I'm not sure what reasons the comment here could have
had for not setting these. Without these set, there is
an assertion hit during DWARF emission.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@243661 91177308-0d34-0410-b5e6-96231b3b80d8
Copy implementation of applyFixup from AArch64 with AArch64 bits
ripped out.
Tests will be included with a later commit. Several other
problems must be fixed before binary debug info emission
will work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@243660 91177308-0d34-0410-b5e6-96231b3b80d8
If the pointer is the store's value operand, this would produce
a broken module. Make sure the use is actually for the pointer operand.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@243462 91177308-0d34-0410-b5e6-96231b3b80d8
The pointer size of the addrspacecasted pointer might not have matched,
so this would have hit an assert in accumulateConstantOffset.
I think this was here to allow constant folding of a load of an
addrspacecasted constant. Accumulating the offset through the
addrspacecast doesn't make much sense, so something else is necessary
to allow folding the load through this cast.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@243300 91177308-0d34-0410-b5e6-96231b3b80d8
This can be done only with moves which theoretically
will optimize better later.
Although this transform increases the instruction count,
it should be code size / cycle count neutral in the worst
VALU case. It also seems to slightly improve a couple
of testcases due to other DAG combines this exposes.
This is probably slightly worse for the SALU case, so
it might be better to handle this during moveToVALU,
although then you lose some simplifications like
the load width reducing in the simple testcase.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@242177 91177308-0d34-0410-b5e6-96231b3b80d8
If the read2 produced was supposed to be writing into a
super register, it would use the wrong subregister indices.
Fix this by inserting copies, so we only ever write to a vreg_64.
Run the register coalescer again to clean this up, although this
isn't ideal and often does result in an extra move.
Also remove the assert that offset1 > offset0.
There isn't a real reason to not allow this other than a minor
convenience in the compiler, and it doesn't seem worth the effort
of avoiding it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@242174 91177308-0d34-0410-b5e6-96231b3b80d8
Most loads and stores are derived from pointers derived from
a kernel argument load inserted during argument lowering.
This was just using the EntryToken chain for the argument loads,
and any users of these loads were also on the EntryToken chain.
Return the chain of the lowered argument load so that dependent loads
end up on the correct chain.
No test since I'm not aware of any case where this actually
broke.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@241960 91177308-0d34-0410-b5e6-96231b3b80d8
The motivation is to allow GatherAllAliases / FindBetterChain
to not give up on dependent loads of a pointer from constant memory.
This is important for AMDGPU, because most loads are pointers
derived from a load of a kernel argument from constant memory.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@241948 91177308-0d34-0410-b5e6-96231b3b80d8
Running this after the scheduler enables scheduling
waits later so other ALU instructions can run while
this would be waiting.
When combined with enabling the post-RA scheduler, this
gives about a ~20% improvement on sgemm.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@241473 91177308-0d34-0410-b5e6-96231b3b80d8
We don't have a good way to detect most situations where
DS offsets are usable on SI, so add an option to force using
them even if unsafe for debugging performance problems.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@241462 91177308-0d34-0410-b5e6-96231b3b80d8
Reapply r239539. Don't assume the collected number of
stores is the same vector size. Just take the first N
stores to fill the vector.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@239825 91177308-0d34-0410-b5e6-96231b3b80d8
This only updates one of the uses. The other is used in cases
that may never touch memory, so I'm not sure why this is even
calling it. Perhaps there should be a new, similar hook for such
cases or pass -1 for unknown address space.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@239540 91177308-0d34-0410-b5e6-96231b3b80d8
Now actually stores the non-zero constant instead of 0.
I somehow forgot to include this part of r238108.
The test change was just an independent instruction order swap,
so just add another check line to satisfy CHECK-NEXT.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@239539 91177308-0d34-0410-b5e6-96231b3b80d8
No test since the kinds of transforms this prevents seem to not really
be relevant for SI's different addressing modes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@239261 91177308-0d34-0410-b5e6-96231b3b80d8
Now that we sometimes know the address space, this can
theoretically do a better job.
