| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
stack
This patch makes sure that if we tag some memory, we untag that memory before
the function returns/throws via any exit, reachable from the tag operation. For
that we place the untag operation either at:
a) the lifetime end call for the alloca, if that call post-dominates the
lifetime start call (where the tag operation is placed), or it (the
lifetime end call) dominates all reachable exits, otherwise
b) at the reachable exits
Differential Revision: https://reviews.llvm.org/D68469
llvm-svn: 374182
|
| |
|
|
| |
llvm-svn: 374165
|
| |
|
|
|
|
|
|
|
|
|
| |
The `expandLoadImmReal` handles four different and almost non-overlapping
cases: loading a "single" float immediate into a GPR, loading a "single"
float immediate into a FPR, and the same couple for a "double" float
immediate.
It's better to move each `else if` branch into separate methods.
llvm-svn: 374164
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A bpf specific clang intrinsic is introduced:
u32 __builtin_preserve_field_info(member_access, info_kind)
Depending on info_kind, different information will
be returned to the program. A relocation is also
recorded for this builtin so that bpf loader can
patch the instruction on the target host.
This clang intrinsic is used to get certain information
to facilitate struct/union member relocations.
The offset relocation is extended by 4 bytes to
include relocation kind.
Currently supported relocation kinds are
enum {
FIELD_BYTE_OFFSET = 0,
FIELD_BYTE_SIZE,
FIELD_EXISTENCE,
FIELD_SIGNEDNESS,
FIELD_LSHIFT_U64,
FIELD_RSHIFT_U64,
};
for __builtin_preserve_field_info. The old
access offset relocation is covered by
FIELD_BYTE_OFFSET = 0.
An example:
struct s {
int a;
int b1:9;
int b2:4;
};
enum {
FIELD_BYTE_OFFSET = 0,
FIELD_BYTE_SIZE,
FIELD_EXISTENCE,
FIELD_SIGNEDNESS,
FIELD_LSHIFT_U64,
FIELD_RSHIFT_U64,
};
void bpf_probe_read(void *, unsigned, const void *);
int field_read(struct s *arg) {
unsigned long long ull = 0;
unsigned offset = __builtin_preserve_field_info(arg->b2, FIELD_BYTE_OFFSET);
unsigned size = __builtin_preserve_field_info(arg->b2, FIELD_BYTE_SIZE);
#ifdef USE_PROBE_READ
bpf_probe_read(&ull, size, (const void *)arg + offset);
unsigned lshift = __builtin_preserve_field_info(arg->b2, FIELD_LSHIFT_U64);
#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
lshift = lshift + (size << 3) - 64;
#endif
#else
switch(size) {
case 1:
ull = *(unsigned char *)((void *)arg + offset); break;
case 2:
ull = *(unsigned short *)((void *)arg + offset); break;
case 4:
ull = *(unsigned int *)((void *)arg + offset); break;
case 8:
ull = *(unsigned long long *)((void *)arg + offset); break;
}
unsigned lshift = __builtin_preserve_field_info(arg->b2, FIELD_LSHIFT_U64);
#endif
ull <<= lshift;
if (__builtin_preserve_field_info(arg->b2, FIELD_SIGNEDNESS))
return (long long)ull >> __builtin_preserve_field_info(arg->b2, FIELD_RSHIFT_U64);
return ull >> __builtin_preserve_field_info(arg->b2, FIELD_RSHIFT_U64);
}
There is a minor overhead for bpf_probe_read() on big endian.
