radeonsi: add VS blit shader creation

no users yet

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: split declare_default_desc_pointers

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium/u_blitter: let drivers decide which VS to use for draw_rectangle

This approach allows drivers to set their own vertex shader and skip
compilation of u_blitter vertex shaders.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium/u_blitter: let drivers set the vertex elements state

radeonsi won't set it.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium/u_blitter: remove blitter_context_priv::viewport

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: don't use util_draw_arrays_instanced in si_draw_rectangle

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: move si_draw_rectangle into si_state_draw.c

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: remove wrappers si_decompress_xx_textures

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium/radeon: remove r600_atom::num_dw

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium/radeon: remove old r600g code checking chip_class and family

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

st/va: Implement vaExportSurfaceHandle()

This is a new interface in libva2 to support wider use-cases of passing
surfaces to external APIs.  In particular, this allows export of NV12 and
P010 surfaces.

v2: Convert surfaces to progressive before exporting them (Christian).

v3: Set destination rectangle to match source when converting (Leo).
    Add guards to allow building with libva1.

Signed-off-by: Mark Thompson <sw@jkqxz.net>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-and-Tested-by: Leo Liu <leo.liu@amd.com>

gallivm: don't use pabs intrinsic with llvm version >= 6

The intrinsic is gone, causing shader compilation to crash.
While here, also change the fallback code to match what llvm's auto-updater
of these intrinsics would do (except that there will still be zext/trunc
instructions in there), which should ensure that the sequence gets recognized
and fused back into a pabs in the end (I didn't test this, and it's possible
even the old sequence would get recognized, but I don't see a reason why we
shouldn't use the same sequence in any case).

Tested-by: Vinson Lee <vlee@freedesktop.org>

swr/rast: use proper alignment for debug transposedPrims

Causing a crash in ParaView waveletcontour.py test when
_DEBUG defined due to vector aligned copy with unaligned
address.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

anv/cmd_buffer: Reset state in cmd_buffer_destroy

This ensures that everything gets cleaned up properly. In particular,
it fixes a memory leak where we were leaking the push constants
structs.

Valgrind stats on
dEQP-VK.pipeline.push_constant.graphics_pipeline.range_size_128 :

Before:
HEAP SUMMARY:
    in use at exit: 2,467,513 bytes in 1,305 blocks
  total heap usage: 697,853 allocs, 696,530 frees, 138,466,600 bytes allocated

LEAK SUMMARY:
   definitely lost: 1,068 bytes in 11 blocks
   indirectly lost: 24,669 bytes in 412 blocks
     possibly lost: 0 bytes in 0 blocks
   still reachable: 2,441,776 bytes in 882 blocks
        suppressed: 0 bytes in 0 blocks

After:
HEAP SUMMARY:
    in use at exit: 2,467,381 bytes in 1,304 blocks
  total heap usage: 697,853 allocs, 696,531 frees, 138,466,600 bytes allocated

LEAK SUMMARY:
   definitely lost: 936 bytes in 10 blocks
   indirectly lost: 24,669 bytes in 412 blocks
     possibly lost: 0 bytes in 0 blocks
   still reachable: 2,441,776 bytes in 882 blocks
        suppressed: 0 bytes in 0 blocks

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: "17.2 17.1" <mesa-stable@lists.freedesktop.org>

anv/cmd_buffer: fix push descriptors with set > 0

When writing to set > 0, we were just wrongly writing to set 0. This
commit fixes this by lazily allocating each set as we write to them.

We didn't go for having them directly into the command buffer as this
would require an additional ~45Kb per command buffer.

v2: Allocate push descriptors from system memory rather than in BO
    streams. (Lionel)

Cc: "17.2 17.1" <mesa-stable@lists.freedesktop.org>
Fixes: 9f60ed98e501 ("anv: add VK_KHR_push_descriptor support")
Reported-by: Daniel Ribeiro Maciel <daniel.maciel@gmail.com>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: enable VK_KHR_sampler_ycbcr_conversion

v2: Make GetImageMemoryRequirements2KHR() iterate over all pInfo
    structs (Lionel)
    Handle VkSamplerYcbcrConversionImageFormatPropertiesKHR (Andrew/Jason)
    Iterator over BindImageMemory2KHR's pNext structs correctly (Jason)

v3: Revert GetImageMemoryRequirements2KHR() change from v2 (Jason)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: enable multiple planes per image/imageView

This change introduce the concept of planes for image & views. It
matches the planes available in new formats.

