Summary:
Multi-dword constant loads generated unnecessary moves from SGPRs into VGPRs,
increasing the code size and VGPR pressure. These moves are now folded away.
Note that this lack of operand folding was not a problem for VMEM loads,
because COPY nodes from VReg_Nnn to VGPR32 are eliminated by the register
coalescer.
Some tests are updated, note that the fsub.ll test explicitly checks that
the move is elided.
With the IR generated by current Mesa, the changes are obviously relatively
minor:
7063 shaders in 3531 tests
Totals:
SGPRS: 351872 -> 352560 (0.20 %)
VGPRS: 199984 -> 200732 (0.37 %)
Code Size: 9876968 -> 9881112 (0.04 %) bytes
LDS: 91 -> 91 (0.00 %) blocks
Scratch: 1779712 -> 1767424 (-0.69 %) bytes per wave
Wait states: 295164 -> 295337 (0.06 %)
Totals from affected shaders:
SGPRS: 65784 -> 66472 (1.05 %)
VGPRS: 38064 -> 38812 (1.97 %)
Code Size: 1993828 -> 1997972 (0.21 %) bytes
LDS: 42 -> 42 (0.00 %) blocks
Scratch: 795648 -> 783360 (-1.54 %) bytes per wave
Wait states: 54026 -> 54199 (0.32 %)
Reviewers: tstellarAMD, arsenm, mareko
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15875
llvm-svn: 257074
Summary:
Somehow, I first interpreted the docs as saying space for xnack_mask is only
reserved when XNACK is enabled via SH_MEM_CONFIG. I felt uneasy about this and
went back to actually test what is happening, and it turns out that xnack_mask
is always reserved at least on Tonga and Carrizo, in the sense that flat_scr
is always fixed below the SGPRs that are used to implement xnack_mask, whether
or not they are actually used.
I confirmed this by writing a shader using inline assembly to tease out the
aliasing between flat_scratch and regular SGPRs. For example, on Tonga, where
we fix the number of SGPRs to 80, s[74:75] aliases flat_scratch (so
xnack_mask is s[76:77] and vcc is s[78:79]).
This patch changes both the calculation of the total number of SGPRs and the
various register reservations to account for this.
It ought to be possible to use the gap left by xnack_mask when the feature
isn't used, but this patch doesn't try to do that. (Note that the same applies
to vcc.)
Note that previously, even before my earlier change in r256794, the SGPRs that
alias to xnack_mask could end up being used as well when flat_scr was unused
and the total number of SGPRs happened to fall on the right alignment
(e.g. highest regular SGPR being used s29 and VCC used would lead to number
of SGPRs being 32, where s28 and s29 alias with xnack_mask). So if there
were some conflict due to such aliasing, we should have noticed that already.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15898
llvm-svn: 257073
Summary:
This is admittedly something that you could only run into by manually
playing around with shader assembly because the SITypeWriter pass is
skipped for compute.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15902
llvm-svn: 256980
Due to the SGPR init bug, every program claims to use the same number
of SGPRs anyway, so there's no point in trying to shift those registers
down from their initial spot of reservation.
Add a test that uses VGPR spilling and blocks most SGPRs from being used for
the scratch resource register. Previously, this would run into an assertion.
Differential Revision: http://reviews.llvm.org/D15724
llvm-svn: 256870
Summary:
Enabling this feature will account for the two SGPRs used by the hardware
to store the XNACK_MASK physically.
The hardware only requires this reservation when the XNACK feature is
explicitly enabled. At some point, HSA will probably want to do that, but
it does increase SGPR register pressure, so leave it disabled by default
for now (but do add a small test).
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15869
llvm-svn: 256794
Summary:
For some reason doing executing an MUBUF instruction with the addr64
bit set and a zero base pointer in the resource descriptor causes
the memory operation to be dropped when the shader is executed using
the HSA runtime.
This kind of MUBUF instruction is commonly used when the pointer is
stored in VGPRs. The base pointer field in the resource descriptor
is set to zero and and the pointer is stored in the vaddr field.
This patch resolves the issue by only using flat instructions for
global memory operations when targeting HSA. This is an overly
conservative fix as all other configurations of MUBUF instructions
appear to work.
NOTE: re-commit by fixing a failure in Codegen/AMDGPU/llvm.dbg.value.ll
Reviewers: tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15543
llvm-svn: 256282
Summary:
For some reason doing executing an MUBUF instruction with the addr64
bit set and a zero base pointer in the resource descriptor causes
the memory operation to be dropped when the shader is executed using
the HSA runtime.
This kind of MUBUF instruction is commonly used when the pointer is
stored in VGPRs. The base pointer field in the resource descriptor
is set to zero and and the pointer is stored in the vaddr field.