The code and relocation generated for field_read where bpf_probe_read() is
used to access argument data on little endian mode:
r3 = r1
r1 = 0
r1 = 4 <=== relocation (FIELD_BYTE_OFFSET)
r3 += r1
r1 = r10
r1 += -8
r2 = 4 <=== relocation (FIELD_BYTE_SIZE)
call bpf_probe_read
r2 = 51 <=== relocation (FIELD_LSHIFT_U64)
r1 = *(u64 *)(r10 - 8)
r1 <<= r2
r2 = 60 <=== relocation (FIELD_RSHIFT_U64)
r0 = r1
r0 >>= r2
r3 = 1 <=== relocation (FIELD_SIGNEDNESS)
if r3 == 0 goto LBB0_2
r1 s>>= r2
r0 = r1
LBB0_2:
exit
Compare to the above code between relocations FIELD_LSHIFT_U64 and
FIELD_LSHIFT_U64, the code with big endian mode has four more
instructions.
r1 = 41 <=== relocation (FIELD_LSHIFT_U64)
r6 += r1
r6 += -64
r6 <<= 32
r6 >>= 32
r1 = *(u64 *)(r10 - 8)
r1 <<= r6
r2 = 60 <=== relocation (FIELD_RSHIFT_U64)
The code and relocation generated when using direct load.
r2 = 0
r3 = 4
r4 = 4
if r4 s> 3 goto LBB0_3
if r4 == 1 goto LBB0_5
if r4 == 2 goto LBB0_6
goto LBB0_9
LBB0_6: # %sw.bb1
r1 += r3
r2 = *(u16 *)(r1 + 0)
goto LBB0_9
LBB0_3: # %entry
if r4 == 4 goto LBB0_7
if r4 == 8 goto LBB0_8
goto LBB0_9
LBB0_8: # %sw.bb9
r1 += r3
r2 = *(u64 *)(r1 + 0)
goto LBB0_9
LBB0_5: # %sw.bb
r1 += r3
r2 = *(u8 *)(r1 + 0)
goto LBB0_9
LBB0_7: # %sw.bb5
r1 += r3
r2 = *(u32 *)(r1 + 0)
LBB0_9: # %sw.epilog
r1 = 51
r2 <<= r1
r1 = 60
r0 = r2
r0 >>= r1
r3 = 1
if r3 == 0 goto LBB0_11
r2 s>>= r1
r0 = r2
LBB0_11: # %sw.epilog
exit
Considering verifier is able to do limited constant
propogation following branches. The following is the
code actually traversed.
r2 = 0
r3 = 4 <=== relocation
r4 = 4 <=== relocation
if r4 s> 3 goto LBB0_3
LBB0_3: # %entry
if r4 == 4 goto LBB0_7
LBB0_7: # %sw.bb5
r1 += r3
r2 = *(u32 *)(r1 + 0)
LBB0_9: # %sw.epilog
r1 = 51 <=== relocation
r2 <<= r1
r1 = 60 <=== relocation
r0 = r2
r0 >>= r1
r3 = 1
if r3 == 0 goto LBB0_11
r2 s>>= r1
r0 = r2
LBB0_11: # %sw.epilog
exit
For native load case, the load size is calculated to be the
same as the size of load width LLVM otherwise used to load
the value which is then used to extract the bitfield value.
Differential Revision: https://reviews.llvm.org/D67980
llvm-svn: 374099
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
There were 2 problems here. First, these patterns were duplicated to
handle the inverted shift operands instead of using the commuted
PatFrags.
Second, the point of the zext folding patterns don't apply to the
non-0ing high subtargets. They should be skipped instead of inserting
the extension. The zeroing high code would be emitted when necessary
anyway. This was also emitting unnecessary zexts in cases where the
high bits were undefined.
llvm-svn: 374092
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
separately in loop-vectorize"
Also Revert "[LoopVectorize] Fix non-debug builds after rL374017"
This reverts commit 9f41deccc0e648a006c9f38e11919f181b6c7e0a.
This reverts commit 18b6fe07bcf44294f200bd2b526cb737ed275c04.
The patch is breaking PowerPC internal build, checked with author, reverting
on behalf of him for now due to timezone.
llvm-svn: 374091
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary:
Without offsets on the MachineMemOperands (MMOs),
MachineInstr::mayAlias() will return true for all reads and writes to the
same resource descriptor. This leads to O(N^2) complexity in the MachineScheduler
when analyzing dependencies of buffer loads and stores. It also limits
the SILoadStoreOptimizer from merging more instructions.