We also refactor depth & stencil support through the usage of planes
for the sake of uniformity. In the backend (genX_cmd_buffer.c) we have
to take some care though with regard to auxilliary surfaces.
Multiplanar color buffers can have multiple auxilliary surfaces but
depth & stencil share the same HiZ one (only store in the depth
plane).

v2: by Jason
    Remove unused aspect parameters from anv_blorp.c
    Assert when attempting to resolve YUV images
    Drop redundant logic for plane offset in make_surface()
    Rework anv_foreach_plane_aspect_bit()

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: Take an image in can_sample_with_hiz

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

anv: Take a single aspect in anv_layout_to_aux_usage

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

anv/cmd_buffer: Make get_fast_clear_state return an address

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

anv/blorp: Add a concept of default aux usage

A good chunk of anv_blorp just wants the aux usage from the image.  This
magic aux_usage value means just that.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

anv: add nir lowering pass for ycbcr textures

This pass implements all the implicit conversions required by the
VK_KHR_sampler_ycbcr_conversion specification.

It also inserts plane sources onto sampling instructions that we then
let the pipeline layout pass deal with, when mapping things correctly
to descriptors.

v2: Add new file to meson build (Lionel)
    Use nir_frcp() rather than (1.0f / x) (Jason)
    Reuse nir_tex_instr_dest_size() rather than handwritten one (Jason)
    Return progress (Jason)
    Account for array of samplers (Jason)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: prepare sampler emission code for multiplanar images

New settings from the KHR_sampler_ycbcr_conversion specifications
might require different sampler settings for luma and chroma planes.
This change makes the sampler table emission ready to handle multiple
planes.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/apply_pipeline_layout: Prepare for multi-planar images

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: add new formats KHR_sampler_ycbcr_conversion

Adding new downsampling factors for each planes.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: modify the internal concept of format to express multiple planes

A given Vulkan format can now be decomposed into a set of planes. We
now use 'struct anv_format_plane' to represent the format of those
planes.

v2: by Jason
    Rename anv_get_plane_format() to anv_get_format_plane()
    Don't rename anv_get_isl_format()
    Replace ds_fmt() by fmt2()
    Introduce fmt_unsupported()

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: prepare formats to handle disjoints sets

Newer format enums start at offset 1000000000, making it impossible to
have them all in one table. This change splits the formats into sets
that we then access through indirection.

v2: rename format_extract to vk_to_anv_format (Chad/Jason)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

isl: fill out layout descriptions for yuv formats

Some description was missing.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

isl: check whether a format is rgb if colorspace is yuv

Suggested by Chad.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

isl: make format layout channels accessible by index

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Chad Versace <chadversary@chromium.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

vulkan: util: add macros to extract extension/offset number from enums

v2: Simplify offset enum computation (Jason)

v3: capitalize macros (Chad)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

radv: convert all COMPUTE operations to the RADV_META_SAVE_XXX flags

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add RADV_META_SAVE_COMPUTE_PIPELINE flag

This will allow use to merge the compute save/restore helpers.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add radv_meta_save() helper

And merge radv_meta_save_novertex() with
radv_meta_save_graphics_reset_vport_scissor_novertex().

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: merge radv_meta_{save,restore}_pass() with RADV_META_SAVE_PASS

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: convert all GFX operations to the RADV_META_SAVE_XXX flags

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: introduce the concept of meta save flags

This will allow us to save/restore the different states on-demand
based on the meta operation. For now, this saves/restores all
states. Compute will follow once the graphics part is done.

The main idea is to merge all save/restore helpers.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: remove unused RADV_META_VERTEX_BINDING_COUNT

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: select the pipeline outside of the loop when decompressing htile

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add radv_htile_enabled() helper

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

i965: pass wanted format to intel_miptree_create_for_dri_image

Change b3a44ae7a4 caused regressions on Android where DRI and renderbuffer
can disagree on the format being used. This patch removes the colorspace
parameter and instead we pass renderbuffer format. For non-winsys images we
still do srgb/linear modification in same manner as change b3a44ae7a4 wanted
but take format from renderbuffer instead of DRI image.

