This includes gfx908 which only has a no-return version of the
global_atomic_add_f32 instruction, using the same hack that was
previously implemented for selecting from the
llvm.amdgcn.global.atomic.fadd intrinsic.
Differential Revision: https://reviews.llvm.org/D97767
Look throught G_PTRTOINT and G_PTR_ADD nodes when looking for constant
offset for buffer stores. This also helps with merging of these instructions
later on.
Differential Revision: https://reviews.llvm.org/D95242
Returning int64_t was arbitrarily limiting for wide integer types, and
the functions should handle the full generality of the IR.
Also changes the full form which returns the originally defined
vreg. Add another wrapper for the common case of just immediately
converting to int64_t (arguably this would be useful for the full
return value case as well).
One possible issue with this change is some of the existing uses did
break without conversion to getConstantVRegSExtVal, and it's possible
some without adequate test coverage are now broken.
It does not seem to fold offsets but this is not specific
to the flat scratch as getPtrBaseWithConstantOffset() does
not return the split for these tests unlike its SDag
counterpart.
Differential Revision: https://reviews.llvm.org/D93670
Remove immediate operand from SI_ELSE which indicates if EXEC has
been modified. Instead always emit code that handles EXEC and
remove unnecessary instructions during pre-RA optimisation.
This facilitates passes (i.e. SIWholeQuadMode) adding exec mask
manipulation post control flow lowering, and pre control flow
lower passes do not need to be aware of SI_ELSE handling.
Reviewed By: nhaehnle
Differential Revision: https://reviews.llvm.org/D89644
Summary:
This implements a workaround for a hardware bug in gfx8 and gfx9,
where register usage is not estimated correctly for image_store and
image_gather4 instructions when D16 is used.
Change-Id: I4e30744da6796acac53a9b5ad37ac1c2035c8899
Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81172
uint8_t types are implicitly promoted to int, leading to a
unsigned-signed comparison.
Thanks for the heads-up @uabelho.
Differential Revision: https://reviews.llvm.org/D88876
Use tablegen generic tables to get the index of image intrinsic
arguments.
Before, the computation of which image intrinsic argument is at which
index was scattered in a few places, tablegen, the SDag instruction
selection and GlobalISel. This patch changes that, so only tablegen
contains code to compute indices and the ImageDimIntrinsicInfo table
provides these information.
Differential Revision: https://reviews.llvm.org/D86270
The addend in a REL32 reloc needs to be adjusted to account for the
offset from the PC value returned by the s_getpc instruction to the
point where the reloc is applied. This was being done correctly for
(GOTPC)REL32_LO but not for (GOTPC)REL32_HI. This will only make a
difference if the target symbol happens to get loaded almost exactly
a multiple of 4G away from the relocated instructions.
Differential Revision: https://reviews.llvm.org/D86938
If the condition output is negated, swap the branch targets. This is
similar to what SelectionDAG does for when SelectionDAGBuilder
decides to invert the condition and swap the branches.
This is leaving behind a dead constant def for some reason.
Most notably, we were incorrectly reporting <3 x s16> as a legal type
for these. Make sure these aren't legal to help make progress on
fixing the artifact combiner and vector legalizer
rules. Unfortunately, this means spreading the -global-isel-abort=0
hack, although this doesn't change the legalizer result in any
situation.
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
Summary:
- HIP uses an unsized extern array `extern __shared__ T s[]` to declare
the dynamic shared memory, which size is not known at the
compile time.
Reviewers: arsenm, yaxunl, kpyzhov, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, dstuttard, tpr, t-tye, hiraditya, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82496
Assuming this is used to split a memory access into smaller pieces,
the new access should still have the same aliasing properties as the
original memory access. As far as I can tell, this wasn't
intentionally dropped. It may be necessary to drop this if you are
moving the operand outside of the bounds of the original object in
such a way that it may alias another IR object, but I don't think any
of the existing users are doing this. Some of the uses widen into
unused alignment padding, which I think is OK.
