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.
This reverts commit 37114036aa57e53217a57afacd7f47b36114edfb.
The output of mbcnt does not depend on other active lanes, and hence it is not
convergent. The original change was made as a possible fix for
https://github.com/ROCm-Developer-Tools/HIP/issues/3172
But changing mbcnt does not fix that issue.
Reviewed By: ruiling, foad, yaxunl
Differential Revision: https://reviews.llvm.org/D153953
Atomic optimizer is turned on by default through D152649. This patch
removes the usage of old command line option amdgpu-atomic-optimizations
and transfer the responsibility to `amdgpu-atomic-optimizer-strategy`.
We can safely remove old option when LLPC remove its all usage.
Reviewed By: foad, arsenm, #amdgpu, cdevadas
Differential Revision: https://reviews.llvm.org/D153007
This patch provides an alternative implementation to DPP for Scan Computations.
An alternative implementation iterates over all active lanes of Wavefront
using llvm.cttz and performs the following steps:
1. Read the value that needs to be atomically incremented using
llvm.amdgcn.readlane intrinsic
2. Accumulate the result.
3. Update the scan result using llvm.amdgcn.writelane intrinsic
if intermediate scan results are needed later in the kernel.
Reviewed By: arsenm, cdevadas
Differential Revision: https://reviews.llvm.org/D147408
Add GFX11 test coverage to a bunch of tests where it was easy to do so,
mostly because the checks are autogenerated and/or GFX11 can share the
same checks as GFX10.
Differential Revision: https://reviews.llvm.org/D129295
Pre gfx1030 null for sdst is different.
c97436f8b6e2 [AMDGPU] Use null for dead sdst operand - requires a change to make
it not apply to pre gfx1030
Differential Revision: https://reviews.llvm.org/D127869
Nic Curtis done the experiments to prove it is faster than a
separate mul and add.
Fixes: SWDEV-332806
Differential Revision: https://reviews.llvm.org/D127253
Only fold for uniform values on pre-GFX9 chips. GFX9+ allow us
to keep the calculation entirely on the SALU.
For subtargets where integer multiplication isn't full-rate, avoid
folding if the multiply has too many uses.
Finally, we expand 64x32 and 64x64 multiplies here as well, if they
feed into an addition. This results in better code generation than
the generic expansion for such multiplies because we end up using
the accumulator of the MAD instructions.
Differential Revision: https://reviews.llvm.org/D123835
The WWM register has unmodeled register liveness, For v_set_inactive_*,
clobberring source register is dangerous because it will overwrite the
inactive lanes. When the source vgpr is dead at v_set_inactive_lane,
the inactive lanes may be not really dead. This may make common
optimizations doing wrong.
For example in a simple if-then cfg in Machine IR:
bb.if:
%src =
bb.then:
%src1 = COPY %src
%dst = V_SET_INACTIVE %src1(tied-def 0), %inactive
bb.end
... = PHI [0, %bb.then] [%src, %bb.if]
The register coalescer will think it is safe to optimize "%src1 = COPY %src"
in bb.then. And at the same time, there is no interference for the PHI in
bb.end. The source and destination values of the PHI will be assigned
the same register. The single PHI register will be overwritten by the
v_set_inactive, then we would get wrong value in bb.end.
With this change, we will copy the content of the source register before
setting inactive lanes after register allocation. Yes, this will sacrifice
the WWM code generation a little, but I don't have any better idea to do things
correctly.
Differential Revision: https://reviews.llvm.org/D117482
Using a BufferSize of one for memory ProcResources will result in better
ILP since it more accurately models the dependencies between memory ops
and their consumers on an in-order processor. After this change, the
scheduler will treat the data edges from loads as blocking so that
stalls are guaranteed when waiting for data to be retreaved from memory.
Since we don't actually track waitcnt here, this should do a better job
at modeling their behavior.
Practically, this means that the scheduler will trigger the 'STALL'
heuristic more often.
This type of change needs to be evaluated experimentally. Preliminary
results are positive.
Fixes: SWDEV-282962
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D114777
Select SelectionDAG ops smul_lohi/umul_lohi to
v_mad_i64_i32/v_mad_u64_u32 respectively, with an addend of 0.
v_mul_lo, v_mul_hi and v_mad_i64/u64 are all quarter-rate instructions
so it is better to use one instruction than two.
Further improvements are possible to make better use of the addend
operand, but this is already a strict improvement over what we have
now.
Differential Revision: https://reviews.llvm.org/D113986
The compiler was generating symbols in the final code object for local
branch target labels. This bloats the code object, slows down the loader,
and is only used to simplify disassembly.
Use '--symbolize-operands' with llvm-objdump to improve readability of the
branch target operands in disassembly.
Fixes: SWDEV-312223
Reviewed By: scott.linder
Differential Revision: https://reviews.llvm.org/D114273
Use GCNHazardRecognizer in postra sched.
Updated tests for the new schedules.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D109536
Change-Id: Ia86ba2ae168f12fb34b4d8efdab491f84d936cde
The requested register class priorities weren't respected
globally. Not sure why this is a target option, and not just the
expected behavior (recently added in
1a6dc92be7d68611077f0fb0b723b361817c950c). This avoids an allocation
failure when many wide tuple spills are introduced. I think this is a
workaround since I would not expect the allocation priority to be
required, and only a performance hint. The allocator should be smarter
about when only a subregister needs to be spilled and restored.
