Many places need to know the maximum or minimum possible value for a
given size integer... so everyone just open-codes their favorite
version. There is some potential to hit either undefined or
implementation-defined behavior, so having one version that Just Works
seems beneficial.
v2: Fix copy-and-pasted bug (INT64_MAX instead of INT64_MIN) in
u_intmin. Noticed by CI. Lol. Rename functions
`s/u_(uint|int)(min|max)/u_\1N_\2/g`. Suggested by Jason. Add some
unit tests that would have caught the copy-and-paste bug before wasting
CI time. Change the implementation of u_intN_min to use the same
pattern as stdint.h. This avoids the integer division. Noticed by
Jason.
v3: Add changes to convert_clear_color
(src/gallium/drivers/iris/iris_clear.c). Suggested by Nanley.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12177>
It notably didn't fit the pattern of RGB5_A1_UNORM, and violated the
general pattern for bitmask format BE channels (channels are ordered
right-to-left in the BE columns in the CSV due to the parser walking them
in that order for historical reasons).
Acked-by: Adam Jackson <ajax@redhat.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10505>
These tests passed for LE, and the BE channel ordering specified obviously
didn't fit the pattern of the other BE formats (channels are listed
right-to-left in the BE columns for historical reasons).
Note that we can't write pure-integer format tests in u_format_tests.c
currently.
Acked-by: Adam Jackson <ajax@redhat.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10505>
gcc 11 warns:
../src/util/format/u_format_fxt1.c:940:22: warning: ‘fxt1_variance.constprop’ accessing 128 bytes in a region of size 64 [-Wstringop-overflow=]
940 | int32_t maxVarR = fxt1_variance(NULL, &input[N_TEXELS / 2], n_comp);
But, suspiciously, if you inline fxt1_variance the warning goes away.
What's happening is that the 2nd arg is uint8_t[N_TEXELS][MAX_COMP], so
it looks like we're passing too small of an array in since gcc knows
that `input` is also [N_TEXELS][MAX_COMP]. Fair enough. Fix the
signature to reflect what's actually going on, and remove some unused
arguments while we're at it.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10671>
Before we can fix big endian, we kind of have to agree on how pixel data
should be laid out in memory.
For this we have to use packed data for !is_array declared formats and per
element values for is_array ones.
Also we need proper PACKED_1xn definitions for BE.
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7776>
channel datatypes in Mesa are the host's endianness. For example,
PIPE_FORMAT_R32_UINT doesn't do a bswap in and out in u_format_table.c's
pack/unpack functions. So, z32_unorm shouldn't be byte swapping either,
and neither should z24s8 which is also a packed format, and once you've
got those it becomes clear that all of the swaps in this file were
mistaken.
Things would mostly work out because it's unusual to read/write Z/S data
through the GL API, and even for drivers like softpipe as long as the pack
and unpack both swap it could work anyway. However, the bug would be
visible in glReadPixels() with the matching datatype which would hit the
memcpy fastpath without doing another swap.
Caught by a mesa/main unit test on transitioning to using these
pack/unpack functions.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10336>
In looking at the profile of dEQP, GLES3 was spending 5-10% of its time in
ReadPixels, and almost all of that is b8g8r8a8_unorm8. It's really slow
because we're getting about 47MB/s by doing uncached reads 32 bits at a
time in the code-generated unpack. If we use NEON to generate larger bus
transactions, we can speed things up to 136MB/s. In comparison, raw
ldr/str read/writes with no byte swapping can hit a max of 216MB/sec.
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10014>
We have only a few callers of unpack that do rects, so add a helper that
iterates over y adding the strides. This saves us 36kb of generated code
and means that adding cpu-specific variants for RGBA format unpack will be
much simpler.
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10014>
MSVC warns (loudly) about a mismatch between the generated functions in
u_format_table.c and the definition of util_format_[un]pack_description,
specifically having 'restrict' in the function but not in the pointer
type in the struct.
So, add 'restrict' everywhere - to the function types, and to the rest
of the implementations that are assigned to those structs.
v2: util_format_unpack_description is used in gallium/auxiliary/translate
Fixes: 5785fdac ("u_format: Mark the generated pack/unpack src/dst args as restrict.")
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9670>
Calling code to pack/unpack with overlap would be already be undefined.
Cuts 50k of text on x86_64 release builds from the compiler having more
freedom in the src/dst loads knowing that they don't interfere with each
other.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9500>
Adrian Bunk