This patch resolves the issue by only using flat instructions for
global memory operations when targeting HSA. This is an overly
conservative fix as all other configurations of MUBUF instructions
appear to work.
Reviewers: tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15543
llvm-svn: 256273
noduplicate prevents unrolling of small loops that happen to have
barriers in them. If a loop has a barrier in it, it is OK to duplicate
it for the unroll.
llvm-svn: 256075
Summary:
When copying aggregate registers within the same register class, there may
be an overlap between source and destination that forces us to do the copy
backwards.
Do the simplest possible thing that guarantees the correct order of moves
when there are overlaps, and does whatever when there is no overlap. (The
last part forces some trivial adjustments to test cases.)
Together with r255906, this fixes a VM fault in Unreal Elemental Demo.
While at it, change the generation of kill and def flags to something that
looks more reasonable. This method is used very late during compilation, so
it probably doesn't matter in practice, and to be honest, I don't know if
this change is actually correct because the semantics in connection with
aggregate registers vs. sub-registers are not clear to me.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93264
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15622
llvm-svn: 256072
Summary: I'm not sure how things worked before without this.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15492
llvm-svn: 255692
Summary:
We were previously selecting all constant loads to SMRD instructions and legalizing
the SMRDs with non-uniform addresses during the SIFixSGPRCopesPass.
This new solution is more simple and also generates much better code, because
the instruction selector is able to take advantage of all the MUBUF addressing
modes that are legalization pass wasn't able to.
We also no longer need to generate v_add_* instructions when we
have a uniform pointer and a non-uniform offset, as this is now folded into the
MUBUF instruction during instruction selection.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15425
llvm-svn: 255672
Summary:
These are meant to be used instead of the llvm.SI.tid intrinsic which will
be deprecated at some point.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15475
llvm-svn: 255652
Summary:
These are meant to be used instead of the llvm.SI.fs.interp intrinsic which
will be deprecated at some point.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15474
llvm-svn: 255651
Summary:
This allows us to remove the END_OF_TEXT_LABEL hack we had been using
and simplifies the fixups used to compute the address of constant
arrays.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15257
llvm-svn: 255204
Re-comitting with a change that avoids undefined uses getting put into
the VRegUses list.
The new algorithm remembers the uses encountered while walking backwards
until a matching def is found. Contrary to the previous version this:
- Works without LiveIntervals being available
- Allows to increase the precision to subregisters/lanemasks
(not used for now)
The changes in the AMDGPU tests are necessary because the R600 scheduler
is not stable with respect to the order of nodes in the ready queues.
Differential Revision: http://reviews.llvm.org/D9068
llvm-svn: 254683
Summary: This is done only when targeting HSA.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D13807
llvm-svn: 254587
This works mostly fine but breaks some stage 1 builders when compiling
compiler-rt on i386. Revert for further investigation as I can't see an
obvious cause/fix.
This reverts commit r254577.
llvm-svn: 254586
The new algorithm remembers the uses encountered while walking backwards
until a matching def is found. Contrary to the previous version this:
- Works without LiveIntervals being available
- Allows to increase the precision to subregisters/lanemasks
(not used for now)
The changes in the AMDGPU tests are necessary because the R600 scheduler
is not stable with respect to the order of nodes in the ready queues.
Differential Revision: http://reviews.llvm.org/D9068
llvm-svn: 254577
Summary: Only global or readonly segment variables should appear in object files.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15111
llvm-svn: 254519
If we know we have stack objects, we reserve the registers
that the private buffer resource and wave offset are passed
and use them directly.
If not, reserve the last 5 SGPRs just in case we need to spill.
After register allocation, try to pick the next available registers
instead of the last SGPRs, and then insert copies from the inputs
to the reserved registers in the progloue.
This also only selectively enables all of the input registers
which are really required instead of always enabling them.
llvm-svn: 254331
It does not work because of emergency stack slots.
This pass was supposed to eliminate dummy registers for the
spill instructions, but the register scavenger can introduce
more during PrologEpilogInserter, so some would end up
left behind if they were needed.
The potential for spilling the scratch resource descriptor
and offset register makes doing something like this
overly complicated. Reserve registers to use for the resource
descriptor and use them directly in eliminateFrameIndex.
Also removes creating another scratch resource descriptor
when directly selecting scratch MUBUF instructions.
The choice of which registers are reserved is temporary.
For now it attempts to pick the next available registers
after the user and system SGPRs.
llvm-svn: 254329
Summary:
This returns a pointer to the dispatch packet, which can be used to load
information about the kernel dispach.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D14898
llvm-svn: 254116
v2: added more tests, moved the SALU->VALU conversion to a separate function
It looks like it's not possible to get subregisters in the S_ABS lowering
code, and I don't feel like guessing without testing what the correct code
would look like.
llvm-svn: 254095
The one regression in the builtin tests is in the read2 test which now
(again) has many extra copies, but this should be solved once the pass
is replaced with a DAG combine.
llvm-svn: 253974
Note, this was reviewed (and more details are in) http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
These intrinsics currently have an explicit alignment argument which is
required to be a constant integer. It represents the alignment of the
source and dest, and so must be the minimum of those.
