We can not use the VUE Dereference flags combination for EOT
message under ILK and SNB because the threads are not initialized
there with initial VUE handle unlike Pre-IL.
So to avoid GPU hangs on SNB and ILK we need
to avoid usage of the VUE Dereference flags combination.
(Was tested only on SNB but according to the specification
SNB Volume 2 Part 1: 1.6.5.3, 1.6.5.6
the ILK must behave itself in the similar way)
v2: Approach to fix this issue was changed.
Instead of different EOT flags in the program end
we will create VUE every time even if GS produces no output.
v3: Clean up the patch.
Signed-off-by: Andrii Simiklit <andrii.simiklit@globallogic.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105399
CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Tested-by: Mark Janes <mark.a.janes@intel.com>
This extension provides fences and frame count information to direct
display contexts. It uses new kernel ioctls to provide 64-bits of
vblank sequence and nanosecond resolution.
v2: Adopt Jason Ekstrand's coding conventions
Declare variables at first use, eliminate extra whitespace between
types and names. Wrap lines to 80 columns.
Add extension to list in alphabetical order
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
v3: Adapt to WSI fence API change. It now returns VkResult and
no longer has an option for relative timeouts.
v4: wsi_register_display_event and wsi_register_device_event now
use the default allocator when NULL is provided, so remove the
computation of 'alloc' here.
v5: use zalloc2 instead of alloc2 for the WSI fence.
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Handle the case where the set of fences to wait for is not all of the
same type by either waiting for them sequentially (waitAll), or
polling them until the timer has expired (!waitAll). We hope the
latter case is not common.
While the current code makes sure that it always has fences of only
one type, that will not be true when we add WSI fences. Split out this
refactoring to make merging that clearer.
v2: Adopt Jason Ekstrand's coding conventions
Declare variables at first use, eliminate extra whitespace between
types and names. Wrap lines to 80 columns.
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
v2:
Cast INT64_MAX to uint64_t to make of its use as the maximum
possible timeout clearly unsigned to the reader.
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
Make anv_wait_for_fences with !waitAll check all fences at least
once, even if the requested timeout has already passed.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
This completely reworks the pass to support deref instructions and
delete support for old deref chains
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Now that it's rewritten for deref instructions, we can turn it back on.
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
nir_tex_instr_create uses rzalloc so it's already NULL
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Everything else should already be handled.
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Since we had to rewrite the deref walking loop anyway, I took the
opportunity to make it a bit clearer and more efficient. In particular,
in the AoA case, we will now emit one minmax instead of one per array
level.
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This commit removes most of the deref instruction lowering. Instead of
lowering early, we only lower textures and images and we only do so
right before any of the anv image lowering passes.
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This commit completely reworks function calls in NIR. Instead of having
a set of variables for the parameters and return value, nir_call_instr
now has simply has a number of sources which get mapped to load_param
intrinsics inside the functions. It's up to the client API to build an
ABI on top of that. In SPIR-V, out parameters are handled by passing
the result of a deref through as an SSA value and storing to it.
This virtue of this approach can be seen by how much it allows us to
delete from core NIR. In particular, nir_inline_functions gets halved
and goes from a fairly difficult pass to understand in detail to almost
trivial. It also simplifies spirv_to_nir somewhat because NIR functions
never were a good fit for SPIR-V.
Unfortunately, there is no good way to do this without a mega-commit.
Core NIR and SPIR-V have to be changed at the same time. This also
requires changes to anv and radv because nir_inline_functions couldn't
handle deref instructions before this change and can't work without them
after this change.
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Before, we were doing structure splitting in spirv_to_nir.
Unfortunately, this doesn't really work when you think about passing
struct pointers into functions. Doing it later in NIR is a much better
plan.
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This pass doesn't handle deref instructions yet. Making it handle both
legacy derefs and deref instructions would be painful. Since it's not
important for correctness, just disable it for now.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This inserts a call to nir_lower_deref_instrs at every call site of
glsl_to_nir, spirv_to_nir, and prog_to_nir.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
vtn_variable_mode_image and _sampler are instead replaced with
vtn_variable_mode_uniform which encompasses both of them. In the few
places where it was neccessary to distinguish between the two, the
GLSL type of the pointer is used instead.
The main reason to do this is that on OpenGL it is permitted to put
images and samplers into structs and declare a uniform with them. That
means that variables can now have a mix of uniform, sampler and image
modes so picking a single one of those modes for a variable no longer
makes sense.
This fixes OpLoad on a sampler within a struct which was previously
using the variable mode to determine whether it was a sampler or not.