This patch reduces the compile time of one pathological compute shader
from 12 seconds to 1 second.
Reviewers: arsenm, nhaehnle
Reviewed By: arsenm
Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, hiraditya, jfb, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D65097
llvm-svn: 374087
|
| |
|
|
|
|
|
|
|
|
|
| |
Inhibit generation of unused real dpp instructions on gfx10 just
like it is done on other subtargets. This does not change anything
because these are illegal anyway and not accepted, but it does
reduce the number of instruction definitions generated.
Differential Revision: https://reviews.llvm.org/D68607
llvm-svn: 374083
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary:
When searching for local expression tree created by stackified
registers, for 'block' placement, we start the search from the previous
instruction of a BB's terminator. But in 'try''s case, we should start
from the previous instruction of a call that can throw, or a EH_LABEL
that precedes the call, because the return values of the call's previous
instructions can be stackified and consumed by the throwing call.
For example,
```
i32.call @foo
call @bar ; may throw
br $label0
```
In this case, if we start the search from the previous instruction of
the terminator (`br` here), we end up stopping at `call @bar` and place
a 'try' between `i32.call @foo` and `call @bar`, because `call @bar`
does not have a return value so it is not a local expression tree of
`br`.
But in this case, unlike when placing 'block's, we should start the
search from `call @bar`, because the return value of `i32.call @foo` is
stackified and used by `call @bar`.
Reviewers: dschuff
Subscribers: sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68619
llvm-svn: 374073
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
During the If-Converter optimization pay attention when copying or
deleting call instructions in order to keep call site information in
valid state.
Reviewers: aprantl, vsk, efriedma
Reviewed By: vsk, efriedma
Differential Revision: https://reviews.llvm.org/D66955
llvm-svn: 374068
|
| |
|
|
|
|
|
|
|
|
|
| |
When -pg option is present than a call to _mcount is inserted into every
function. However since the proper ABI was not followed then the generated
gmon.out did not give proper results. By inserting needed instructions
before every _mcount we can fix this.
Differential Revision: https://reviews.llvm.org/D68390
llvm-svn: 374055
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Tim Northover remarked that the added patterns for fmls fp16
produce wrong code in case the fsub instruction has a
multiplication as its first operand, i.e., all the patterns FMLSv*_OP1:
> define <8 x half> @test_FMLSv8f16_OP1(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
> ; CHECK-LABEL: test_FMLSv8f16_OP1:
> ; CHECK: fmls {{v[0-9]+}}.8h, {{v[0-9]+}}.8h, {{v[0-9]+}}.8h
> entry:
>
> %mul = fmul fast <8 x half> %c, %b
> %sub = fsub fast <8 x half> %mul, %a
> ret <8 x half> %sub
> }
>
> This doesn't look right to me. The exact instruction produced is "fmls
> v0.8h, v2.8h, v1.8h", which I think calculates "v0 - v2*v1", but the
> IR is calculating "v2*v1-v0". The equivalent <4 x float> code also
> doesn't emit an fmls.
This patch generates an fmla and negates the value of the operand2 of the fsub.
Inspecting the pattern match, I found that there was another mistake in the
opcode to be selected: matching FMULv4*16 should generate FMLSv4*16
and not FMLSv2*32.
Tested on aarch64-linux with make check-all.
Differential Revision: https://reviews.llvm.org/D67990
llvm-svn: 374044
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Adds a TypeSize struct to represent the known minimum size of a type
along with a flag to indicate that the runtime size is a integer multiple
of that size
* Converts existing size query functions from Type.h and DataLayout.h to
return a TypeSize result
* Adds convenience methods (including a transparent conversion operator
to uint64_t) so that most existing code 'just works' as if the return
values were still scalars.
* Uses the new size queries along with ElementCount to ensure that all
supported instructions used with scalable vectors can be constructed
in IR.
