 12K
sna_present.c 2019-05-10 15:22 12K
kgem_debug_gen7.c 2019-05-10 15:22 17K
kgem_debug.c 2019-05-10 15:22 17K
kgem_debug_gen5.c 2019-05-10 15:22 18K
kgem_debug_gen4.c 2019-05-10 15:22 18K
sna_video.c 2019-05-10 15:22 19K
sna_video_sprite.c 2019-05-10 15:22 20K
kgem_debug_gen2.c 2019-05-10 15:22 20K
sna_render.h 2019-05-10 15:22 21K
kgem.h 2019-05-10 15:22 22K
sna_video_overlay.c 2019-05-10 15:22 23K
gen2_render.h 2019-05-10 15:22 28K
blt.c 2019-05-10 15:22 28K
kgem_debug_gen6.c 2019-05-10 15:22 31K
sna.h 2019-05-10 15:22 31K
sna_tiling.c 2019-05-10 15:22 31K
sna_trapezoids.c 2019-05-10 15:22 32K
sna_composite.c 2019-05-10 15:22 32K
Makefile.in 2019-05-10 15:22 33K
gen8_eu.c 2019-05-10 15:22 34K
sna_driver.c 2019-05-10 15:22 34K
sna_trapezoids_boxes.c 2019-05-10 15:22 38K
sna_trapezoids_mono.c 2019-05-10 15:22 39K
kgem_debug_gen3.c 2019-05-10 15:22 43K
gen8_render.h 2019-05-10 15:22 45K
sna_damage.c 2019-05-10 15:22 45K
sna_io.c 2019-05-10 15:22 50K
gen7_render.h 2019-05-10 15:22 53K
gen3_render.h 2019-05-10 15:22 55K
gen6_render.h 2019-05-10 15:22 59K
sna_glyphs.c 2019-05-10 15:22 60K
sna_render.c 2019-05-10 15:22 61K
gen4_vertex.c 2019-05-10 15:22 78K
gen4_render.h 2019-05-10 15:22 78K
gen5_render.h 2019-05-10 15:22 82K
gen4_render.c 2019-05-10 15:22 86K
sna_trapezoids_precise.c 2019-05-10 15:22 87K
sna_dri2.c 2019-05-10 15:22 88K
gen5_render.c 2019-05-10 15:22 91K
gen2_render.c 2019-05-10 15:22 92K
sna_trapezoids_imprecise.c 2019-05-10 15:22 95K
gen6_render.c 2019-05-10 15:22 99K
gen8_render.c 2019-05-10 15:22 103K
gen7_render.c 2019-05-10 15:22 103K
sna_blt.c 2019-05-10 15:22 114K
gen3_render.c 2019-05-10 15:22 165K
kgem.c 2019-05-10 15:22 187K
sna_display.c 2019-05-10 15:22 194K
sna_accel.c 2019-05-10 15:22 463K
SandyBridge's New Acceleration
------------------------------
The guiding principle behind the design is to avoid GPU context switches.
On SandyBridge (and beyond), these are especially pernicious because the
RENDER and BLT engine are now on different rings and require
synchronisation of the various execution units when switching contexts.
They were not cheap on early generation, but with the increasing
complexity of the GPU, avoiding such serialisations is important.
Furthermore, we try very hard to avoid migrating between the CPU and GPU.
Every pixmap (apart from temporary "scratch" surfaces which we intend to
use on the GPU) is created in system memory. All operations are then done
upon this shadow copy until we are forced to move it onto the GPU. Such
migration can only be first triggered by: setting the pixmap as the
scanout (we obviously need a GPU buffer here), using the pixmap as a DRI
buffer (the client expects to perform hardware acceleration and we do not
want to disappoint) and lastly using the pixmap as a RENDER target. This
last is chosen because when we know we are going to perform hardware
acceleration and will continue to do so without fallbacks, using the GPU
is much, much faster than the CPU. The heuristic I chose therefore was
that if the application uses RENDER, i.e. cairo, then it will only be
using those paths and not intermixing core drawing operations and so
unlikely to trigger a fallback.
The complicating case is front-buffer rendering. So in order to accommodate
using RENDER on an application whilst running xterm without a composite
manager redirecting all the pixmaps to backing surfaces, we have to
perform damage tracking to avoid excess migration of portions of the
buffer.