update() calls mem.indexOfScalar() which uses `==` for comparing items,
which fails when the operator is not supported.
As PirorityQueue needs a comparing function anyway we can use `.eq` results
to identify matching objects.
closes#9918
architectures
The idea behind this is using the register capabilities in safe amounts,
there is still some consideration to be done.
+ Fixed compile error using std.Target.<arch>.featureSetHas
+ X86 MMX and "3DNOW" 64 bits register usage considered for vector size
+ Added ARM Neon recommened usage of 128 bits (The size of the register)
+ Added AARCH64 Neon and SVE for 128 bits. SVE could use in theory up to
2048 bits, but found only evidence of functional 512 bits on a super
computer, decided on using 128 bits as a safety
+ Added Altivec recommendation of using the 128 bits long register
+ Using MIPS msa 2x64bits capabilities, usage of 64 bits registers for MDMX
systems, need testing on how using bigger values affect performance
+ Using V extension on RISC-V, which is extendable via instructions, decided on 128 bits
as a value to not use all registers
+ in SPARC the 64 bits registers are used, a max of 32 registers
are to be used for configurable simd instructions, decided on using
the size of the register, need testing on performance hit on using a
bigger sized register vector size
When lowering a struct type to an LLVM struct type, keep track of
whether there are any underaligned fields. If so, then make it a packed
llvm struct. This works because we already insert manual padding bytes
regardless.
We could unconditionally use an LLVM packed struct; the reason we bother
checking for underaligned fields is that it is a conservative choice, in
case LLVM handles packed structs less optimally. A future improvement
could simplify this code by unconditionally using packed LLVM structs
and then make sure measure perf is unaffected.
closes#12190
`validateExternType` does not require the type to be resolved so we can
check it earlier. Only doing it in `resolveTypeFully` lead to worse or
missing compile errors.
* Use a debug build of stage3 instead of a debug build of stage2 for
our self-hosted compiler test coverage.
* Move coverage from stage1 to stage3 for:
- building self-hosted without LLVM
- building self-hosted for 32-bit arm
- test-compiler-rt
- test-behavior
- test-std
- test-compare-output
- test-asm-link
- test-fmt
We were accidentally returning a pointer to stack memory, because these
arguments were passed by value. It's just an accident that stage 1 was
passing these by reference, so things were alright until stage 3.
This is a possible workaround for
https://github.com/llvm/llvm-project/issues/56585
On my computer it makes stage3-release go from false positive
compilation errors on the behavior tests to "segmentation fault".
Is this forwards progress or backwards progress? I have no idea.
See #11450
Since we know the offset, we may as well read starting there. Still expects
rpath to fit in 4096 bytes; that might be worth fixing in the future.
Fixes issue #12112
This is a workaround for https://github.com/llvm/llvm-project/issues/56585
which causes writes to i1 in memory to be optimized to an incorrect value.
Unfortunately, this does not save users from running into this bug with u1
in their own types.
However, this does seem to be enough to get the behavior tests working.
This resolves#11450 on my machine.