This change allows source and dest to each have their own alignments
by using the alignment attribute on their arguments. The alignment
argument itself is removed.
There are a few places in the code for which the code needs to be
checked by an expert as to whether using only src/dest alignment is
safe. For those places, they currently take the minimum of src/dest
alignments which matches the current behaviour.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 500, i32 8, i1 false)
will now read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %dest, i8* align 8 %src, i32 500, i1 false)
For out of tree owners, I was able to strip alignment from calls using sed by replacing:
(call.*llvm\.memset.*)i32\ [0-9]*\,\ i1 false\)
with:
$1i1 false)
and similarly for memmove and memcpy.
I then added back in alignment to test cases which needed it.
A similar commit will be made to clang which actually has many differences in alignment as now
IRBuilder can generate different source/dest alignments on calls.
In IRBuilder itself, a new argument was added. Instead of calling:
CreateMemCpy(Dst, Src, getInt64(Size), DstAlign, /* isVolatile */ false)
you now call
CreateMemCpy(Dst, Src, getInt64(Size), DstAlign, SrcAlign, /* isVolatile */ false)
There is a temporary class (IntegerAlignment) which takes the source alignment and rejects
implicit conversion from bool. This is to prevent isVolatile here from passing its default
parameter to the source alignment.
Note, changes in future can now be made to codegen. I didn't change anything here, but this
change should enable better memcpy code sequences.
Reviewed by Hal Finkel.
llvm-svn: 253511
Several places in AsmPrinter.cpp print comments describing MachineOperand
registers using MCRegisterInfo, which uses MCOperand-oriented names. This
doesn't work for targets that use virtual registers exclusively, as
WebAssembly does, since virtual registers are represented and printed
differently.
This patch preserves what seems to be the spirit of r229978, avoiding the
use of TM.getSubtargetImpl(), while still using MachineOperand-oriented
printing for MachineOperands.
Differential Revision: http://reviews.llvm.org/D14709
llvm-svn: 253338
This reverts commit r252565.
This also includes the revert of the commit mentioned below in order to
avoid breaking tests in AMDGPU:
Revert "AMDGPU: Set isAllocatable = 0 on VS_32/VS_64"
This reverts commit r252674.
llvm-svn: 252956
Mark kernels that use certain features that require user
SGPRs to support with kernel attributes. We need to know
before instruction selection begins because it impacts
the kernel calling convention lowering.
For now this only detects the workitem intrinsics.
llvm-svn: 252323
For some reason VS_32 ends up factoring into the pressure heuristics
even though we should never see a virtual register with this class.
When SGPRs are reserved for register spilling, this for some reason
triggers reg-crit scheduling.
Setting isAllocatable = 0 may help with this since that seems to remove
it from the default implementation's generated table.
llvm-svn: 252321
Previously, subprograms contained a metadata reference to the function they
described. Because most clients need to get or set a subprogram for a given
function rather than the other way around, this created unneeded inefficiency.
For example, many passes needed to call the function llvm::makeSubprogramMap()
to build a mapping from functions to subprograms, and the IR linker needed to
fix up function references in a way that caused quadratic complexity in the IR
linking phase of LTO.
This change reverses the direction of the edge by storing the subprogram as
function-level metadata and removing DISubprogram's function field.
Since this is an IR change, a bitcode upgrade has been provided.
Fixes PR23367. An upgrade script for textual IR for out-of-tree clients is
attached to the PR.
Differential Revision: http://reviews.llvm.org/D14265
llvm-svn: 252219
The operand layout is slightly different for the atomic
opcodes from the usual MUBUF loads and stores.
This should only fix it on SI/CI. VI is still broken
because it still emits the addr64 replacement.
llvm-svn: 252140
This was causing a variety of test failures when v2i64
is added as a legal type.
SIFixSGPRCopies should correctly handle the case of vector inputs
to a scalar reg_sequence, so this isn't necessary anymore. This
was hiding some deficiencies in how reg_sequence is handled later,
but this shouldn't be a problem anymore since the register class
copy of a reg_sequence is now done before the reg_sequence.
llvm-svn: 251860
I've found myself pointlessly debugging problems from running
graphics tests with an HSA triple a few times, so stop this from
happening again.
llvm-svn: 251858
There may be other use operands that also need their kill flags cleared.
This happens in a few tests when SIFoldOperands is moved after
PeepholeOptimizer.
PeepholeOptimizer rewrites cases that look like:
%vreg0 = ...
