Let's use the `.space` directive to simplify AVR's branch relaxation
tests, as noticed by @MaskRay at
https://github.com/llvm/llvm-project/pull/118015#issuecomment-2728254164.
Having `.space 4100` instead of `.space 4096` is little bit awkward, but
I'd rather avoid changing two things at once (i.e. current assertions
are fine).
This fixes the handling of subregister extract copies. This
will allow AMDGPU to remove its implementation of
shouldRewriteCopySrc, which exists as a 10 year old workaround
to this bug. peephole-opt-fold-reg-sequence-subreg.mir will
show the expected improvement once the custom implementation
is removed.
The copy coalescing processing here is overly abstracted
from what's actually happening. Previously when visiting
coalescable copy-like instructions, we would parse the
sources one at a time and then pass the def of the root
instruction into findNextSource. This means that the
first thing the new ValueTracker constructed would do
is getVRegDef to find the instruction we are currently
processing. This adds an unnecessary step, placing
a useless entry in the RewriteMap, and required skipping
the no-op case where getNewSource would return the original
source operand. This was a problem since in the case
of a subregister extract, shouldRewriteCopySource would always
say that it is useful to rewrite and the use-def chain walk
would abort, returning the original operand. Move the process
to start looking at the source operand to begin with.
This does not fix the confused handling in the uncoalescable
copy case which is proving to be more difficult. Some currently
handled cases have multiple defs from a single source, and other
handled cases have 0 input operands. It would be simpler if
this was implemented with isCopyLikeInstr, rather than guessing
at the operand structure as it does now.
There are some improvements and some regressions. The
regressions appear to be downstream issues for the most part. One
of the uglier regressions is in PPC, where a sequence of insert_subrgs
is used to build registers. I opened #125502 to use reg_sequence instead,
which may help.
The worst regression is an absurd SPARC testcase using a <251 x fp128>,
which uses a very long chain of insert_subregs.
We need improved subregister handling locally in PeepholeOptimizer,
and other pasess like MachineCSE to fix some of the other regressions.
We should handle subregister composes and folding more indexes
into insert_subreg and reg_sequence.
Similar to 806761a7629df268c8aed49657aeccffa6bca449
-mtriple= specifies the full target triple while -march= merely sets the
architecture part of the default target triple, leaving a target triple which
may not make sense.
Therefore, -march= is error-prone and not recommended for tests without a target
triple. The issue has been benign as we recognize avr-apple-darwin as ELF instead
of rejecting it outrightly.
Some targets (e.g. PPC and Hexagon) already did this. I think it's best
to do this consistently so that frontend authors don't run into
inconsistent results when they emit `naked` functions. For example, in
Zig, we had to change our emit code to also set `frame-pointer=none` to
get reliable results across targets.
Note: I don't have commit access.
Ever since 6859685a87ad093d60c8bed60b116143c0a684c7 (or, precisely,
84428dafc0941e3a31303fa1b286835ab2b8e234) relative jumps emitted by the
AVR codegen are off by two bytes - this pull request fixes it.
## Abstract
As compared to absolute jumps, relative jumps - such as rjmp, rcall or
brsh - have an implied `pc+2` behavior; that is, `jmp 100` is `pc =
100`, but `rjmp 100` gets understood as `pc = pc + 100 + 2`.
This is not reflected in the AVR codegen:
f95026dbf6/llvm/lib/Target/AVR/MCTargetDesc/AVRAsmBackend.cpp (L89)
... which always emits relative jumps that are two bytes too far - or
rather it _would_ emit such jumps if not for this check:
f95026dbf6/llvm/lib/Target/AVR/MCTargetDesc/AVRAsmBackend.cpp (L517)
... which causes most of the relative jumps to be actually resolved
late, by the linker, which applies the offsetting logic on its own,
hiding the issue within LLVM.
[Some time
ago](697a162fa6)
we've had a similar "jumps are off" problem that got solved by touching
`shouldForceRelocation()`, but I think that has worked only by accident.
It's exploited the fact that absolute vs relative jumps in the parsed
assembly can be distinguished through a "side channel" check relying on
the existence of labels (i.e. absolute jumps happen to named labels, but
relative jumps are anonymous, so to say). This was an alright idea back
then, but it got broken by 6859685a87ad093d60c8bed60b116143c0a684c7.