Reviewers: hfinkel, lattner, rkruppe, greened, rovka, rengolin, sdesmalen
Reviewed By: rovka, sdesmalen
Differential Revision: https://reviews.llvm.org/D53137
llvm-svn: 374042
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary: Issue: https://github.com/GPUOpen-Drivers/llpc/issues/204
Reviewers: arsenm, rampitec
Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68184
llvm-svn: 374041
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Support for tracking registers that forward function parameters into the
following function frame. For now we only support cases when parameter
is forwarded through single register.
Reviewers: aprantl, vsk, t.p.northover
Reviewed By: vsk
Differential Revision: https://reviews.llvm.org/D66953
llvm-svn: 374033
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Based on the discussion in
http://lists.llvm.org/pipermail/llvm-dev/2019-October/135574.html, the
conclusion was reached that the ARM backend should produce vcmp instead
of vcmpe instructions by default, i.e. not be producing an Invalid
Operation exception when either arguments in a floating point compare
are quiet NaNs.
In the future, after constrained floating point intrinsics for floating
point compare have been introduced, vcmpe instructions probably should
be produced for those intrinsics - depending on the exact semantics
they'll be defined to have.
This patch logically consists of the following parts:
- Revert http://llvm.org/viewvc/llvm-project?rev=294945&view=rev and
http://llvm.org/viewvc/llvm-project?rev=294968&view=rev, which
implemented fine-tuning for when to produce vcmpe (i.e. not do it for
equality comparisons). The complexity introduced by those patches
isn't needed anymore if we just always produce vcmp instead. Maybe
these patches need to be reintroduced again once support is needed to
map potential LLVM-IR constrained floating point compare intrinsics to
the ARM instruction set.
- Simply select vcmp, instead of vcmpe, see simple changes in
lib/Target/ARM/ARMInstrVFP.td
- Adapt lots of tests that tested for vcmpe (instead of vcmp). For all
of these test, the intent of what is tested for isn't related to
whether the vcmp should produce an Invalid Operation exception or not.
Fixes PR43374.
Differential Revision: https://reviews.llvm.org/D68463
llvm-svn: 374025
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
in loop-vectorize
In loop-vectorize, interleave count and vector factor depend on target register number. Currently, it does not
estimate different register pressure for different register class separately(especially for scalar type,
float type should not be on the same position with int type), so it's not accurate. Specifically,
it causes too many times interleaving/unrolling, result in too many register spills in loop body and hurting performance.
So we need classify the register classes in IR level, and importantly these are abstract register classes,
and are not the target register class of backend provided in td file. It's used to establish the mapping between
the types of IR values and the number of simultaneous live ranges to which we'd like to limit for some set of those types.
For example, POWER target, register num is special when VSX is enabled. When VSX is enabled, the number of int scalar register is 32(GPR),
float is 64(VSR), but for int and float vector register both are 64(VSR). So there should be 2 kinds of register class when vsx is enabled,
and 3 kinds of register class when VSX is NOT enabled.
It runs on POWER target, it makes big(+~30%) performance improvement in one specific bmk(503.bwaves_r) of spec2017 and no other obvious degressions.
Differential revision: https://reviews.llvm.org/D67148
llvm-svn: 374017
|
| |
|
|
| |
llvm-svn: 373989
|
| |
|
|
|
|
|
|
|
|
| |
larger than i32.
Gather instructions can use i32 or i64 elements for indices. If
the index is zero extended from a type smaller than i32 to i64, we
can shrink the extend to just extend to i32.
llvm-svn: 373982
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When the target option GuaranteedTailCallOpt is specified, calls with
the fastcc calling convention will be transformed into tail calls if
they are in tail position. This diff adds a new calling convention,
tailcc, currently supported only on X86, which behaves the same way as
fastcc, except that the GuaranteedTailCallOpt flag does not need to
enabled in order to enable tail call optimization.
Patch by Dwight Guth <dwight.guth@runtimeverification.com>!
Reviewed By: lebedev.ri, paquette, rnk
Differential Revision: https://reviews.llvm.org/D67855
llvm-svn: 373976
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary:
There was a bug when computing the number of unwind destination
mismatches in CFGStackify. When there are many mismatched calls that
share the same (original) destination BB, they have to be counted
separately.
