Eliminate the `std.Thread.Pool` used in the compiler for concurrency and
asynchrony, in favour of the new `std.Io.async` and `std.Io.concurrent`
primitives.
This removes the last usage of `std.Thread.Pool` in the Zig repository.
On the x86_64 self hosted backend, thread locals are accessed through
__tls_get_addr on PIC. Usually this goes through a fast path which does
not lose any registers, however in some cases (notably any dlopened
library on my machine) this can take a slow path which calls out to C
ABI functions
Catch this case and backup registers as necessary
Fix a few other ones while we're here. Credit to mlugg
Fixes#30183
A new `Legalize.Feature` tag is introduced for each float bit width
(16/32/64/80/128). When e.g. `soft_f16` is enabled, all arithmetic and
comparison operations on `f16` are converted to calls to the appropriate
compiler_rt function using the new AIR tag `.legalize_compiler_rt_call`.
This includes casts where the source *or* target type is `f16`, or
integer<=>float conversions to or from `f16`. Occasionally, operations
are legalized to blocks because there is extra code required; for
instance, legalizing `@floatFromInt` where the integer type is larger
than 64 bits requires calling an arbitrary-width integer conversion
function which accepts a pointer to the integer, so we need to use
`alloc` to create such a pointer, and store the integer there (after
possibly zero-extending or sign-extending it).
No backend currently uses these new legalizations (and as such, no
backend currently needs to implement `.legalize_compiler_rt_call`).
However, for testing purposes, I tried modifying the self-hosted x86_64
backend to enable all of the soft-float features (and implement the AIR
instruction). This modified backend was able to pass all of the behavior
tests (except for one `@mod` test where the LLVM backend has a bug
resulting in incorrect compiler-rt behavior!), including the tests
specific to the self-hosted x86_64 backend.
`f16` and `f80` legalizations are likely of particular interest to
backend developers, because most architectures do not have instructions
to operate on these types. However, enabling *all* of these legalization
passes can be useful when developing a new backend to hit the ground
running and pass a good amount of tests more easily.
I had tried unrolling the loops to avoid requiring the
`vector_store_elem` instruction, but it's arguably a problem to generate
O(N) code for an operation on `@Vector(N, T)`. In addition, that
lowering emitted a lot of `.aggregate_init` instructions, which is
itself a quite difficult operation to codegen.
This requires reintroducing runtime vector indexing internally. However,
I've put it in a couple of instructions which are intended only for use
by `Air.Legalize`, named `legalize_vec_elem_val` (like `array_elem_val`,
but for indexing a vector with a runtime-known index) and
`legalize_vec_store_elem` (like the old `vector_store_elem`
instruction). These are explicitly documented as *not* being emitted by
Sema, so need only be implemented by backends if they actually use an
`Air.Legalize.Feature` which emits them (otherwise they can be marked as
`unreachable`).
I started this diff trying to remove a little dead code from the C
backend, but ended up finding a bunch of dead code sprinkled all over
the place:
* `packed` handling in the C backend which was made dead by `Legalize`
* Representation of pointers to runtime-known vector indices
* Handling for the `vector_store_elem` AIR instruction (now removed)
* Old tuple handling from when they used the InternPool repr of structs
* Straightforward unused functions
* TODOs in the LLVM backend for features which Zig just does not support
The memory operand might use one of the extended GPRs R8 through R15 and
hence require a REX prefix, but having a REX prefix makes the high-byte
register AH unencodeable as the src operand. This latent bug was exposed
by this branch, presumably because `select` now happens to be putting
something in an extended GPR instead of a legacy GPR.
In theory this could be fixed with minimal cost by introducing a way to
communicate to `select` that neither the destination memory nor the
other temporary can be in an extended GPR. However, I just went for the
simple solution which comes at a cost of one trivial instruction: copy
the remainder from AH to AL, and *then* copy AL to the destination.
`rep movsb` isn't usually a great idea here. This commit makes the logic
which tentatively existed in `genInlineMemcpy` apply in more cases, and
in particular applies it to the "new" backend logic. Put simply, all
copies of 128 bytes or fewer will now attempt this path first,
where---provided there is an SSE register and/or a general-purpose
register available---we will lower the operation using a sequence of 32,
16, 8, 4, 2, and 1 byte copy operations.
The feedback I got on this diff was "Push it to master and if it
miscomps I'll revert it" so don't blame me when it explodes