I propose a different approach:
- when emitting relative jumps, offset them by `-2` (well, `-1`,
strictly speaking, because those instructions rely on right-shifted
offset),
- when parsing relative jumps, treat `.` as `+2` and read `rjmp .+1234`
as `rjmp (1234 + 2)`.
This approach seems to be sound and now we generate the same assembly as
avr-gcc, which can be confirmed with:
```cpp
// avr-gcc test.c -O3 && avr-objdump -d a.out
int main() {
asm(
" foo:\n\t"
" rjmp .+2\n\t"
" rjmp .-2\n\t"
" rjmp foo\n\t"
" rjmp .+8\n\t"
" rjmp end\n\t"
" rjmp .+0\n\t"
" end:\n\t"
" rjmp .-4\n\t"
" rjmp .-6\n\t"
" x:\n\t"
" rjmp x\n\t"
" .short 0xc00f\n\t"
);
}
```
avr-gcc is also how I got the opcodes for all new tests like `inst-brbc.s`, so we should be good.
Previously, we just check if the source is a virtual register and
this prevents some potential hoists.
We can see some improvements in AArch64/RISCV tests.
PR #66334 tried to renumber slot indexes before register allocation, but
the numbering was still affected by list entries for instructions which
had been erased. Fix this to make the register allocator's live range
length heuristics even less dependent on the history of how instructions
have been added to and removed from SlotIndexes's maps.
If a virtual register is not assigned preferred physical register, it means some
COPY instructions will be changed to real register move instructions. In this
case we can try to split the virtual register in colder blocks, if success, the
original COPY instructions can be deleted, and the new COPY instructions in
colder blocks will be generated as register move instructions. It results in
fewer dynamic register move instructions executed.
The new test case split-reg-with-hint.ll gives an example, the hot path contains
24 instructions without this patch, now it is only 4 instructions with this
patch.
Differential Revision: https://reviews.llvm.org/D156491
The issue is uncovered by #47698: for IR files without a target triple,
-mtriple= specifies the full target triple while -march= merely sets the
architecture part of the default target triple, leaving a target triple which
may not make sense, e.g. riscv64-apple-darwin.
Therefore, -march= is error-prone and not recommended for tests without a target
triple. The issue has been benign as we recognize $unknown-apple-darwin as ELF instead
of rejecting it outrightly.
Currently our AVRShiftExpand pass expands only 32-bit shifts, with the
assumption that other kinds of shifts (e.g. 64-bit ones) are
automatically reduced to 8-bit ones by LLVM during ISel.
However this is not always true and causes problems in the rust-lang runtime.
This commit changes the logic a bit, so that instead of expanding only
32-bit shifts, we expand shifts of all types except 8-bit and 16-bit.
This is not the most optimal solution, because 64-bit shifts can be
expanded to 32-bit shifts which has been deeply optimized.
I've checked the generated code using rustc + simavr, and all shifts
seem to behave correctly.
Spotted in the wild in rustc:
https://github.com/rust-lang/compiler-builtins/issues/523https://github.com/rust-lang/rust/issues/112140
Reviewed By: benshi001
Differential Revision: https://reviews.llvm.org/D154785
This patch fixes the failed test of verifyInstructionPredicates which is caused by verifyInstructionPredicates. verifyInstructionPredicates will add JMPk without checking the target predicate.
Reviewed By: benshi001
Differential Revision: https://reviews.llvm.org/D155570
Since ROLBRd needs an implicit R1 (on AVR) or an implicit R17 (on AVRTiny),
we split ROLBRd to ROLBRdR1 (on AVR) and ROLBRdR17 (on AVRTiny).
Reviewed By: aykevl, Patryk27
Differential Revision: https://reviews.llvm.org/D152248
This is a follow-up to b71edfaa4ec3c998aadb35255ce2f60bba2940b0
since I forgot the lit.local.cfg files in that one.
Reformatting is done with `black`.
If you end up having problems merging this commit because you
have made changes to a python file, the best way to handle that
is to run git checkout --ours <yourfile> and then reformat it
with black.
If you run into any problems, post to discourse about it and
we will try to help.