The type of the variable is a struct so it was not being considered to
be uniform mode even though the member being loaded should be sampler
mode.
The previous code appeared to be using var->interface_type as a place
to store the type of the variable without the enclosing array for
images and samplers. I guess this worked because opaque types can not
appear in interfaces so the interface_type is sort of unused. This
patch removes the overloading of var->interface_type and any places
that needed the type without the array can now just deduce it from
var->type.
v2: squash in this patch the changes to anv/nir (Timothy)
Signed-off-by: Eduardo Lima <elima@igalia.com>
Signed-off-by: Neil Roberts <nroberts@igalia.com
Signed-off-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
When VK_USE_PLATFORM_XLIB_XRANDR_EXT is defined, vulkan.h includes
X11/extensions/Xrandr.h for the RROutput typedef which is used in
the vkGetRandROutputDisplayEXT interface.
Make sure we have the required header by checking during the build,
and also set CFLAGS to point at the right directory.
We don't need to link against the library as we don't use any
functions from there, so don't add the _LIBS value in the autotools
build.
Signed-off-by: Keith Packard <keithp@keithp.com>
Fixes: dbac8e25f8 "radv: Add EXT_acquire_xlib_display to radv driver [v2]"
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
This extension is required to support EXT_display_control as it offers
a way to query whether the vblank counter is supported.
v2:
Add extension to list in alphabetical order
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This extension adds the ability to borrow an X RandR output for
temporary use directly by a Vulkan application to the anv driver.
v2:
Simplify addition of VK_USE_PLATFORM_XLIB_XRANDR_KHR to
vulkan_wsi_args
Suggested-by: Eric Engestrom <eric.engestrom@imgtec.com>
v3:
Add extension to list in alphabetical order
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Add support for the EXT_direct_mode_display extension. This just
provides the vkReleaseDisplayEXT function.
v2: Add extension to list in alphabetical order
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This adds support for the KHR_display extension to the anv Vulkan
driver. The driver now attempts to open the master DRM node when the
KHR_display extension is requested so that the common winsys code can
perform the necessary operations.
v2: Make sure primary fd is usable
When KHR_display is selected, we try to open the primary node
instead of the render node in case the user wants to use
KHR_display for presentation. However, if we're actually going
to end up using RandR leases, then we don't care if the
resulting fd can't be used for display, but the kernel also
prevents us from using it for drawing when someone else has
master.
v3:
Simplify addition of VK_USE_PLATFORM_DISPLAY_KHR to vulkan_wsi_args
Suggested-by: Eric Engestrom <eric.engestrom@imgtec.com>
v4:
Adapt primary node usage to new wsi_device_init API
v5:
Adopt Jason Ekstrand's coding conventions
Declare variables at first use, eliminate extra whitespace between
types and names. Wrap lines to 80 columns.
Remove spurious MM_PER_PIXEL define
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
v6:
Open DRM master before initializing WSI layer.
The DRM master FD is passed to the WSI layer during
initialization, so we need to open the device slightly earlier
in the function.
Close DRM master in device_finish.
Use anv_gem_get_param to detect working master_fd instead of
directly using the ioctl.
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
v7:
Add vkCreateDisplayModeKHR. This doesn't actually create
new modes, it only looks to see if the requested parameters
matches an existing mode and returns that.
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This adds support for the KHR_display extension support to the vulkan
WSI layer. Driver support will be added separately.
v2:
* fix double ;; in wsi_common_display.c
* Move mode list from wsi_display to wsi_display_connector
* Fix scope for wsi_display_mode andwsi_display_connector
allocs
* Switch all allocations to vk_zalloc instead of vk_alloc.
* Fix DRM failure in
wsi_display_get_physical_device_display_properties
When DRM fails, or when we don't have a master fd
(presumably due to application errors), just return 0
properties from this function, which is at least a valid
response.
* Use vk_outarray for all property queries
This is a bit less error-prone than open-coding the same
stuff.
* Remove VK_COMPOSITE_ALPHA_INHERIT_BIT_KHR from surface caps
Until we have multi-plane support, we shouldn't pretend to
have any multi-plane semantics, even if undefined.
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
* Simplify addition of VK_USE_PLATFORM_DISPLAY_KHR to
vulkan_wsi_args
Suggested-by: Eric Engestrom <eric.engestrom@imgtec.com>
v3:
Add separate 'display_fd' and 'render_fd' arguments to
wsi_device_init API. This allows drivers to use different FDs
for the different aspects of the device.
