Adjust SITargetLowering::allowsMisalignedMemoryAccessesImpl for
unaligned flat scratch support. Mostly needed for global isel.
Differential Revision: https://reviews.llvm.org/D93669
The support is disabled by default. So far there is instruction
selection, spilling, and frame elimination. It also changes SP
from unswizzled to swizzled as used by flat scratch instructions,
so it cannot be mixed with MUBUF stack access.
At the very least missing:
- GlobalISel;
- Some optimizations in frame elimination in between vector
and scalar ALU;
- It shall finally allow to always materialize frame index
as an SGPR, but that is not implemented and frame elimination
cannot handle it yet;
- Unaligned and/or multidword flat scratch shall work, but it
is legalized now for MUBUF;
- Operand folding cannot optimize FI like with MUBUF yet;
- It will need scaling the value of the SP/FP in the DWARF
expression to recover the unswizzled scratch address;
Differential Revision: https://reviews.llvm.org/D89170
This reverts commit ca907bfb57.
According to michel.daenzer,
> This completely broke the Mesa radeonsi driver on Navi 14. Xorg +
> xterm come up with major corruption & psychedelic colours.
When memory operations are outstanding on function calls, either the
caller or the callee can insert a waitcnt to ensure that all reads are
finished.
Calls need some time to be executed, so if the callee inserts the
waitcnt, filling the instruction buffer and waiting for memory will be
interleaved, hiding some latency. This comes at the cost of having a
waitcnt inside functions that may not be needed as no memory operations
are outstanding.
For function calls, this is already implemented. The same principal
applies to returns: If the caller inserts a waitcnt after the call, the
callee does not have to wait and the return and memory operation can be
run in parallel.
This commit implements waiting in the caller after returning from a
function call.
Differential Revision: https://reviews.llvm.org/D87674
Features UnalignedBufferAccess and UnalignedDSAccess are now used to determine
whether hardware supports such access.
UnalignedAccessMode should be used to enable them.
hasUnalignedBufferAccessEnabled() and hasUnalignedDSAccessEnabled() can be
now used to quickly check both.
Differential Revision: https://reviews.llvm.org/D84522
tryLatency compares two sched candidates. For the top zone it prefers
the one with lesser depth, but only if that depth is greater than the
total latency of the instructions we've already scheduled -- otherwise
its latency would be hidden and there would be no stall.
Unfortunately it only tests the depth of one of the candidates. This can
lead to situations where the TopDepthReduce heuristic does not kick in,
but a lower priority heuristic chooses the other candidate, whose depth
*is* greater than the already scheduled latency, which causes a stall.
The fix is to apply the heuristic if the depth of *either* candidate is
greater than the already scheduled latency.
All this also applies to the BotHeightReduce heuristic in the bottom
zone.
Differential Revision: https://reviews.llvm.org/D72392
Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in
the entry function prologue. This allows us to removes the scratch wave
offset register from the calling convention ABI.
As part of this change, allow the use of an inline constant zero for the
SOffset of MUBUF instructions accessing the stack in entry functions
when a frame pointer is not requested/required. Entry functions with
calls still need to set up the calling convention ABI stack pointer
register, and reference it in order to address arguments of called
functions. The ABI stack pointer register remains unswizzled, but is now
wave-relative instead of queue-relative.
Non-entry functions also use an inline constant zero SOffset for
wave-relative scratch access, but continue to use the stack and frame
pointers as before. When the stack or frame pointer is converted to a
swizzled offset it is now scaled directly, as the scratch wave offset no
longer needs to be subtracted first.
Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling
convention.
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D75138
This is apparently worse than 1-byte alignment. This does not attempt
to decompose 2-byte aligned wide stores, but will stop trying to
produce them.
Also fix bug in LoadStoreVectorizer which was decreasing the alignment
and vectorizing stack accesses. It was assuming a stack object was an
alloca that could have its base alignment changed, which is not true
if the pointer is derived from a function argument.
Summary:
Mem op clustering adds a weak edge in the DAG between two loads or
stores that should be clustered, but the direction of this edge is
pretty arbitrary (it depends on the sort order of MemOpInfo, which
represents the operands of a load or store). This often means that two
loads or stores will get reordered even if they would naturally have
been scheduled together anyway, which leads to test case churn and goes
against the scheduler's "do no harm" philosophy.
The fix makes sure that the direction of the edge always matches the
original code order of the instructions.
Reviewers: atrick, MatzeB, arsenm, rampitec, t.p.northover
Subscribers: jvesely, wdng, nhaehnle, kristof.beyls, hiraditya, javed.absar, arphaman, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72706
Function return instruction lowering, currently uses the fixed register pair s[30:31] for holding
the return address. It can be any SGPR pair other than the CSRs. Created an SGPR pair sub-register class
exclusive of the CSRs, and used this regclass while lowering the return instruction.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D63924
llvm-svn: 365512
Every called function could possibly need this to calculate the
absolute address of stack objectst, and this avoids inserting a copy
around every call site in the kernel. It's also somewhat cleaner to
keep this in a callee saved SGPR.
llvm-svn: 363990
When matching half of the build_vector to a load, there could still be
a hidden dependency on the other half of the build_vector the pattern
wouldn't detect. If there was an additional chain dependency on the
other value, a cycle could be introduced.
I don't think a tablegen pattern is capable of matching the necessary
conditions, so move this into PreprocessISelDAG. Check isPredecessorOf
for the other value to avoid a cycle. This has a warning that it's
expensive, so this should probably be moved into an MI pass eventually
that will have more freedom to reorder instructions to help match
this. That is currently complicated by the lack of a computeKnownBits
type mechanism for the selected function.
llvm-svn: 355731
This avoids breaking possible value dependencies when sorting loads by
offset.
AMDGPU has some load instructions that write into the high or low bits
of the destination register, and have a tied input for the other input
bits. These can easily have the same base pointer, but be a swizzle so
the high address load needs to come first. This was inserting glue
forcing the opposite ordering, producing a cycle the InstrEmitter
would assert on. It may be potentially expensive to look for the
dependency between the other loads, so just skip any where this could
happen.
Fixes bug 40936 by reverting r351379, which added a hacky attempt to
fix this by adding chains in this case, which I think was just working
around broken glue before the InstrEmitter. The core of the patch is
re-implementing the fix for that problem.
llvm-svn: 355728
Summary:
For these loads that write to the HI part of a register, we should chain them to the op that writes to the LO part
of the register to maintain the appropriate order.
Reviewers:
rampitec, arsenm
Differential Revision:
https://reviews.llvm.org/D56454
llvm-svn: 351379