This also fixes a typo and runs `fixUnwindMismatches` only when the wasm
exception handling is enabled. This is to prevent unnecessary
computations and does not change behavior.
Reviewers: dschuff
Subscribers: sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68552
llvm-svn: 373975
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary:
Previously, `WebAssembly::mayThrow()` assumed all inputs are global
addresses. But when intrinsics, such as `memcpy`, `memmove`, or `memset`
are lowered to external symbols in instruction selection and later
emitted as library calls. And these functions don't throw.
This patch adds handling to those memory intrinsics to `mayThrow`
function. But while most of libcalls don't throw, we can't guarantee all
of them don't throw, so currently we conservatively return true for all
other external symbols.
I think a better way to solve this problem is to embed 'nounwind' info
in `TargetLowering::CallLoweringInfo`, so that we can access the info
from the backend. This will also enable transferring 'nounwind'
properties of LLVM IR instructions. Currently we don't transfer that
info and we can only access properties of callee functions, if the
callees are within the module. Other targets don't need this info in the
backend because they do all the processing before isel, but it will help
us because that info will reduce code size increase in fixing unwind
destination mismatches in CFGStackify.
But for now we return false for these memory intrinsics and true for all
other libcalls conservatively.
Reviewers: dschuff
Subscribers: sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68553
llvm-svn: 373967
|
| |
|
|
|
|
|
|
|
| |
Start manually writing a table to get the subreg index. TableGen
should probably generate this, but I'm not sure what it looks like in
the arbitrary case where subregisters are allowed to not fully cover
the super-registers.
llvm-svn: 373947
|
| |
|
|
| |
llvm-svn: 373946
|
| |
|
|
| |
llvm-svn: 373945
|
| |
|
|
| |
llvm-svn: 373944
|
| |
|
|
|
|
|
| |
This hides some defects in SIFoldOperands when the immediates are
split.
llvm-svn: 373943
|
| |
|
|
|
|
| |
Continue making a mess of merge/unmerge legality.
llvm-svn: 373942
|
| |
|
|
|
|
|
|
|
|
| |
At minimum handle the s64 insert type, which are emitted in real cases
during legalization.
We really need TableGen to emit something to emit something like the
inverse of composeSubRegIndices do determine the subreg index to use.
llvm-svn: 373938
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Allows targets to introduce regbankselectable
pseudo-instructions. Currently the closet feature to this is an
intrinsic. However this requires creating a public intrinsic
declaration. This litters the public intrinsic namespace with
operations we don't necessarily want to expose to IR producers, and
would rather leave as private to the backend.
Use a new instruction bit. A previous attempt tried to keep using enum
value ranges, but it turned into a mess.
llvm-svn: 373937
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Doing this makes MSVC complain that `empty(someRange)` could refer to
either C++17's std::empty or LLVM's llvm::empty, which previously we
avoided via SFINAE because std::empty is defined in terms of an empty
member rather than begin and end. So, switch callers over to the new
method as it is added.
https://reviews.llvm.org/D68439
llvm-svn: 373935
|
| |
|
|
|
|
|
|
| |
NFCI.
Stop all the callers from having to check the value type before calling getTargetShuffleInputs.
llvm-svn: 373915
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This ensures that frame-based unwinding will continue to work when
calling a noreturn function; there is not much use having the caller's
frame pointer saved if you don't also have the caller's program counter.
Patch by James Clarke.
Differential Revision: https://reviews.llvm.org/D68542
llvm-svn: 373907
|
| |
|
|
|
|
|
|
|
|
|
|
| |
J/JAL/JALX/JALS are absolute branches, but stay within the current
256 MB-aligned region, so we must include the high bits of the
instruction address when calculating the branch target.
Patch by James Clarke.
Differential Revision: https://reviews.llvm.org/D68548
llvm-svn: 373906
|
| |
|
|
|
|
| |
Fix comment.
llvm-svn: 373901
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
possible.