Use largest mode as display size when no preferred mode.
If the display doesn't provide a preferred mode, we'll assume
that the largest supported mode is the "physical size" of the
device and report that.
v4:
Make wsi_image_state enumeration values uppercase.
Follow more common mesa conventions.
Remove 'render_fd' from wsi_device_init API. The
wsi_common_display code doesn't use this fd at all, so stop
passing it in. This avoids any potential confusion over which
fd to use when creating display-relative object handles.
Remove call to wsi_create_prime_image which would never have
been reached as the necessary condition (use_prime_blit) is
never set.
whitespace cleanups in wsi_common_display.c
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Add depth/bpp info to available surface formats. Instead of
hard-coding depth 24 bpp 32 in the drmModeAddFB call, use the
requested format to find suitable values.
Destroy kernel buffers and FBs when swapchain is destroyed. We
were leaking both of these kernel objects across swapchain
destruction.
Note that wsi_display_wait_for_event waits for anything to
happen. wsi_display_wait_for_event is simply a yield so that
the caller can then check to see if the desired state change
has occurred.
Record swapchain failures in chain for later return. If some
asynchronous swapchain activity fails, we need to tell the
application eventually. Record the failure in the swapchain
and report it at the next acquire_next_image or queue_present
call.
Fix error returns from wsi_display_setup_connector. If a
malloc failed, then the result should be
VK_ERROR_OUT_OF_HOST_MEMORY. Otherwise, the associated ioctl
failed and we're either VT switched away, or our lease has
been revoked, in which case we should return
VK_ERROR_OUT_OF_DATE_KHR.
Make sure both sides of if/else brace use matches
Note that we assume drmModeSetCrtc is synchronous. Add a
comment explaining why we can idle any previous displayed
image as soon as the mode set returns.
Note that EACCES from drmModePageFlip means VT inactive. When
vt switched away drmModePageFlip returns EACCES. Poll once a
second waiting until we get some other return value back.
Clean up after alloc failure in
wsi_display_surface_create_swapchain. Destroy any created
images, free the swapchain.
Remove physical_device from wsi_display_init_wsi. We never
need this value, so remove it from the API and from the
internal wsi_display structure.
Use drmModeAddFB2 in wsi_display_image_init. This takes a drm
format instead of depth/bpp, which provides more control over
the format of the data.
v5:
Set the 'currentStackIndex' member of the
VkDisplayPlanePropertiesKHR record to zero, instead of
indexing across all displays. This value is the stack depth of
the plane within an individual display, and as the current
code supports only a single plane per display, should be set
to zero for all elements
Discovered-by: David Mao <David.Mao@amd.com>
v6:
Remove 'platform_display' bits from the build and use the
existing 'platform_drm' instead.
v7:
Ensure VK_ICD_WSI_PLATFORM_MAX is large enough by
setting to VK_ICD_WSI_PLATFORM_DISPLAY + 1
v8:
Simplify wsi_device_init failure from wsi_display_init_wsi
by using the same pattern as the other wsi layers.
Adopt Jason Ekstrand's white space and variable declaration
suggestions. Declare variables at first use, eliminate extra
whitespace between types and names, add list iterator helpers,
switch to lower-case list_ macros.
Respond to Jason's April 8 review:
* Create a function to convert relative to absolute timeouts
to catch overflow issues in one place
* use VK_NULL_HANDLE to clear prop->currentDisplay
* Get rid of available_present_modes array.
* return OUT_OF_DATE_KHR when display_queue_next called after
display has been released.
* Make errors from mode setting fatal in display_queue_next
* Remove duplicate pthread_mutex_init call
* Add wsi_init_pthread_cond_monotonic helper function to
isolate pthread error handling from wsi_display_init_wsi
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
v9:
Fix vscan handling by using MAX2(vscan, 1) everywhere. Vscan
can be zero anywhere, which is treated the same as 1.
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
v10:
Respond to Vulkan CTS failures.
1. Initialize planeReorderPossible in display_properties code
2. Only report connected displays in
get_display_plane_supported_displays
3. Return VK_ERROR_OUT_OF_HOST_MEMORY when pthread cond
initialization fails.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
4. Add vkCreateDisplayModeKHR. This doesn't actually create
new modes, it only looks to see if the requested parameters
matches an existing mode and returns that.
