In foldSelectIntoOp we sometimes transform a select of a fadd into a
fadd of a select, where we select between data and an identity value.
For both fadd and fsub the identity is always -0.0, but if the nsz
flag is set on the select instruction we can use +0.0 instead. Doing
so then triggers other optimisations, such as when folding the select
of masked load into a new masked load.
Differential Revision: https://reviews.llvm.org/D126774
This patch improves the fix in D110529 to prevent from crashing on value
with byval attribute that is not added in SCCP solver.
Authored-by: sinan.lin@linux.alibaba.com
Reviewed By: ChuanqiXu
Differential Revision: https://reviews.llvm.org/D126355
We already have patterns for matching fadd(select(..., -0.0)),
but an upcoming patch will lead to patterns using +0.0 as the
identity instead of -0.0. I'm adding support for these patterns
now to avoid any regressions for MVE.
Differential Revision: https://reviews.llvm.org/D127275
This adds a fold for aggressive instcombine that converts
smin(smax(fptosi(x))) into a llvm.fptosi.sat, providing that the
saturation constants are correct and the cost of the llvm.fptosi.sat is
lower.
Unfortunately, a llvm.fptosi.sat cannot always be converted back to a
smin/smax/fptosi. The llvm.fptosi.sat intrinsic is more defined that the
original, which produces poison if the original fptosi was out of range.
The llvm.fptosi.sat will saturate any value, so needs to be expanded to
a fptosi(fpmin(fpmax(x))), which can be worse for codegeneration
depending on the target.
So this change thais conditional on the backend reporting that the
llvm.fptosi.sat is cheaper that the original smin+smax+fptost. This is
a change to the way that AggressiveInstrcombine has worked in the past.
Instead of just being a canonicalization pass, that canonicalization can
be dependant on the target in certain specific cases.
Differential Revision: https://reviews.llvm.org/D125755
Update the EarlyCSE tests to use opaque pointers.
Worth noting that this leaves some bitcast ptr to ptr instructions
in the input IR behind which are no longer necessary. This is
because these use numbered instructions, so it's hard to drop them
in an automated fashion (as it would require renumbering all other
instructions as well). I'm leaving that as a problem for another day.
The test updates have been performed using
https://gist.github.com/nikic/98357b71fd67756b0f064c9517b62a34.
Differential Revision: https://reviews.llvm.org/D127278
D125887 changed the ctlz/cttz despeculation transform to insert
a freeze for the introduced branch on zero. While this does fix
the "branch on poison" issue, we may still get in trouble if we
pick a different value for the branch and for the ctz argument
(i.e. non-zero for the branch, but zero for the ctz). To avoid
this, we should use the same frozen value in both positions.
This does cause a regression in RISCV codegen by introducing an
additional sext. The DAG looks like this:
t0: ch = EntryToken
t2: i64,ch = CopyFromReg t0, Register:i64 %3
t4: i64 = AssertSext t2, ValueType:ch:i32
t23: i64 = freeze t4
t9: ch = CopyToReg t0, Register:i64 %0, t23
t16: ch = CopyToReg t0, Register:i64 %4, Constant:i64<32>
t18: ch = TokenFactor t9, t16
t25: i64 = sign_extend_inreg t23, ValueType:ch:i32
t24: i64 = setcc t25, Constant:i64<0>, seteq:ch
t28: i64 = and t24, Constant:i64<1>
t19: ch = brcond t18, t28, BasicBlock:ch<cond.end 0x8311f68>
t21: ch = br t19, BasicBlock:ch<cond.false 0x8311e80>
I don't see a really obvious way to improve this, as we can't push
the freeze past the AssertSext (which may produce poison).
Differential Revision: https://reviews.llvm.org/D126638
Add new intrinsic and codegen support for the s_sendmsg_rtn_b32 and
s_sendmsg_rtn_b64 instructions.
Differential Revision: https://reviews.llvm.org/D127315
In GFX10 dlc controlled L1 cache bypass. In GFX11 it has been repurposed
to control MALL NOALLOC, and glc controls L1 as well as L0 cache bypass.
