`swizzle` had the wrong operands, and the vector masking boolean logic
was incorrect in the internal `shuffle_or_0` implementation. `shuffle`
was previously implemented as a dynamic swizzle, when it uses an
immediate operand for lane indices in the spec.
(cherry picked from commit 9cc3e7d32d150dd30d683c1a8cf0bd59676f14ab)
Previously we relied on signed overflow, this commit makes the same
behaviour explicit (avoiding UB in the process).
(cherry picked from commit 8c8310f0bddc874a9f7f07c4158f0abc799357d4)
With this we pass an additional ~2100 tests.
We are left with 7106 WASM fails :).
There's still some test cases in the iNxM tests that fail with
this PR, but they are somewhat weird.
(cherry picked from commit b4acd4fb0b7f4105c7ef673ccc00904114c3c468)
Co-authored-by: Diego Frias <styx5242@gmail.com>
NaN bit patterns are now (hopefully) preserved. `static_cast` does not
preserve the bit pattern of a given NaN, so ideally we'd use some other
sort of cast and avoid `static_cast` altogether, but that's a large
change for this commit. For now, this fixes the issues found in spec
tests.
(cherry picked from commit c882498d4450c4c2e46d77a8ab36afc4eb412c00)
Otherwise we'd hit a VERIFY in AK::SIMD::shuffle() when that operand
contains an out-of-range value, the spec tests indicate that a swizzle
with an out-of-range index should return 0.
Adds all the arithmetic ops for f32x4 and f64x2 SIMD instructions.
With this, we pass 8375 additional tests :)
Quite a few of the spec tests for this are still failing.
I confirmed with the wasmer runtime manually for a number of them,
and we seem to match their and results. I'm not really sure
what's happening here, a spec bug or wasmer is broken in
the same way.
18476 failed before.
10101 failed after.
It's possible to construct a floating point value that when converted to
double is not larger than i64::max(), but when remaining a float is
larger. This patch avoids that edge case with some even less exciting if
constexpr branches to fix a float-cast-overflow UBSAN error on macOS
with llvm 15.0.6.
Each of these strings would previously rely on StringView's char const*
constructor overload, which would call __builtin_strlen on the string.
Since we now have operator ""sv, we can replace these with much simpler
versions. This opens the door to being able to remove
StringView(char const*).
No functional changes.
In order to reduce our reliance on __builtin_{ffs, clz, ctz, popcount},
this commit removes all calls to these functions and replaces them with
the equivalent functions in AK/BuiltinWrappers.h.
This should make debugging and profiling much better, at little to no
runtime cost.
Also moves off the operator definitions to a separate header, so it
should also improve the editing experience quite a bit.