This patch fixes regressions seen with following test sets:

   dEQP-EGL.functional.color_clears*
   dEQP-EGL.functional.render*

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102999
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

radeonsi: add a drirc workaround for HTILE corruption in ARK: Survival Evolved

v2: use DB_META | PS_PARTIAL_FLUSH

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102955
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> (v1)
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com> (v1)

radeonsi: inline struct si_sampler_views

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: rename si_textures_info -> si_samplers, si_images_info -> si_images

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: fold needs_*_decompress_mask update into si_set_sampler_view

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: simplify a loop in si_update_fb_dirtiness_after_rendering

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

ac: properly document a buffer.store LLVM workaround

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use f32_0 and f32_1

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: fold *gallivm

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: lp_type::length is always 1

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: don't use bld.elem_type

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: don't use lp_build_const_*

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use ctx->ac.context and ctx->types

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use ctx->ac.builder

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use ctx->i/f32 types more

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use i32_0 and i32_1 more

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use bitcast in a few places

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use ac helpers for bitcasts

Reviewed-by: Nicolai Hähnle <nicolai.haehnle at amd.com>

glsl_to_tgsi: skip UARL for 1D registers if the driver doesn't need it

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

glsl_to_tgsi: handle reladdr as TEMP in rename_temp_registers and dead_code

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

glsl_to_tgsi: each reladdr object should have only one parent

required by rename_temp_registers.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

glsl_to_tgsi: fix instruction order for bindless textures

We emitted instructions loading the bindless handle after the memory
instruction.

Cc: 17.2 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

glsl_to_tgsi: enable copy propagation for tessellation shaders

just don't propagate output reads

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: implement PIPE_CAP_TGSI_ANY_REG_AS_ADDRESS

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use si_get_indirect_index for TEMP indexing

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use si_get_indirect_index for CONST indexing

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

tgsi/ureg: allow any register file in address operands

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium: add PIPE_CAP_TGSI_ANY_REG_AS_ADDRESS

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

tgsi/scan: scan address operands (v2)

v2: set swizzled usage mask

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

tgsi/scan: set correct usage mask for tex offsets in scan_src_operand

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

tgsi/scan: take advantage of already swizzled usage mask in scan_src_operand

It has always been a usage mask *after* swizzling.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

tgsi/scan: set non-valid src_index for tex offsets in scan_src_operand

tex offsets are not "Src" operands.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

tgsi: implement tgsi_util_get_inst_usage_mask properly

All opcodes are handled.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

tgsi: add docs for some existing pack opcodes

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radv: Enable VK_KHR_maintenance2 extension.

Reviewed-by: Dave Airlie <airlied@redhat.com>

radv: Make tess winding order a bit more intuitive.

Reviewed-by: Dave Airlie <airlied@redhat.com>

radv: Allow setting the domain origin in tess.

Reviewed-by: Dave Airlie <airlied@redhat.com>

radv: Disable usage checks in metadata for images with extended usage data.

The app can extend the usage, so knowing that the usage is limitied
does not help us here.

Reviewed-by: Dave Airlie <airlied@redhat.com>

radv: Implement querying the point clipping behavior.

Reviewed-by: Dave Airlie <airlied@redhat.com>

broadcom: Fix out-of-tree build include path

Reviewed-by: Eric Anholt <eric@anholt.net>
Fixes: 5b102160ae ("broadcom/genxml: Introduce a V3D packet/struct decoder.")

meson: generate builddir/src/amd/vulkan/dev_icd.json

Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

mesa: Use a 565 format for GL_RGB and GL_UNSIGNED_SHORT_5_6_5 textures.

Found while trying to optimize an application.

Not observed to help performance on i965, but should at least reduce
the memory usage of such textures a bit.

Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Eero Tamminen <eero.t.tamminen@intel.com>

intel/compiler: Don't propagate cmod into integer multiplies

No shader-db change on Sky Lake.

Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org

intel/compiler: Don't cmod propagate into a saturated operation

Shader-db results on Sky Lake:

    total instructions in shared programs: 12954445 -> 12955125 (0.01%)
    instructions in affected programs: 141862 -> 142542 (0.48%)
    helped: 0
    HURT: 626

Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org

broadcom/vc4: Don't advertise tiled dmabuf modifiers if we can't use them

If the DRM_VC4_GET_TILING ioctl isn't present then we can't tell
if a dmabuf bo is tiled or linear, so will always assume it's
linear.

By not advertising tiled formats in this situation we ensure the
assumption is correct.

This fixes a bug where most attempts to render a gl wayland client
under weston will result in a client side abort.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Acked-by: Daniel Stone <daniels@collabora.com> (on irc)

egl: Simplify the "driver" interface

"Driver" isn't a great word for what this layer is, it's effectively a
build-time choice about what OS you're targeting. Despite that both of
the extant backends totally ignore the display argument, the old code
would only set up the backend relative to a display.