Move the erasing and iterator updating inside to match the
other slow LEA function.
I've adapted code from optTwoAddrLEA and basically rebuilt the
implementation here. We do lose the kill flags now just like
optTwoAddrLEA. This runs late enough in the pipeline that
shouldn't really be a problem.
llvm-svn: 373877
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
load (PR43217)
If a fp scalar is loaded and then used as both a scalar and a vector broadcast, perform the load as a broadcast and then extract the scalar for 'free' from the 0th element.
This involved switching the order of the X86ISD::BROADCAST combines so we only convert to X86ISD::BROADCAST_LOAD once all other canonicalizations have been attempted.
Adds a DAGCombinerInfo::recursivelyDeleteUnusedNodes wrapper.
Fixes PR43217
Differential Revision: https://reviews.llvm.org/D68544
llvm-svn: 373871
|
| |
|
|
| |
llvm-svn: 373870
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This is patch aims to group together the `CRNotPat` multi class instantiations
within the `PPCInstrInfo.td` file.
Integer instantiations of the multi class are grouped together into a section,
and the floating point patterns are separated into its own section.
Differential Revision: https://reviews.llvm.org/D67975
llvm-svn: 373869
|
| |
|
|
|
|
|
|
| |
directly.
Move the resolveTargetShuffleInputsAndMask call to after the shuffle mask combine before the undef/zero constant fold instead.
llvm-svn: 373868
|
| |
|
|
|
|
|
|
| |
Replaces setTargetShuffleZeroElements with getTargetShuffleAndZeroables which reports the Zeroable elements but doesn't merge them into the decoded target shuffle mask (the merging has been moved up into getTargetShuffleInputs until we can get rid of it entirely).
This is part of the work to fix PR43024 and allow us to use SimplifyDemandedElts to simplify shuffle chains - we need to get to a point where the target shuffle mask isn't adjusted by its source inputs but instead we cache them in a parallel Zeroable mask.
llvm-svn: 373867
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
v8i64->v8i8 truncate when v8i64 isn't legal
Summary:
The default legalization for v16i64->v16i8 tries to create a multiple stage truncate concatenating after each stage and truncating again. But avx512 implements truncates with multiple uops. So it should be better to truncate all the way to the desired element size and then concatenate the pieces using unpckl instructions. This minimizes the number of 2 uop truncates. The unpcks are all single uop instructions.
I tried to handle this by just custom splitting the v16i64->v16i8 shuffle. And hoped that the DAG combiner would leave the two halves in the state needed to make D68374 do the job for each half. This worked for the first half, but the second half got messed up. So I've implemented custom handling for v8i64->v8i8 when v8i64 needs to be split to produce the VTRUNCs directly.
Reviewers: RKSimon, spatel
Reviewed By: RKSimon
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68428
llvm-svn: 373864
|
| |
|
|
|
|
| |
of using setTargetShuffleZeroElements directly. NFCI.
llvm-svn: 373855
|
| |
|
|
|
|
|
|
| |
Summary: Replace 'isDarwin' with 'IsDarwin' based on LLVM naming convention.
Differential Revision: https://reviews.llvm.org/D68336
llvm-svn: 373852
|
| |
|
|
| |
llvm-svn: 373849
|
| |
|
|
|
|
|
|
|
|
| |
We can make use of the Zeroable mask to indicate which elements we can safely set to zero instead of creating a target shuffle mask on the fly.
This allows us to remove createTargetShuffleMask.
This is part of the work to fix PR43024 and allow us to use SimplifyDemandedElts to simplify shuffle chains - we need to get to a point where the target shuffle masks isn't adjusted by its source inputs in setTargetShuffleZeroElements but instead we cache them in a parallel Zeroable mask.
llvm-svn: 373846
|
| |
|
|
| |
llvm-svn: 373845
|
| |
|
|
| |
llvm-svn: 373842
|
| |
|
|
| |
llvm-svn: 373841
|
| |
|
|
| |
llvm-svn: 373840
|