This does regress a couple of degenerate store stress lit tests which
shouldn't be too important.
Reuse the existing KnownBits multiplication code to handle the 'extend + multiply + extract high bits' pattern for multiply-high ops.
Noticed while looking at the codegen for D88785 / D98587 - the patch helps division-by-constant expansion code in particular, which suggests that we might have some further KnownBits div/rem cases we could handle - but this was far easier to implement.
Differential Revision: https://reviews.llvm.org/D98857
Prefer to keep uniform (non-divergent) multiplies on the scalar ALU when
possible. This significantly improves some game cases by eliminating
v_readfirstlane instructions when the result feeds into a scalar
operation, like the address calculation for a scalar load or store.
Since isDivergent is only an approximation of whether a value is in
SGPRs, it can potentially regress some situations where a uniform value
ends up in a VGPR. These should be rare in real code, although the test
changes do contain a number of examples.
Most of the test changes are just using s_mul instead of v_mul/mad which
is generally better for both register pressure and latency (at least on
GFX10 where sgpr pressure doesn't affect occupancy and vector ALU
instructions have significantly longer latency than scalar ALU). Some
R600 tests now use MULLO_INT instead of MUL_UINT24.
GlobalISel appears to handle more scenarios in the desirable way,
although it can also be thrown off and fails to select the 24-bit
multiplies in some cases.
Alternative solution considered and rejected was to allow selecting
MUL_[UI]24 to S_MUL_I32. I've rejected this because the definition of
those SD operations works is don't-care on the most significant 8 bits,
and this fact is used in some combines via SimplifyDemandedBits.
Based on a patch by Nicolai Hähnle.
Differential Revision: https://reviews.llvm.org/D97063
Exec mask manipulation inserted by SIWholeQuadMode barriers to
instruction scheduling. Move the entire pass after the machine
instruction scheduler and make changes so pass is correct for
non-SSA operation. These changes should leave the pass still
usable pre-scheduler, although tests have be updated to reflect
post-scheduler results.
Reviewed By: nhaehnle
Differential Revision: https://reviews.llvm.org/D88081
Add a new llvm.amdgcn.ballot intrinsic modeled on the ballot function
in GLSL and other shader languages. It returns a bitfield containing the
result of its boolean argument in all active lanes, and zero in all
inactive lanes.
This is intended to replace the existing llvm.amdgcn.icmp and
llvm.amdgcn.fcmp intrinsics after a suitable transition period.
Use the new intrinsic in the atomic optimizer pass.
Differential Revision: https://reviews.llvm.org/D65088
Summary:
Add support for gfx10, where all DPP operations are confined to work
within a single row of 16 lanes, and wave32.
Reviewers: arsenm, sheredom, critson, rampitec
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, t-tye, hiraditya, jfb, dstuttard, tpr, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D65644
llvm-svn: 369745
Summary:
r363675 changed the exec modification helper function, now called
execMayBeModifiedBeforeUse, so that if no UseMI is specified it checks
all instructions in the basic block, even beyond the last use. That
meant that the DPP combiner no longer worked in any basic block that
ended with a control flow instruction, and in particular it didn't work
on code sequences generated by the atomic optimizer.
Fix it by reinstating the old behaviour but in a new helper function
execMayBeModifiedBeforeAnyUse, and limiting the number of instructions
scanned.
Reviewers: arsenm, vpykhtin
Subscribers: kzhuravl, nemanjai, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, kbarton, MaskRay, jfb, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D64393
llvm-svn: 365910
This change incorporates an effort by Connor Abbot to change how we deal
with WWM operations potentially trashing valid values in inactive lanes.
Previously, the SIFixWWMLiveness pass would work out which registers
were being trashed within WWM regions, and ensure that the register
allocator did not have any values it was depending on resident in those
registers if the WWM section would trash them. This worked perfectly
well, but would cause sometimes severe register pressure when the WWM
section resided before divergent control flow (or at least that is where
I mostly observed it).
This fix instead runs through the WWM sections and pre allocates some
registers for WWM. It then reserves these registers so that the register
allocator cannot use them. This results in a significant register
saving on some WWM shaders I'm working with (130 -> 104 VGPRs, with just
this change!).
Differential Revision: https://reviews.llvm.org/D59295
llvm-svn: 357400
This commit fixes the DPP sequence in the atomic optimizer (which was
previously missing the row_shr:3 step), and works around a read_register
exec bug by using a ballot instead.
Differential Revision: https://reviews.llvm.org/D57737
llvm-svn: 353703
This commit adds a new IR level pass to the AMDGPU backend to perform
atomic optimizations. It works by:
- Running through a function and finding atomicrmw add/sub or uses of
the atomic buffer intrinsics for add/sub.
- If all arguments except the value to be added/subtracted are uniform,
record the value to be optimized.
- Run through the atomic operations we can optimize and, depending on
whether the value is uniform/divergent use wavefront wide operations
(DPP in the divergent case) to calculate the total amount to be
atomically added/subtracted.
- Then let only a single lane of each wavefront perform the atomic
operation, reducing the total number of atomic operations in flight.
- Lastly we recombine the result from the single lane to each lane of
the wavefront, and calculate our individual lanes offset into the
final result.
Differential Revision: https://reviews.llvm.org/D51969
llvm-svn: 343973