Suggested-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
The Vulkan API has only one patch version shared among all of the
major.minor versions. We should also advertise the same patch version
regardless of major.minor.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=106941
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
If we are on gen8+ and have context isolation support, just make that
constant buffer address be absolute, so we can use it for push UBOs too.
v2: Do not duplicate constant_buffer_0_is_relative flag (Jason)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Shouldn't make any functional difference, just that `liblibanv_gen90.a`
will now be called `libanv_gen90.a`.
Fixes: 3218056e0e "meson: Build i965 and dri stack"
Fixes: d1992255bb "meson: Add build Intel "anv" vulkan driver"
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
As the previous use of shuffle_32bit_load_result_to_64bit_data
had a source/destination overlap for 64-bit. Now a temporary destination
is used for 64-bit cases to use shuffle_from_32bit_read that doesn't
handle src/dst overlaps.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Previously, the shuffle function had a source/destination overlap that
needs to be avoided to use shuffle_from_32bit_read. As we can use for
the shuffle destination the destination of removed MOVs.
This change also avoids the internal MOVs done by the previous shuffle
to deal with possible overlaps.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
shuffle_from_32bit_read manages 32-bit reads to 32-bit destination
in the same way that the previous loop so now we just call the new
function for all bitsizes, simplifying also the 64-bit load_input.
v2: Add comment about future 16-bit support (Jason Ekstrand)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This implementation avoids two unneeded MOVs for each 64-bit
component. One was done in the old shuffle, to avoid cases of
src/dst overlap but this is not the case. And the removed MOV
was already being being done in the shuffle.
Copy propagation wasn't able to remove them because shuffle
destination values are defined with partial writes because they
have stride == 2.
v2: Reword commit log summary (Jason Ekstrand)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
do_untyped_vector_read is used at load_ssbo and load_shared.
The previous MOVs are removed because shuffle_from_32bit_read
can handle storing the shuffle results in the expected destination
just using the proper offset.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Using shuffle_from_32bit_read instead of 16-bit shuffle functions
avoids the need of retype. At the same time new function are
ready for 8-bit type SSBO reads.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
shuffle_from_32bit_read can manage the shuffle/unshuffle needed
for different 8/16/32/64 bit-sizes at VARYING PULL CONSTANT LOAD.
To get the specific component the first_component parameter is used.
In the case of the previous 16-bit shuffle, the shuffle operation was
generating not needed MOVs where its results where never used. This
behaviour passed unnoticed on SIMD16 because dead_code_eliminate
pass removed the generated instructions but for SIMD8 they cound't be
removed because of being partial writes.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
These new shuffle functions deal with the shuffle/unshuffle operations
needed for read/write operations using 32-bit components when the
read/written components have a different bit-size (8, 16, 64-bits).
Shuffle from 32-bit to 32-bit becomes a simple MOV.
shuffle_src_to_dst takes care of doing a shuffle when source type is
smaller than destination type and an unshuffle when source type is
bigger than destination. So this new read/write functions just need
to call shuffle_src_to_dst assuming that writes use a 32-bit
destination and reads use a 32-bit source.
As shuffle_for_32bit_write/from_32bit_read components take components
in unit of source/destination types and shuffle_src_to_dst takes units
of the smallest type component, we adjust components and first_component
parameters.
To enable this new functions it is needed than there is no
source/destination overlap in the case of shuffle_from_32bit_read.
That never happens on shuffle_for_32bit_write as it allocates a new
destination register as it was at shuffle_64bit_data_for_32bit_write.
v2: Reword commit log and add comments to explain why first_component
and components parameters are adjusted. (Jason Ekstrand)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This new function takes care of shuffle/unshuffle components of a
particular bit-size in components with a different bit-size.
If source type size is smaller than destination type size the operation
needed is a component shuffle. The opposite case would be an unshuffle.
Component units are measured in terms of the smaller type between
source and destination. As we are un/shuffling the smaller components
from/into a bigger one.
The operation allows to skip first_component number of components from
the source.
Shuffle MOVs are retyped using integer types avoiding problems with
denorms and float types if source and destination bitsize is different.
This allows to simplify uses of shuffle functions that are dealing with
these retypes individually.
Now there is a new restriction so source and destination can not overlap
anymore when calling this shuffle function. Following patches that migrate
to use this new function will take care individually of avoiding source
and destination overlaps.
v2: (Jason Ekstrand)
- Rewrite overlap asserts.
- Manage type_sz(src.type) == type_sz(dst.type) case using MOVs
from source to dest. This works for 64-bit to 64-bits
operation that on Gen7 as it doesn't support Q registers.
- Explain that components units are based in the smallest type.
v3: - Fix unshuffle overlap assert (Jason Ekstrand)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