That causes problems! One problem is it means eglGetProcAddress can
generate X or Wayland protocol when it tries to connect to a default
display so it can call into the backend, which is, you know, completely
bonkers. Any other EGL API that doesn't reference a display, like
EGL_EXT_device_query, would have the same issue.

Fortunately this is a problem that can be solved with the delete key.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

loader/dri3: Don't accidently free buffer holding new back content

Avoid freeing buffers holding new back content
(with GLX_SWAP_COPY_OML and GLX_SWAP_EXCHANGE_OML)
Prevously that would have resulted in back buffer content becoming
incorrect after a swap, although I haven't managed to trigger such a
situation yet.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Sinclair Yeh <syeh@vmware.com>

loader/dri3: Avoid resizing existing buffers in dri3_find_back_alloc

Resize only in loader_dri3_get_buffers(),
where the dri driver has a chance to immediately update the viewport.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Sinclair Yeh <syeh@vmware.com>

loader/dri3: Use local blits and local buffers when resizing

When a drawable is resized, and we fill the resized buffers, with data
from the old buffers, use a local blit if there is a local buffer (back or
fake front), and we have local blitting capability.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Sinclair Yeh <syeh@vmware.com>

gallivm/ppc64le: allow environmental control of Altivec code generation

In check_os_altivec_support(), allow control of Altivec (first PPC vector
instruction set) code generation via a new environmental control,
GALLIVM_ALTIVEC, which is expected to take on a value of 1 or 0.
The default is to enable Altivec code generation.

This environmental control of Altivec code generation is initially
available only #ifdef DEBUG.

Cc: "17.2" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Ben Crocker <bcrocker@redhat.com>
Acked-by: Roland Scheidegger <sroland@vmware.com>

gallivm/ppc64le: adjust VSX code generation control.

In lp_build_create_jit_compiler_for_module(), advance the minimum
version of LLVM for VSX code generation to 4.0; this is the minimum
revision at which several known VSX code generation bugs are fixed:

  https://llvm.org/bugs/show_bug.cgi?id=25503 (fixed in 3.8.1)
  https://llvm.org/bugs/show_bug.cgi?id=26775 (fixed in 3.8.1)
  https://llvm.org/bugs/show_bug.cgi?id=33531 (fixed in 4.0)

An llc performance bug introduced in LLVM 4.0,

  https://llvm.org/bugs/show_bug.cgi?id=34647

is still pending as of LLVM 5.0, but only has a pronounced effect on
one of the Piglit tests: ext_transform_feedback-max-varyings.

All changes tested via Piglit.

Cc: "17.2" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Ben Crocker <bcrocker@redhat.com>
Acked-by: Roland Scheidegger <sroland@vmware.com>

gallivm: allow additional llc options

In init_native_targets, allow the passing of additional options to
the LLC compiler via new GALLIVM_LLC_OPTIONS environmental control.
This option is available only #ifdef DEBUG, initially.
At top, add #include <llvm-c/Support.h> for LLVMParseCommandLineOptions()
declaration.

v2: Fix compile error with old llvm versions (sroland)

Cc: "17.2" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Ben Crocker <bcrocker@redhat.com>
Acked-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

gallivm: fix typo in debug_printf message

In gallivm_compile_module, fix a typo in the
debug_printf("Invoke as \"llc ..." message.

Cc: "17.2" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Ben Crocker <bcrocker@redhat.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>

radv: remove useless checks around radv_CmdBindPipeline()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: check that pipeline is different before binding it

We only need to dirty the descriptors when the pipeline is
a new one, because user SGPRs can be potentially different.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

i965: Validate "Special Requirements for Handling Double Precision Data Types"

I did not implement:

   CNL's restriction on 64-bit int + align16, because I don't think
   we'll ever use this combination regardless of hardware generation.

   The restriction on immediate DF -> F conversions, because there's no
   reason to ever generate that, and I don't even know how DF -> F
   conversions are supposed to work in Align16 since (1) the dst stride
   must be 1, but (2) the dst stride would have to be 2 for src and dst
   strides to be aligned.

i965: Fix and enable forgotten validation test

I seem to have forgotten I still had work to do.

i965: Only insert error message if not already present

Some restrictions require something like strides to match between src
and dest. For multi-source instructions, I'd rather encapsulate the
logic for not inserting already present errors in ERROR_IF than
open-coding it multiple places.

i965: Avoid validation error when src1 is not present

There can be no violation of the restriction that source offsets are
aligned if there is only one source offset.

i965: Remove validate_reg()

Replaced by the assembly validator, and in fact gets in the way of
writing tests for the assembly validator.