Custom lower and widen odd sized loads up to the alignment. The
default set of legalization actions doesn't have a way to represent
this. This fixes naturally aligned <3 x s8> and <3 x s16> loads.
This also starts moving towards eliminating the buggy and
overcomplicated legalization rules for narrowing. All the memory size
changes should be done in the lower or custom action, not NarrowScalar
/ FewerElements. These currently have redundant and ambiguous code
with the lower action.
This mirrors the support for the equivalent extracts. This also
creates a huge mess that would be greatly improved if we had any bit
operation combines.
Use the same basic strategy as LegalizeVectorTypes. Try to index into
smaller pieces if there's a constant index, and otherwise fall back to
a stack temporary.
If we were to have an operation with an s16 def that needs to be
executed in a waterfall loop, not having s16 legal would place an
avoidable burden on RegBankSelect to widen it.
Get the argument register and ensure there's a copy to the virtual
register. AMDGPU and AArch64 have similarish code to get the livein
value, and I also want to use this in multiple places.
This is a bit more aggressive about setting the register class than
the original function, but that's probably OK.
I think we're missing a few verifier checks for function live ins. I
noticed AArch64's calling convention code is not actually adding
liveins to functions, only the entry block (which apparently might not
matter that much?). There should probably be a verifier check that
entry block live ins are also live into the function. We also might
need a verifier check that the copy to the livein virtual register is
in the entry block.
For AMDGPU, vectors with elements < 32 bits should be indexed in
32-bit elements and the desired bits extracted from there. For
elements > 64-bits, these should be reduce to 64/32 elements to enable
the normal dynamic indexing paths.
In the dynamic index cases, this produces shorter code most of the
time. This does immediately regress the constant index cases, but this
should be fixed once we have the most basic of shift combines.
The element size > 64 case is pretty much ported from the exisiting
DAG implementation for extract element promote. The increasing element
size case is new.
I still think it's highly questionable that we have two intrinsics
with identical behavior and only vary by the name of the libcall used
if it happens to be lowered that way, but try to reduce the feature
delta between SDAG and GlobalISel for recently added intrinsics. I'm
not sure which opcode should be considered the canonical one, but
lower roundeven back to round.
These aren't implemented and we're still relying on the AtomicExpand
pass, but mark these as lower to eliminate a few of the few remaining
no rules defined cases.
We don't really need these asserts. The LegalizerInfo is also
overly-aggressivly constructed, even when not in use. It needs to not
assert on dummy targets that have manually specified, unrelated
features.
Widen or narrow a type to a type with the same scalar size as
another. This can be used to force G_PTR_ADD/G_PTRMASK's scalar
operand to match the bitwidth of the pointer type. Use this to
disallow narrower types for G_PTRMASK.
Add support in LegalizerHelper for lowering G_SADDSAT etc. either
using add/subtract-with-overflow or using max/min instructions.
Enable this lowering for AMDGPU so it can be tested. The legalization
rules are still approximate and skips out on using the clamp bit to
treat these as legal, which has never been used before. This also
doesn't yet try to deal with expanding SALU cases.
Add narrowScalarFor action.
Add narrow scalar for typeIndex == 0 for G_FPTOSI/G_FPTOUI.
Legalize using narrowScalarFor as s16->s32 G_FPTOSI/G_FPTOUI
followed by s32->s64 G_SEXT/G_ZEXT.
Differential Revision: https://reviews.llvm.org/D84010
Add widenScalar for TypeIdx == 0 for G_SITOFP/G_UITOFP.
Legailize, using widenScalar, as s64->s32 G_SITOFP/G_UITOFP
followed by s32->s16 G_FPTRUNC.
Differential Revision: https://reviews.llvm.org/D83880
This avoids many instances of failing to legalize a vector truncstore
of <4 x s8> to 2 bytes. We don't perfectly handle every truncstore
yet, largely because the given set of legalization actions can't
actually differentiate between changing the result type and changing
the memory type.