RFC Thread below:
https://discourse.llvm.org/t/rfc-document-and-standardize-python-code-style
Reviewed By: barannikov88, kwk
Differential Revision: https://reviews.llvm.org/D150762
The 'ELPM' instruction has three forms:
--------------------------
| form | feature |
| ----------- | -------- |
| ELPM | hasELPM |
| ELPM Rd, Z | hasELPMX |
| ELPM Rd, Z+ | hasELPMX |
--------------------------
The second form is always used in the expansion of pseudo instructions
LPMWRdZ/ELPMWRdZ. But for devices without ELPMX and with only ELPM,
only the first form can be used.
Reviewed By: aykevl, Miss_Grape
Differential Revision: https://reviews.llvm.org/D141264
Define and use a MachineOperand overload of isPlainlyKilled. This
improves codegen in a couple of tests because it catches the case where
MO does not kill Reg but another operand of the same instruction does.
Differential Revision: https://reviews.llvm.org/D147167
The 'LPM' instruction has three forms:
------------------------
| form | feature |
| ---------- | --------|
| LPM | hasLPM |
| LPM Rd, Z | hasLPMX |
| LPM Rd, Z+ | hasLPMX |
------------------------
The second form is always selected in ISelDAGToDAG, even on devices
without FeatureLPMX. This patch emits "LPM + MOV" on devices with
only FeatureLPM.
Reviewed By: jacquesguan
Differential Revision: https://reviews.llvm.org/D141246
The 'ELPM' instruction has three forms:
--------------------------
| form | feature |
| ----------- | -------- |
| ELPM | hasELPM |
| ELPM Rd, Z | hasELPMX |
| ELPM Rd, Z+ | hasELPMX |
--------------------------
The second form is always used in the expansion of the pseudo
instruction 'ELPMBRdZ'. But for devices without ELPMX but only
with ELPM, only the first form can be emitted.
Reviewed By: jacquesguan
Differential Revision: https://reviews.llvm.org/D141221
In AVRFrameLowering::spillCalleeSavedRegisters(), when a 16-bit
livein register is spilled, two PUSH instructions are generated
for the higher and lower 8-bit registers. But these two 8-bit
registers are marked as killed in the two PUSH instructions, so
any future use of them will cause a crash.
This patch fixes the above issue by adding the two sub 8-bit
registers to the livein list.
Fixes https://github.com/llvm/llvm-project/issues/56423
Reviewed By: jacquesguan
Differential Revision: https://reviews.llvm.org/D144720
All hardware address spaces on AVR can be freely cast between (they keep
the same bit pattern). They just aren't dereferenceable when they're in
a different address space as they really do point to a separate address
space.
This is supported in avr-gcc: https://godbolt.org/z/9Gfvhnhv9
avr-gcc also supports the `__memx` address space which is 24 bits. We
don't support this address space yet but I've added a safeguard just in
case.
Differential Revison: https://reviews.llvm.org/D142107
This pseudo-instruction stores two small (8-bit) registers into one wide
(16-bit) register. But apparently the order matters a lot to the
register allocator.
This patch changes the order of inserting the registers to optimize for
the best register allocation in the tests of shift32.ll. It might be
detrimental in other cases, but keeping the registers in the same
physical register seems like it would be a common case.
Differential Revision: https://reviews.llvm.org/D140573
This optimization turns shifts of almost a multiple of 8 into a shift
into the opposite direction. Unfortunately it doesn't compose well with
the other optimizations (I've tried) so it's separate from them.
Differential Revision: https://reviews.llvm.org/D140572
This uses a complicated shift sequence that avr-gcc also uses, but
extended to work over any number of bytes and in both directions
(logical shift left and logical shift right). Unfortunately it can't be
used for an arithmetic shift right: I've tried to come up with a
sequence but couldn't.
Differential Revision: https://reviews.llvm.org/D140571
This patch optimizes 32-bit constant shifts by renaming registers. This
is very effective as the compiler would otherwise need to do a lot of
single bit shift instructions. Instead, the registers are renamed at the
SSA level which means the register allocator will insert the necessary
mov instructions.
Unfortunately, the register allocator will insert some unnecessary movs
with the current code. This will be fixed in a later patch.
Differential Revision: https://reviews.llvm.org/D140570