Update the documentation and SIMemoryLegalizer accordingly. Set dlc for
nontemporal and volatile accesses.
Differential Revision: https://reviews.llvm.org/D127405
The patch is a replacement of D125199. PseudoReadVL with vtype has worry for
computing same vtypes of VLEFF/VLSEGFF in two different places, DAGToDAG and
InsertVSETVLI. VLEFF/VLSEGFF MI with VL output still could provide the vtype of
VLEFF/VLSEGFF to the users of its VL.
The patch names the new pseudo as original VLEFF/VLSEGFF name suffixed "_VL" and
expand them in RISCVInsertVSETVLI pass.
This patch also reverts commit 4537aae0d57e17c217c192d8977012ba475b130c,
"[RISCV] Make PseudoReadVL have the vtypes of the corresponding VLEFF/VLSEGFF.".
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D126794
Implements Procedure Linkage Table (PLT) for ELF/AARCH64. The aarch64 linux calling convention also uses r16 as the intra-procedure-call scratch register same as MachO/ARM64. We can use the same stub sequence for this reason.
Also, BR regiseter doesn't touch X30 register. External function call by BL instruction (touched by CALL26 relocation) will set X30 to the original PC + 4, which is the intended behavior. External function call by B instruction (touched by JUMP26 relocation) doesn't requite to set X30, so the patch will be correct in this case too.
Reference: https://github.com/ARM-software/abi-aa/blob/main/aapcs64/aapcs64.rst#611general-purpose-registers
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D127061
Implements R_AARCH64_JUMP26. We can use the same generic aarch64 Branch26 edge since B instruction and BL nstruction have the same sized&offseted immediate field, and the relocation address calculation is the same.
Reference: ELF for the ARM ® 64-bit Architecture Tabel 4-10, ARM Architecture Reference Manual ® ARMv8, for ARMv8-A architecture profile C6.2.24, C6.2.31
Reviewed By: sgraenitz
Differential Revision: https://reviews.llvm.org/D127059
For an addition with simm14 and simm15 immediates with 2 or 3 trailing bits,
we can use a shXadd instruction and an addi to do the addition.
This patch teaches RISCVMergeBaseOffset to see through this pattern.
I don't think the sh1add case occurs because we use two addis for that,
but I implemented it for completeness.
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D127376
Teach the unroller(s) how to handle an invalid cost. This avoids crashes when the backend can't provide a cost due to either a fundemental limitation or an unimplemented cost model case.
Differential Revision: https://reviews.llvm.org/D127305
Per the documentation in Support/InstructionCost.h, the purpose of an invalid cost is so that clients can change behavior on impossible to cost inputs. CodeMetrics was instead asserting that invalid costs never occurred.
On a target with an incomplete cost model - e.g. RISCV - this means that transformations would crash on (falsely) invalid constructs - e.g. scalable vectors. While we certainly should improve the cost model - and I plan to do so in the near future - we also shouldn't be crashing. This violates the explicitly stated purpose of an invalid InstructionCost.
I updated all of the "easy" consumers where bailouts were locally obvious. I plan to follow up with loop unroll in a following change.
Differential Revision: https://reviews.llvm.org/D127131
https://alive2.llvm.org/ce/z/hRy3rE
As shown in D123408, we can produce this pattern when moving
cast around, and we already have a related fold for a binop
with a constant operand.
Changes for GFX11:
- Clauses may not mix instructions of different types, and there are
more types. For example image instructions with and without a sampler
are now different types.
- The max size of a clause is explicitly documented as 63 instructions.
Previously it was implicitly assumed to be 64. This is such a tiny
difference that it does not seem worth making it conditional on the
subtarget.
- It can be beneficial to clause stores as well as loads.
Differential Revision: https://reviews.llvm.org/D127391
Some object files produced by Mirosoft tools contain sections whose name field
is not fully null-padded at the end. Microsoft's dumpbin is able to print the
section name correctly, but this causes parsing errors with LLVM tools.