This function is deceptive at best: it doesn't return what you'd expect.
If you have an arbitrary GlobalValue and you want to determine the
alignment of that pointer, Value::getPointerAlignment() returns the
correct value. If you want the actual declared alignment of a function
or variable, GlobalObject::getAlignment() returns that.
This patch switches all the users of GlobalValue::getAlignment to an
appropriate alternative.
Differential Revision: https://reviews.llvm.org/D80368
This was passing in all the parameters needed to construct a
LegalizerHelper in the custom legalization, when it's simpler to just
pass in the existing helper.
This is slightly more annoying to use in the common case where you
don't need the legalizer helper, but we could add back the common
parameters back in addition to the helper.
I didn't propagate this to all the internal target changes that this
logically implies, but did update a sample one for
legalizeMinNumMaxNum.
This is in preparation for moving AMDGPU load/store legalization
entirely into custom lowering. The current set of legalization actions
is really constraining and not really capable of expressing all the
actions needed to legalize loads/stores. In particular there's no way
to express when the memory access itself needs to change size vs. the
result type. There's also a lot of redundancy since the same
split/widen actions need to be applied in both vector and scalar
cases. All of the sub-cases logically belong as steps in the legalizer
helper, but it will be easier to consider everything at once in custom
lowering.
The logic is written for what loads/stores should be selectable. There
are a set of cases that should be selectable, but due to missing MVTs
and/or selection patterns, will fail to select. I think eventually
load/store select patterns should ignore the type and only look at the
value size, but until that happens, bitcast these to equivalent i32
vectors.
This was implicitly assuming the branch instruction was the next after
the pseudo. It's possible for another non-terminator instruction to be
inserted between the intrinsic and the branch, so adjust the insertion
point. Fixes a non-terminator after terminator verifier error (which
without the verifier, manifested itself as an infinite loop in
analyzeBranch much later on).
The baffling thing is this passed the OpenCL conformance test for
32-bit integer divisions, but only failed in the 32-bit path of
BypassSlowDivisions for the 64-bit tests.
This was promoting booleans to i32 to perform a comparison against
them to feed to a select condition. Just use the booleans
directly. This produces the same final code, since the combiner is
unable to undo the mess this creates. I untangled this logic when I
ported this code to GlobalISel, so port the cleanups back.
It was annoying enough that every custom lowering needed to set the
insert point, but this was made worse since now these all needed to be
updated to setInstrAndDebugLoc. Consolidate these so every
legalization action has the right insert position by default.
This should fix dropping debug info in every custom AMDGPU
legalization.
The current set is an incomprehensible mess riddled with ordering
hacks for various limitations in the legalizer at the time of writing,
many of which have been fixed. This takes a very small step in
correcting this.
The core first change is to start checking for fully legal cases
first, rather than trying to figure out all of the actions that could
need to be performed. It's recommended to check the legal cases first
for faster legality checks in the common case. This still has a table
listing some common cases, but it needs measuring whether this really
helps or not.
More significantly, stop trying to allow any arbitrary type with a
legal bitwidth as a legal memory type, and start using the bitcast
legalize action for them. Allowing loads of these weird vector types
produced new burdens we don't need for handling all of the
legalization artifacts. Unlike the SelectionDAG handling, this is
still not casting 64 or 16-bit element vectors to 32-bit
vectors. These cases should still be handled by increasing/decreasing
the number of 16-bit elements. This is primarily to fix 8-bit element
vectors.
Another change is to stop trying to handle the load-widening based on
a higher alignment. We should still do this, but the way it was
handled wasn't really correct. We really need to modify the MMO's size
at the same time, and not just increase the result type. The
LegalizerHelper does not do this, and I think this would really
require a separate WidenMemory action (or to add a memory action
payload to the LegalizeMutation). These will now fail to legalize.