This needs better test coverage, but this mostly depends on
first updating the loop optimizatiosn to provide the address
space.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@239053 91177308-0d34-0410-b5e6-96231b3b80d8
Only set the triple from the command line options.
Some of these were still testing SI features and using the
old r600-- triple.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238958 91177308-0d34-0410-b5e6-96231b3b80d8
This is important because of different addressing modes
depending on the address space for GPU targets.
This only adds the argument, and does not update
any of the uses to provide the correct address space.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238723 91177308-0d34-0410-b5e6-96231b3b80d8
This is to fix problems introduced by r232481. For HSAIL,
this function does essentially nothing desirable, and
injects unwanted / incorrect stuff before the function.
The only thing it really needs to do is call EmitFunctionEntryLabel
in this case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238222 91177308-0d34-0410-b5e6-96231b3b80d8
This was resulting in the addrspacecast being removed and incorrectly
replaced with a ptrtoint when sinking.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238217 91177308-0d34-0410-b5e6-96231b3b80d8
On GPU targets, materializing constants is cheap and stores are
expensive, so only doing this for zero vectors was silly.
Most of the new testcases aren't optimally merged, and are for
later improvements.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238108 91177308-0d34-0410-b5e6-96231b3b80d8
If you somehow added a MachineOperand to an instruction
that did not have the parent set, the verifier would
crash since it attempts to use the operand's parent.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236249 91177308-0d34-0410-b5e6-96231b3b80d8
This should be fixed to properly understand all rematerializable
instructions while ignoring implicit reads of exec.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235671 91177308-0d34-0410-b5e6-96231b3b80d8
Currently symbol names are printed in quotes if it contains something
outside of the arbitrary set of characters that isAcceptableChar tests
for. On somem targets, it is never OK to print a symbol name in quotes
so allow targets to opt out of this behavior.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235670 91177308-0d34-0410-b5e6-96231b3b80d8
This enables very common cases to switch to the
smaller encoding.
All of the standard LLVM canonicalizations of comparisons
are the opposite of what we want. Compares with constants
are moved to the RHS, but the first operand can be an inline
immediate, literal constant, or SGPR using the 32-bit VOPC
encoding.
There are additional bad canonicalizations that should
also be fixed, such as canonicalizing ge x, k to gt x, (k + 1)
if this makes k no longer an inline immediate value.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@232988 91177308-0d34-0410-b5e6-96231b3b80d8
The expansion code does the same thing. Since
the operands were not defined with the correct
types, this has the side effect of fixing operand
folding since the expanded pseudo would never use
SGPRs or inline immediates.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@230072 91177308-0d34-0410-b5e6-96231b3b80d8
This enables a few useful combines that used to only
use fma.
Also since v_mad_f32 apparently does not support denormals,
disable the existing cases that are custom handled if they are
requested.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@230071 91177308-0d34-0410-b5e6-96231b3b80d8
This allows sharing of FMA forming combines to work
with instructions that have the same semantics as a separate
multiply and add.
This is expand by default, and only formed post legalization
so it shouldn't have much impact on targets that do not want it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@230070 91177308-0d34-0410-b5e6-96231b3b80d8
Some formats capitalized these, but most didn't. Change
them all to be consistently lowercase.
Now, non-encoding fields and convenience bits are capitalized.
Also remove weird looking empty line in some of the formats.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@229613 91177308-0d34-0410-b5e6-96231b3b80d8
src1 doesn't have modifiers, but the operand was missing
resulting in an encoding build error when all fields
are required.'
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@229611 91177308-0d34-0410-b5e6-96231b3b80d8
This should allow finally fixing the f64 fdiv implementation.
Test is disabled for VI since there seems to be a problem with one
of the buffer load instructions on it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@229236 91177308-0d34-0410-b5e6-96231b3b80d8
Matt Arsenault