So far, this issue only seems to happen when the section name is longer than 8
bytes. In this case, the section name field contains a slash (/) followed by the
offset into the string table, but the name field is not fully null-padded at the
end.
Reviewed By: mstorsjo
Differential Revision: https://reviews.llvm.org/D127369
Supports encoding existing instrutions on gfx11 and MC support for the new VOPC
dpp instructions.
Patch 19/N for upstreaming of AMDGPU gfx11 architecture
Depends on D126978
Reviewed By: rampitec, #amdgpu
Differential Revision: https://reviews.llvm.org/D126989
UNIX03 conformance requires utilities to flush stdout before exiting and raise
an error if writing fails. Flushing already happens on a call to exit
and thus automatically on a return from main. Write failure is then
detected by LLVM's default SIGPIPE handler. The handler already exits with
a non-zero code, but conformance additionally requires an error message.
First reapply attempt I hadn't noticed the test had changed, hopefully this
goes better.
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
- VOP3 and SDWA forms of V_CMPX were not handled
- Hazard only exists if the compare defines EXEC (i.e. V_CMPX)
forwarded to the permlane.
Differential Revision: https://reviews.llvm.org/D127344
This was assuming the vector types were MVTs, but they don't have to be.
Note that the concrete output of the test isn't very useful, since it's
dominated by nonsensical calling convention lowering for the weird types.
Differential Revision: https://reviews.llvm.org/D126505
In order to make sure the stack point is right through the EH region,
we also need to restore stack pointer from the frame pointer if we
don't preserve stack space within prologue/epilogue for outgoing variables,
normally it's just checking the variable sized object is present or not
is enough, but we also don't preserve that at prologue/epilogue when
have vector objects in stack.
Example to show what happened:
```
try {
sp adjust for outgoing args. // 1. Sp changed.
func_call // 2. Exception raised
sp restore // Oh, not restored
} catch {
// 3. And now we are here.
}
// 4. Prepare to return!, restore return address from stack, but...sp is wrong.
// 5. Screw up!
```
Reviewed By: rogfer01
Differential Revision: https://reviews.llvm.org/D126861
For the longest time we used `AAValueSimplify` and
`genericValueTraversal` to determine "potential values". This was
problematic for many reasons:
- We recomputed the result a lot as there was no caching for the 9
locations calling `genericValueTraversal`.
- We added the idea of "intra" vs. "inter" procedural simplification
only as an afterthought. `genericValueTraversal` did offer an option
but `AAValueSimplify` did not. Thus, we might end up with "too much"
simplification in certain situations and then gave up on it.
- Because `genericValueTraversal` was not a real `AA` we ended up with
problems like the infinite recursion bug (#54981) as well as code
duplication.
This patch introduces `AAPotentialValues` and replaces the
`AAValueSimplify` uses with it. `genericValueTraversal` is folded into
`AAPotentialValues` as are the instruction simplifications performed in
`AAValueSimplify` before. We further distinguish "intra" and "inter"
procedural simplification now.
`AAValueSimplify` was not deleted as we haven't ported the
re-materialization of instructions yet. There are other differences over
the former handling, e.g., we may not fold trivially foldable
instructions right now, e.g., `add i32 1, 1` is not folded to `i32 2`
but if an operand would be simplified to `i32 1` we would fold it still.
We are also even more aware of function/SCC boundaries in CGSCC passes,
which is good.
Fixes: https://github.com/llvm/llvm-project/issues/54981
When determining liveness via Attributor::isAssumedDead(...) we might
end up without a liveness AA or with one pointing into another function.
Neither is helpful and we will avoid both from now on.
This reverts commit 1fbdbb559569641f6d509b569966901c8fb02b63.
All known issues surfaced by this patch should have been fixed now.
The fixes included fixing issues with SCEV expansion in LV and DA's
reliance on LCSSA phis.
This should fix a number of shuffle regressions in D127115 where the re-ordered combines mean we fail to fold a EXTRACT_VECTOR_ELT/INSERT_VECTOR_ELT sequence into a BUILD_VECTOR if we extract from more than one vector source.