The structure of the legalizer rules makes writing concise rules here
difficult. It would be easier if the same function could answer the
query the query, and report the action to perform at the same
time. Instead these two are split into distinct predicate and action
functions. This is mostly tolerable for other cases, but the
load/store rules get pretty complicated so it's difficult to keep two
versions of these functions in sync.
Tweak a few constant expressions involving numbers::pi etc to avoid
rounding errors. NFCI though it's possible some of these will now be
more accurate in the last bit.
I get confused by a lot of the predicate names here, since I would
assume they apply to vectors as well. Rename to reflect they only
apply to scalars.
Also add a few predicates AMDGPU uses that should be generally useful.
Also add any() to complement all. I've wanted to use this a few times
but then worked around it not being there.
Confusingly, these were unrelated and had different semantics. The
G_PTR_MASK instruction predates the llvm.ptrmask intrinsic, but has a
different format. G_PTR_MASK only allows clearing the low bits of a
pointer, and only a constant number of bits. The ptrmask intrinsic
allows an arbitrary mask. Replace G_PTR_MASK to match the intrinsic.
Only selects the cases that look like the old instruction. More work
is needed to select the general case. Also new legalization code is
still needed to deal with the case where the incoming mask size does
not match the pointer size, which has a specified behavior in the
langref.
Unlike SelectionDAGBuilder, IRTranslator omits the unconditional
branch in fallthrough cases. Confusingly, the control flow pseudos
function in the opposite way the intrinsics are used, and the branch
targets always need to be swapped. We're inverting the target blocks,
so we need to figure out the old fallthrough block and insert a branch
to the original unconditional branch target.
Currently this code exists in widenScalar for G_MERGE_VALUE
sources. I'm not sure if the existing expansion in widenScalar should
be removed or not. The widenScalar variant tries to extend to the
requested size, but this just uses the original bitwidth.
We currently don't have a way to map to the equivalent intrinsic
opcode, so track immediate 0s in place of the address for the
selection to know to change the final opcode.
This reverts commit 9bca8fc4cf.
Rearrange handling to avoid changing the instruction in the case where
it's going to be erased and replaced with undef.
For normal loads, fully eliminate the load. For the TFE case, adjust
the dmask value in the instruction so the selector doesn't need to
handle it. For the TFE special case, I guess it would be possible to
replace the loaded data register with undef, but as-is this will start
treating it as a well defined value.
Trim elements that won't be written. The equivalent still needs to be
done for writes. Also start widening 3 elements to 4
elements. Selection will get the count from the dmask.
Instead, emit a trap and a warning. We force inlining of this
situation, so any function where this happens should be dead as
indirect or external calls are not yet supported. This should avoid
erroring on dead code.
G_SHUFFLE_VECTOR is legal since it theoretically may help match op_sel
for VOP3P instructions. Expand it in some other way in case it doesn't
fold into the use instructions.
There are few differences from the DAG handling. First, the DAG
handling uses a primitive selection pattern instead of custom
legalizing it. Because of this, this makes use of source modifiers
while the DAG does not.
Also instead of promoting f16, try to use the f16 log/exp. There's no
f16 fmul_legacy, so widen just for the multiply, although I'm not sure
that's the best solution.
AMDGPUCodeGenPrepare expands this most of the time, but not always. We
will always at least need a fallback option here. This is the 3rd
implementation of the same expansion in the backend. Eventually I
would like to eliminate the IR expansion (and the DAG version
obviously).
Currently the new legalizer path produces a better result, since the
IR expansion results in extra operations which need to be combined
out. Notably, the IR expansion results in multiplies by 0.
This is more or less directly ported from the AMDGPU custom lowering
for FP_TO_FP16. I made a few minor fixups (using G_UNMERGE_VALUES
instead of creating shift/trunc to extract the two halves, and zexting
an inverted compare instead of select_cc).
This also does not include the fast math expansion the DAG which
converts to f32 and then to f16. I think that belongs in a
pre-legalize combine instead.
Jay Foad