This makes it easier to debug testcases:
- the entity numbers in a .cton file match the entity numbers used
within Cretonne.
- serializing and deserializing doesn't cause indices to change.
One disadvantage is that if a .cton file uses sparse entity numbers,
deserializing to the in-memory form doesn't compact it. However, the
text format is not intended to be performance-critical, so this isn't
expected to be a big burden.
A CallConv enum on every function signature makes it possible to
generate calls to functions with different calling conventions within
the same ISA / within a single function.
The calling conventions also serve as a way of customizing Cretonne's
behavior when embedded inside a VM. As an example, the SpiderWASM
calling convention is used to compile WebAssembly functions that run
inside the SpiderMonkey virtual machine.
All function signatures must have a calling convention at the end, so
this changes the textual IL syntax.
Before:
sig1 = signature(i32, f64) -> f64
After
sig1 = (i32, f64) -> f64 native
sig2 = (i32) spiderwasm
When printing functions, the signature goes after the return types:
function %r1() -> i32, f32 spiderwasm {
ebb1:
...
}
In the parser, this calling convention is optional and defaults to
"native". This is mostly to avoid updating all the existing test cases
under filetests/. When printing a function, the calling convention is
always included, including for "native" functions.
When making an outgoing call, some arguments may have to be passed on
the stack. Allocate OutgoingArg stack slots for these arguments and
write them immediately before the outgoing call instruction.
Do the same for incoming function arguments on the stack, but use
IncomingArg stack slots instead. This was previously done in the
spiller, but we move it to the legalizer so it is done at the same time
as outgoing stack arguments.
These stack slot assignments are done in the legalizer before live
range analysis because the outgoing arguments usually are in different
SSSA values with their own short live ranges.
ARM has all of these as scalar integer instructions. Intel has band_not
in SSE and as a scalar in BMI1.
Add the trivial legalization patterns that use a bnot instruction.
Register locations can change throughout an EBB. Make sure the
emit_inst() function considers this when encoding instructions and
update the register diversion tracker.
* Function names should start with %
* Create FunctionName from string
* Implement displaying of FunctionName as %nnnn with fallback to #xxxx
* Run rustfmt and fix FunctionName::with_string in parser
* Implement FunctionName::new as a generic function
* Binary function names should start with #
* Implement NameRepr for function name
* Fix examples in docs to reflect that function names start with %
* Rebase and fix filecheck tests
These special-purpose arguments and return values are only relevant for
the function being compiled, so add a `current` flag to
legalize_signature().
- Add the necessary argument values to the entry block to represent
the special-purpose arguments.
- Propagate the link and sret arguments to return instructions if the
legalized signature asks for it.
When converting from ABI types to original program types, the final
conversion instruction can place its result into the original value, so
it doesn't need to be changed to an alias.
Now that we can detach and reuse all values, there is no longer a need
to create a lot of alias values during pattern expansion. Instead, reuse
the values from the source pattern when emitting instructions in the
destination pattern.
If a destination instruction produces the exact same values as a source
instruction, simply leave the values attached and replace the
instruction it. Otherwise, detach the source values, reuse them in the
expansion, and remove the source instruction afterwards.
All values are now references into the value table, so drop the
distinction between direct and table values. Direct values don't exist
any more.
Also remove the parser support for the 'vxNN' syntax. Only 'vNN' values
can be parsed now.
We only ever create table values now.
Simplify legalizer::legalize_inst_results. Instead of calling
detach_secondary_results, just detach all the results and don't treat
the first result specially.
This affects the comparison instructions which now read "icmp ult a, b".
This mimics LLVM's style and makes it simpler to add instruction flags
in the future, such as "load v1" -> "load aligned v1".
These enumerated operands and flags feel like opcode modifiers rather
than value operands, so displaying them differently makes sense.
Value and numeric operands are still comma separated.
Not all br_icmp opcodes are present in the ISA. The missing ones can be
reached by commuting operands.
Don't attempt to encode EBB offsets yet. For now just emit an EBB
relocation for the branch instruction.
The carry and borrow values are boolean, so we have to convert them to
an integer type with bint(c) before we can add them to the result.
Also tweak the default legalizer action for unsupported types: Only
attempt a narrowing pattern for lane types > 32 bits.
This was found by @angusholder's new type checks in the verifier.
If an instruction doesn't have an associated encoding, use the standard
TargetIsa hook to encode it.
The test still fails if an instruction can't be encoded. There is no
legalization step.
The EBB argument splitting may generate concat-split dependencies when
it repairs branch arguments in EBBs that have not yet been fully
legalized. Add a branch argument simplification step that can resolve
these dependency chains.
This means that all split and concatenation instructions will be dead
after legalization for types that have no legal instructions using them.
When the legalizer splits a value into halves, it would previously stop
if the value was an EBB argument. With this change, we also split EBB
arguments and iteratively split arguments on branches to the EBB.
The iterative splitting stops when we hit the entry block arguments or
an instruction that isn't one of the concatenation instructions.
The legalizer often splits values into parts with the vsplit and
isplit_lohi instructions. Avoid doing that for values that are already
defined by the corresponding concatenation instructions.
This reduces the number of instructions created during legalization, and
it simplifies later optimizations. A number of dead concatenation
instructions are left behind. They can be trivially cleaned up by a dead
code elimination pass.
The call arguments on call_indirect should not include the fixed callee
argument.
Add legalizer assertions to verify that signatures are actually valid
after legalization. If not, we would get infinite legalizer loops.
Like the entry block arguments, the return values from a call
instruction need to be converted back from their ABI representation.
Add tests of call instruction legalization.
- abi.cton is for testing the actual RISC-V ABI.
- legalize-abi.cton is for testing the legalizer around ABI boundaries.
- parse-encoding.cton is for testing the parser's handling of RISC-V
encoding and register annotations.
The type signatures of functions can change when they are legalized for
a specific ABI. This means that all call and return instructions need to
be rewritten to use the correct arguments.
- Fix arguments to call instructions.
- Fix arguments to return instructions.
TBD:
- Fix return values from call instructions.
Add support for two new type variable functions: half_vector() and
double_vector().
Use these two instructions to break down unsupported SIMD types and
build them up again.