The generic legalizer framework is still used to reduce the problem
to scalar multiplication with the bit size a multiple of 32.
Generating optimal code sequences for big integer multiplication is
somewhat tricky and has a number of target-specific intricacies:
- The target has V_MAD_U64_U32 instructions that multiply two 32-bit
factors and add a 64-bit accumulator. Most partial products should
use this instruction.
- The accumulator is mapped to consecutive 32-bit GPRs, and partial-
product multiply-adds can feed the accumulator into each other
directly. (The register allocator's support for that is somewhat
limited, but that only matters for 128-bit integers and larger.)
- OTOH, on some hardware, V_MAD_U64_U32 requires the accumulator
to be stored in an even-aligned pair of GPRs. To avoid excessive
register copies, it makes sense to compute odd partial products
separately from even partial products (where a partial product
src0[j0] * src1[j1] is "odd" if j0 + j1 is odd) and add both
halves together as a final step.
- We can combine G_MUL+G_ADD into a single cascade of multiply-adds.
- The target can keep many carry-bits in flight simultaneously, so
combining carries using G_UADDE is preferable over G_ZEXT + G_ADD.
- Not addressed by this patch: When the factors are sign-extended,
the V_MAD_I64_I32 instruction (signed version!) can be used.
It is difficult to address these points generically:
1) Finding matching pairs of G_MUL and G_UMULH to find a wide
multiply is expensive. We could add a G_UMUL_LOHI generic instruction
and conditionally use that in the generic legalizer, but by itself
this wouldn't allow us to use the accumulation capability of
V_MAD_U64_U32. One could attempt to find matching G_ADD + G_UADDE
post-legalization, but this is also expensive.
2) Similarly, making sense of the legalization outcome of a wide
pre-legalization G_MUL+G_ADD pair is extremely expensive.
3) How could the generic legalizer possibly deal with the
particular idiosyncracy of "odd" vs. "even" partial products.
All this points in the direction of directly emitting an ideal code
sequence during legalization, but the generic legalizer should not
be burdened with such overly target-specific concerns. Hence, a
custom legalization.
Note that the implemented approach is different from that used by
SelectionDAG because narrowing of scalars works differently in
general. SelectionDAG iteratively cuts wide scalars into low and
high halves until a legal size is reached. By contrast, GlobalISel
does the narrowing in a single shot, which should be better for
compile-time and for the quality of the generated code.
This patch leaves three gaps open:
1. When the factors are uniform, we should execute the multiplication on
the SALU. Register bank mapping already ensures this.
However, the resulting code sequence is not optimal because it doesn't
fully use the carry-in capabilities of S_ADDC_U32. (V_MAD_U64_U32
doesn't have a carry-in.) It is very difficult to fix this after the
fact, so we should really use a different legalization sequence in
this case. Unfortunately, we don't have a divergence analysis and so
cannot make that choice.
(This only matters for 128-bit integers and larger.)
2. Avoid unnecessary multiplies when sources are known to be zero- or
sign-extended. The challenge is that the legalizer does not currently
have access to GISelKnownBits.
3. When the G_MUL is followed by a G_ADD, we should consider combining
the two instructions into a single multiply-add sequence, to utilize
the accumulator of V_MAD_U64_U32 fully. (Unless the multiply has
multiple uses and the implied duplication of the multiply is an
overall negative). However, this is also not true when the factors
are uniform: in that case, it is generally better to *not* combine
the two operations, so that the multiply can be done on the SALU.
Again, we don't have a divergence analysis available and so cannot
make an informed choice.
Differential Revision: https://reviews.llvm.org/D124844
All information is already available in VPlan. Note that there are some
test changes, because we now can correctly look through instructions
like truncates to analyze the actual users.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D123541
We used to be very conservative when integer states were merged.
Instead of adding the known range (which is large due to uncertainty)
into the assumed range (which is hopefully small), we can also only
allow to merge in both at the same time into their respective
counterpart. This will ensure we keep the invariant that assumed is part
of known.
