This patch adds support for controlling the post-RA machine scheduler
direction
(topdown, bottomup, bidirectional) on a per-function basis using the
"amdgpu-post-ra-direction" function attribute.
This patch updates `overrideSchedPolicy` and `overridePostRASchedPolicy`
to take a
`SchedRegion` parameter instead of just `NumRegionInstrs`. This provides
access to both the
instruction range and the parent `MachineBasicBlock`, which enables
looking up function-level
attributes.
With this change, targets can select post-RA scheduling direction per
function using a function
attribute. For example:
```cpp
void overridePostRASchedPolicy(MachineSchedPolicy &Policy,
const SchedRegion &Region) const {
const Function &F = Region.RegionBegin->getMF()->getFunction();
Attribute Attr = F.getFnAttribute("amdgpu-post-ra-direction");
...
}
This reapplies 067caaa and 382a085 (reverting b35f6e2) with fixes to
issues detected by the address sanitizer (MIs have to be removed from
live intervals before being removed from their parent MBB).
Original commit description below.
AMDGPU scheduler's `PreRARematStage` attempts to increase function
occupancy w.r.t. ArchVGPR usage by rematerializing trivial
ArchVGPR-defining instruction next to their single use. It first
collects all eligible trivially rematerializable instructions in the
function, then sinks them one-by-one while recomputing occupancy in all
affected regions each time to determine if and when it has managed to
increase overall occupancy. If it does, changes are committed to the
scheduler's state; otherwise modifications to the IR are reverted and
the scheduling stage gives up.
In both cases, this scheduling stage currently involves repeated queries
for up-to-date occupancy estimates and some state copying to enable
reversal of sinking decisions when occupancy is revealed not to
increase. The current implementation also does not accurately track
register pressure changes in all regions affected by sinking decisions.
This commit refactors this scheduling stage, improving RP tracking and
splitting the stage into two distinct steps to avoid repeated occupancy
queries and IR/state rollbacks.
- Analysis and collection (`canIncreaseOccupancyOrReduceSpill`). The
number of ArchVGPRs to save to reduce spilling or increase function
occupancy by 1 (when there is no spilling) is computed. Then,
instructions eligible for rematerialization are collected, stopping as
soon as enough have been identified to be able to achieve our goal
(according to slightly optimistic heuristics). If there aren't enough of
such instructions, the scheduling stage stops here.
- Rematerialization (`rematerialize`). Instructions collected in the
first step are rematerialized one-by-one. Now we are able to directly
update the scheduler's state since we have already done the occupancy
analysis and know we won't have to rollback any state. Register
pressures for impacted regions are recomputed only once, as opposed to
at every sinking decision.
In the case where the stage attempted to increase occupancy, and if both
rematerializations alone and rescheduling after were unable to improve
occupancy, then all rematerializations are rollbacked.
And related "[AMDGPU] Regenerate mfma-loop.ll test"
Introduce memory error detected by Asan #125885.
This reverts commit 382a085a95b0abeac77b150b7b644b372bd08e78.
This reverts commit 067caaafb58a156d0d77229422607782a639f5b5.
AMDGPU scheduler's `PreRARematStage` attempts to increase function
occupancy w.r.t. ArchVGPR usage by rematerializing trivial
ArchVGPR-defining instruction next to their single use. It first
collects all eligible trivially rematerializable instructions in the
function, then sinks them one-by-one while recomputing occupancy in all
affected regions each time to determine if and when it has managed to
increase overall occupancy. If it does, changes are committed to the
scheduler's state; otherwise modifications to the IR are reverted and
the scheduling stage gives up.
In both cases, this scheduling stage currently involves repeated queries
for up-to-date occupancy estimates and some state copying to enable
reversal of sinking decisions when occupancy is revealed not to
increase. The current implementation also does not accurately track
register pressure changes in all regions affected by sinking decisions.
This commit refactors this scheduling stage, improving RP tracking and
splitting the stage into two distinct steps to avoid repeated occupancy
queries and IR/state rollbacks.
- Analysis and collection (`canIncreaseOccupancyOrReduceSpill`). The
number of ArchVGPRs to save to reduce spilling or increase function
occupancy by 1 (when there is no spilling) is computed. Then,
instructions eligible for rematerialization are collected, stopping as
soon as enough have been identified to be able to achieve our goal
(according to slightly optimistic heuristics). If there aren't enough of
such instructions, the scheduling stage stops here.
- Rematerialization (`rematerialize`). Instructions collected in the
first step are rematerialized one-by-one. Now we are able to directly
update the scheduler's state since we have already done the occupancy
analysis and know we won't have to rollback any state. Register
pressures for impacted regions are recomputed only once, as opposed to
at every sinking decision.
