| Age | Commit message (Collapse) | Author |
|---|
|
When we process one block, we
start by allocating registers
for all the temporaries live
at the exit of the block.
Before this patch we processed
temps first, then in doblk() we
would mark globally live registers
allocated. This meant that temps
could get wrongly assigned a live
register.
The fix is simple: we now process
registers first at block exits,
then allocate temps.
|
|
This reverts commit 028534d9897079bf64559dca0402bc59a956ce46.
riscv64 will have jump arguments with type RTmp.
|
|
|
|
On both amd64 & arm64, the jumps
making it to rega won't have any
argument.
|
|
|
|
|
|
In fact, after spilling, a phi
can be a temporary or a slot.
I am now pondering whether this
is a good idea or not because
it causes annoying mem->mem
movs after register allocation.
|
|
Since ce0ab53ed7, we skip over predecessors that spilled the
temporary. However, if all predecessors spilled, then we might not have
an entry in `rl`, triggering an assertion failure in the following loop.
|
|
The worst one was that "part 3" of rega()
could break the critical invariant that
two interferring temporaries get assigned
different registers. This is fixed by
being careful when changing the register
of a temporary based on predecessor
blocks.
Thanks to Michael Forney for reporting
these bugs and helping with the analysis.
|
|
That was silly... I believe qbe still
managed to work because bitsets are only
used inside a basic block where rcopy()
is not used.
|
|
Compiler warned about comparison between signed and unsigned values.
|
|
The previous heuristics were ad hoc and it was
hard to understand why they worked at all.
This patch can be summarized in three points:
1. When a register is freed (an instruction
assigns it), we try to find if a temporary
would like to be in it, and if we find one,
we move it in the newly freed register.
I call this an "eager move".
2. Temporaries now remember in what register
they were last allocated; this information
is stored in the field Tmp.visit, and
prevails on the field Tmp.hint when it is
set. (This makes having the same hint for
interfering temporaries not so disastrous.)
3. Blocks are now allocated in "onion" order,
from the innermost loop to the outermost.
This is the change I am the least sure
about; it should be evaluated thorougly.
|
|
This big diff does multiple changes to allow
the addition of new targets to qbe. The
changes are listed below in decreasing order
of impact.
1. Add a new Target structure.
To add support for a given target, one has to
implement all the members of the Target
structure. All the source files where changed
to use this interface where needed.
2. Single out amd64-specific code.
In this commit, the amd64 target T_amd64_sysv
is the only target available, it is implemented
in the amd64/ directory. All the non-static
items in this directory are prefixed with either
amd64_ or amd64_sysv (for items that are
specific to the System V ABI).
3. Centralize Ops information.
There is now a file 'ops.h' that must be used to
store all the available operations together with
their metadata. The various targets will only
select what they need; but it is beneficial that
there is only *one* place to change to add a new
instruction.
One good side effect of this change is that any
operation 'xyz' in the IL now as a corresponding
'Oxyz' in the code.
4. Misc fixes.
One notable change is that instruction selection
now generates generic comparison operations and
the lowering to the target's comparisons is done
in the emitter.
GAS directives for data are the same for many
targets, so data emission was extracted in a
file 'gas.c'.
5. Modularize the Makefile.
The Makefile now has a list of C files that
are target-independent (SRC), and one list
of C files per target. Each target can also
use its own 'all.h' header (for example to
define registers).
|
|
The register allocation now has stricter assertions
about global registers. The stricter assertions
required changes in the spiller: We now correctly
indicate to the register allocator what registers
are used by "ret" instructions.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
