Run the verify_contexti() function after invoking the legalize() and
regalloc() context functions. This will help catch bad code produced by
these passes.
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.
Legalizing some instructions may require modifications to the control
flow graph, and some operations need to use the CFG analysis.
The CFG reference is threaded through all the legalization functions to
reach the generated expansion functions as well as the legalizer::split
module where it will be used first.
When the CRETONNE_DBG environment variable is set, send debug messages
to a file named cretonne.dbg.*.
The trace facility is only enabled when debug assertions are on.
The Branch format also stores its fixed argument in the value list. This
requires the value pool to be passed to a few more functions.
Note that this actually makes the Branch and Jump variants of
InstructionData identical. The instruction format hashing does not yet
understand that all value operands are stored in the value list. We'll
fix that in a later patch.
Also convert IndirectCall, noting that Call and IndirectCall remain
separate instruction formats because they have different immediate
fields.
Add a Function::display() method which can include ISA-specific
information when printing the function.
If a test file has a unique ISA, use that in the `test cat`
implementation.
Move the flow graph computation into a compute method which can be
called with multiple functions.
This allows us to reuse the ControlFlowGraph memory and keep an instance
in the Context.
Also rework the algorithm to be more robust against unreachable blocks.
- Add an is_reachable(ebb) method.
- Change idom(ebb) to just return an instruction.
- Make idom() return None for the entry block as well as unreachable
blocks.
Include the test file preamble comments when building a filecheck
instance for every function in the file.
This makes it possible to define common regex variables in the preamble
and use these definitions for all the functions.
Give these crates each a more standard directory layout with sources in
a 'src' sub-sirectory and Cargo.toml in the top lib/foo directory.
Add license and description fields to each.
The build script for the cretonne crate now lives in
'lib/cretonne/build.rs' separating it from the normal library sources
under 'lib/cretonne/src'.
The 'lib/cretonne' directory will be the new root of a stand-alone
cretonne crate containg both Python and Rust sources.
This is in preparation for publishing crates on crates.io.
The Builder keeps track of a position in the layout and inserts new
instructions there.
Add insert_ebb() and ebb() methods to Builder.
Use Builder in the cfg tests.
The Builder provides a convenient interface for inserting instructions
into an extended basic block.
The bulk of the builder methods are generated automatically from the
meta language instruction descriptions.
Still TODO: Keep track of an insertion position.
In instruction formats that have multiple results AND boxed storage,
place the second_result field outside the boxed storage. The 16-byte
instruction format has room for opcode, type, second_result in the first
8 bytes, and the boxed pointer in the last 8 bytes.
This provides a simpler implementation of the second_result() and
second_result_mut() InstructionData methods.
These refer to external functions and function signatures declared in
the preamble. Since we're already using the type names 'Signature' and
'Function', these entity references don't folow the usual EntityData /
Entity naming convention.
The isplit_lohi instruction breaks an integer into two halves. This will
typically be used to get the two halves of an `i64` value on 32-bit
CPUs.
The iconcat_lohi is the reverse operation. It reconstructs the `i64`
from the low and high bits.
All instructions with associated encodings are now annotated with
encoding information in a column before the code.
When write_function() is givan a TargetIsa reference, the annotations use
ISA-specific names. Otherwise everything is numeric.
The # is a more conventional prefix for hexadecimal, and when ISA
information is not available, there may be a decimal number in front
which would be confusing.
So prefer [1#10c] for the ISA-less encoding format. Here '1' is decimal
and '#10c' is hexadecimal.
Add a `needs_isa()` method to the SubTest trait, and pass a TargetIsa
trait object to those sub-tests that request it.
When multiple sub-tests and ISAs are specified, test the cross product.
If a sub-test requires an ISA, but none are specified, fail the test. In
the future, it may be a good idea to generate a default set of ISAs and
test against those.
Some tests are only applicable to specific ISAs. This can be indicated
with an ISA specification:
test legalizer
isa riscv
function foo() {
....
}
The ISA specifications have the same format as the test lines: The name
of the ISA following by optional settings until the end of the line.
Also parse `set` commands mixed in with the `isa` commands. These are
used to set ISA-independent settings as defined in
meta/cretonne/settings.py.
When constructing the Flags object from the Builder, don't consume it,
but take a reference instead.
This makes it possible for the parser to accept multiple 'set' lines and
apply them to different ISA specifications.