In the case where the stage attempted to increase occupancy, and if both
rematerializations alone and rescheduling after were unable to improve
occupancy, then all rematerializations are rollbacked.
Previously the AnnotateKernelFeatures pass infers two attributes:
amdgpu-calls and amdgpu-stack-objects, which are used to help determine
if flat scratch init is allowed. PR #118907 created the
amdgpu-no-flat-scratch-init attribute. Continuing with that work, this
patch makes use of this attribute to determine flat scratch init,
replacing amdgpu-calls and amdgpu-stack-objects. This also leads to the
removal of the AnnotateKernelFeatures pass.
Occupancy (i.e., the number of waves per EU) depends, in addition to
register usage, on per-workgroup LDS usage as well as on the range of
possible workgroup sizes. Mirroring the latter, occupancy should
therefore be expressed as a range since different group sizes generally
yield different achievable occupancies.
`getOccupancyWithLocalMemSize` currently returns a scalar occupancy
based on the maximum workgroup size and LDS usage. With respect to the
workgroup size range, this scalar can be the minimum, the maximum, or
neither of the two of the range of achievable occupancies. This commit
fixes the function by making it compute and return the range of
achievable occupancies w.r.t. workgroup size and LDS usage; it also
renames it to `getOccupancyWithWorkGroupSizes` since it is the range of
workgroup sizes that produces the range of achievable occupancies.
Computing the achievable occupancy range is surprisingly involved.
Minimum/maximum workgroup sizes do not necessarily yield maximum/minimum
occupancies i.e., sometimes workgroup sizes inside the range yield the
occupancy bounds. The implementation finds these sizes in constant time;
heavy documentation explains the rationale behind the sometimes
relatively obscure calculations.
As a justifying example, consider a target with 10 waves / EU, 4 EUs/CU,
64-wide waves. Also consider a function with no LDS usage and a flat
workgroup size range of [513,1024].
- A group of 513 items requires 9 waves per group. Only 4 groups made up
of 9 waves each can fit fully on a CU at any given time, for a total of
36 waves on the CU, or 9 per EU. However, filling as much as possible
the remaining 40-36=4 wave slots without decreasing the number of groups
reveals that a larger group of 640 items yields 40 waves on the CU, or
10 per EU.
- Similarly, a group of 1024 items requires 16 waves per group. Only 2
groups made up of 16 waves each can fit fully on a CU ay any given time,
for a total of 32 waves on the CU, or 8 per EU. However, removing as
many waves as possible from the groups without being able to fit another
equal-sized group on the CU reveals that a smaller group of 896 items
yields 28 waves on the CU, or 7 per EU.
Therefore the achievable occupancy range for this function is not [8,9]
as the group size bounds directly yield, but [7,10].
Naturally this change causes a lot of test churn as instruction
scheduling is driven by achievable occupancy estimates. In most unit
tests the flat workgroup size range is the default [1,1024] which,
ignoring potential LDS limitations, would previously produce a scalar
occupancy of 8 (derived from 1024) on a lot of targets, whereas we now
consider the maximum occupancy to be 10 in such cases. Most tests are
updated automatically and checked manually for sanity. I also manually
changed some non-automatically generated assertions when necessary.
Fixes#118220.
This is a refinement for the existing hack. With this,
the default target will have neither wavefrontsize feature
present, unless it was explicitly specified. That is,
getWavefrontSize() == 64 no longer implies +wavefrontsize64.
getWavefrontSize() == 32 does imply +wavefrontsize32.
Continue to assume the value is 64 with no wavesize feature.
This maintains the codegenable property without any code
that directly cares about the wavesize needing to worry about it.
Introduce an isWaveSizeKnown helper to check if we know the
wavesize is accurate based on having one of the features explicitly
set, or a known target-cpu.
I'm not sure what's going on in wave_any.s. It's testing what
happens when both wavesizes are enabled, but this is treated
as an error in codegen. We now treat wave32 as the winning
case, so some cases that were previously printed as vcc are now
vcc_lo.
In graphics shaders it is better overall to use NSA encoding for IMAGE
instructions, because the benefit of less constrained register
allocation outweighs the cost of larger encoding. In particular NSA form
often avoids the need for extra V_MOV_B32 instructions between IMAGE
instructions, which can allow the IMAGE instructions to be claused.
Note that in GFX12 there is no longer a bit in the encoding to choose
between NSA and non-NSA forms, so this only affects GFX10 and GFX11.
Change the names of the TableGen features to match the names used by
AMDGPUSubtarget. "Addressable" refers to the amount that can be accessed
by a single workgroup. Add some explanatory comments. NFC.
The generic subtarget has neither of these features. Rather than forcing
HasMovrel on, it is simpler to expand dynamic vector indexing to a
sequence of compare/select instructions.
NFC for real subtargets.
