Add PCI IDs for the Intel Kabylake platforms. The IDs are taken
directly from the Linux kernel patches, which are under review:
http://lists.freedesktop.org/archives/intel-gfx/2015-October/078967.htmlhttp://cgit.freedesktop.org/~vivijim/drm-intel/log/?h=kbl-upstream-v2
The Kabylake PCI IDs taken from the kernel are rearranged to be in order
of GT type, then PCI ID.
Please note that if this patch is backported, the following fixes will
need to be added before this patch:
commit 28ed1e08e8 "i965/skl: Remove early platform support"
commit c1e38ad370 "i965/skl: Use larger URB size where available."
Thanks to Ben for fixing a bug around setting urb.size, and being
patient with my questions about what the various fields mean.
Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Suggested-by: Ben Widawsky <benjamin.widawsky@intel.com>
Tested-by: Rodrigo Vivi <rodrigo.vivi@intel.com> (KBL-GT2)
Cc: "11.1" <mesa-stable@lists.freedesktop.org>
Rename SVGA_HINT_FLAG_DRAW_EMITTED to SVGA_HINT_FLAG_CAN_PRE_FLUSH
because preemptive flush can be unblocked by more commands than
draw.
Reviewed-by: Brian Paul <brianp@vmware.com>
The existing code effectively turns off preemptive flushing for all
but the regions used for draws. This turns out to be overly
restrictive as some memory regions, e.g. GMR, may never get a draw
when used as a DMA upload staging area, causing problems for apps
that upload a large amount of textures, e.g. Unigine Heaven.
This patch fixes the Unigine Heaven memory allocation error and
has been verified to not cause a regression in the previous extended
retina display issue.
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
In emit_input_declarations(), we are skipping declarations for those
registers that are not being used. But in emit_vertex_attrib_instructions(),
we are still emitting instructions to tweak the vertex attributes even if
they are not being used. This causes an assert in the backend because an
input register is not declared in the shader. This patch fixes the problem
by skipping the instruction if the vertex attribute is not being used.
Changes in this patch is originated from the code snippet from Jose as
suggested in bug 1530161.
Tested with piglit, Heaven, Turbine, glretrace.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
If the only dirty state is mesa's _NEW_PROGRAM_CONSTANTS flag, we can
skip state validation before drawing a bitmap since that state doesn't
effect bitmap rendering.
This further increases the performance of the ipers demo on llvmpipe
to about what it was before commit 36c93a6fae.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
We were checking the dirty->st flags but not the dirty->mesa flags.
When we took the early return, we didn't clear the dirty->mesa flags
so the next time we called st_validate_state() we'd often flush the
glBitmap cache. And since st_validate_state() is called from
st_Bitmap(), it meant we flushed the bitmap cache for every glBitmap()
call.
This change seems to recover most of the performance loss observed
with the ipers demo on llvmpipe since commit commit 36c93a6fae.
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Previously each member was being counted as using a single slot,
count_attribute_slots() fixes the count for array and struct members.
Also don't assign a negitive to the unsigned expl_location variable.
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Previously we were only reserving a single location for arrays and
structs.
We also didn't take into account implicit locations clashing with
explicit locations when assigning locations for their arrays or
structs.
This patch fixes both issues.
V5: fix regression for patch inputs/outputs in tessellation shaders
V4: just use count_attribute_slots() to get the number of slots,
also calculate the correct number of slots to reserve for gs and
tess stages by making use of the new get_varying_type() helper.
V3: handle arrays of structs
V2: also fix for arrays of arrays and structs.
Acked-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
This will be used in the following patch for calculating array sizes correctly
when reserving explicit varying locations.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Previously we would pack varyings before trying to remove them, this
relied on the packing pass not packing varyings with a location of -1
to avoid packing varyings that should be removed.
However this meant unused varyings with an explicit location would be
packed before they could be removed when we enable packing of them in a
later patch.
V2: fix regression in V1 removing unused varyings in multi-stage SSO,
fix regression with single stage programs.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
ALIGN_DIVUP is a driver specific(r600g) macro that duplicates DIV_ROUND_UP functionality.
Replacing it with DIV_ROUND_UP eliminates this problems.
Signed-off-by: Krzysztof A. Sobiecki <sobkas@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
I had the driver all tested for the last series, and in my last build I
noticed that get_swizzled_channel was unused now, and removed
it... apparently without testing to find that I removed the wrong channel
swizzle function.
We routinely have code like:
vec1 ssa_220 = fge ssa_104, ssa_61
vec1 ssa_199 = bcsel ssa_220, ssa_106, ssa_105
and we would compare fge's args and choose between ~0 and 0 to generate
ssa_220, then compare ssa_220 to 0 and choose between bcsel's args.
Instead, try to notice the pattern and compare between fge's args to
select between bcsel's args.
total instructions in shared programs: 88019 -> 87574 (-0.51%)
instructions in affected programs: 9985 -> 9540 (-4.46%)
total estimated cycles in shared programs: 245752 -> 245237 (-0.21%)
estimated cycles in affected programs: 17232 -> 16717 (-2.99%)
We can't use its other features currently (mostly because we don't want
Newton-Raphson on rcps for texture coordinates), but it gets us started.
This eliminates some comparisons with constants in GLB2.7 and ETQW traces
at the QIR level by moving the comparisons into NIR, where they get
constant-folded out.
instructions in affected programs: 165 -> 156 (-5.45%)
total uniforms in shared programs: 32087 -> 32085 (-0.01%)
total estimated cycles in shared programs: 245762 -> 245752 (-0.00%)
estimated cycles in affected programs: 461 -> 451 (-2.17%)
I'm moving away from QIR being SSA (since NIR is doing lots of SSA
optimization for us now) and instead having QIR just be QPU operations
with virtual registers. By making our SELs be composed of two MOVs, we
could potentially coalesce the registers for the MOV's src and dst and
eliminate the MOV.
total instructions in shared programs: 88448 -> 88028 (-0.47%)
instructions in affected programs: 39845 -> 39425 (-1.05%)
total estimated cycles in shared programs: 246306 -> 245762 (-0.22%)
estimated cycles in affected programs: 162887 -> 162343 (-0.33%)
If you want the SF of the value of a register produced from a series of
packing MOVs or conditional MOVs, we can't just SF on the last MOV into
the register.
QPitch is usually expressed as rows of surface elements (where a surface
element is an compression block or a single surface sample. Skylake 1D
is an outlier; there QPitch is expressed as individual surface
elements.
For 1D surfaces and for surfaces with Yf or Ys tiling, the hardware
ignores SurfaceVerticalAlignment and SurfaceHorizontalAlignment.
Moreover, the anv_halign[] and anv_valign[] lookup tables may not even
contain the surface's actual alignment values. So don't do the lookup
for those surfaces.
Fix a 's/unsigned int/unsigned/' consistency case while here.
Found-by: Coccinelle
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Fix silly issue with MSVC case fall-though support to need
a extra 'break;'
Found-by: Coccinelle
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
This patch converts the SSE-optimized lp_rast_triangle_32_3_16()
to VMX/VSX.
I measured the results on POWER8 machine with 32 cores at 3.4GHz and
16GB of RAM.
FPS/Score
Name Before After Delta
------------------------------------------------
openarena 16.35 16.7 2.14%
xonotic 4.707 4.97 5.57%
glmark2 didn't show a significant (more than 1%) difference.
v2: Make sure code is build only on POWER8 LE machine
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