%vreg1 = COPY %vreg0
use %vreg1<kill>
%vreg2 = COPY %vreg0
use %vreg2<kill>
to use the earlier source to
%vreg0 = ...
use %vreg0
use %vreg0
Currently SIFoldOperands sees the copied registers, so there is
only one use. So far I haven't managed to come up with a test
that currently has multiple uses of a foldable VGPR -> VGPR copy.
llvm-svn: 250960
This wasn't doing anything useful. They weren't explicitly used
anywhere, and the RegScavenger ignores reserved registers.
This for some reason caused a random scheduling change in the test.
Getting the check lines to pass is too frustrating, and there's probably
not too much value in checking the vector case's operands N times.
llvm-svn: 250794
The comment says this was stopped because it was unlikely to be
profitable. This is not true if you want to combine vector loads
with multiple components.
For a simple case that looks like
t0 = load t0 ...
t1 = load t0 ...
t2 = load t0 ...
t3 = load t0 ...
t4 = store t0:1, t0:1
t5 = store t4, t1:0
t6 = store t5, t2:0
t7 = store t6, t3:0
We want to get all of these stores onto a chain
that is a TokenFactor of these N loads. This mostly
solves the AMDGPU merge-stores.ll regressions
with -combiner-alias-analysis for merging vector
stores of vector loads.
llvm-svn: 250138
This basic combine was surprisingly missing.
AMDGPU legalizes many operations in terms of 32-bit vector components,
so not doing this results in many extra copies and subregister extracts
that need to be cleaned up later.
InstCombine already does this for the hasOneUse case. The target hook
is to fix a handful of tests which break (e.g. ARM/vmov.ll) which turn
from a vector materialize repeated immediate instruction to a constant
vector load with more scalar copies from it.
llvm-svn: 250129
This stops using an unknown reg class operand.
Currently build_vector selection has a broken looking check
where it tries to use a VGPR reg class and an SGPR one if it
sees an SGPR use.
With the source operand has an explicit VGPR class,
illegal copies will be inserted that SIFixSGPRCopies will take care
of normally later, which will allow removing the weird check
of build_vector users. Without this, when removed v_movrels_b32 would
still be emitted even though all of the values were only stored in
SGPRs.
llvm-svn: 249494
Summary:
We currently ignore the calling convention, so there is no real reason to
assert on the calling convention of functions.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D13367
llvm-svn: 249468
Summary:
Instead of asserting when the kernel metadata is different than we expect,
we should just skip lowering that function. This fixes assertion
failures with OpenCL argument metadata from older LLVM releases.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D13356
llvm-svn: 249073
v2: Add test (Matt).
Fix capitalization of isEOP (Matt).
Move pattern to class parameter (Matt).
Make the instruction available to Cayman (Matt).
Change name from MEM_RAT WRITE_TYPED to MEM_RAT STORE_TYPED.
Patch by: Zoltan Gilian
llvm-svn: 249042
Trying to use the version with the explicit output operand
would complain because of the missing WriteSALU. I'm not sure
why it doesn't complain about this with the implicit VCC def.
llvm-svn: 248646
If a virtual register is copied and another copy was already
seen, replace with the previous copy. This only handles the
simplest cases for now.
This pattern shows up from various operand restrictions
AMDGPU has which require inserting copies depending
on the register class of the operands.
llvm-svn: 248611
This fixes a select error when the i64 source was also
bitcasted to v2i32 in the original source.
Instead of awkwardly trying to select the modified source value and
the store, replace before isel begins.
Uses a worklist to avoid possible problems from mutating the DAG,
although it seems to work OK without it.
llvm-svn: 248589
These were all using the default 32-bit VALU write class,
but the i64/f64 compares are half rate.
I'm not sure this is really correct, because they are still using
the write to VALU write class, even though they really write
to the SALU.
llvm-svn: 248582
These are necessary for implementing mem_fence for
OpenCL 2.0.
The VI assembler tests are disabled since it seems to be
using the wrong encoding or opcode.
llvm-svn: 248532
Allow a target to do something other than search for copies
that will avoid cross register bank copies.
Implement for SI by only rewriting the most basic copies,
so it should look through anything like a subregister extract.
I'm not entirely satisified with this because it seems like
eliminating a reg_sequence that isn't fully used should work
generically for all targets without them having to override
something. However, it seems to be tricky to have a simple
implementation of this without rewriting to invalid kinds
of subregister copies on some targets.
I'm not sure if there is currently a generic way to easily check
if a subregister index would be valid for the current use.
The current set of TargetRegisterInfo::get*Class functions don't
quite behave like I would expect (e.g. getSubClassWithSubReg
returns the maximal register class rather than the minimal), so
I'm not sure how to make the generic test keep searching if
SrcRC:SrcSubReg is a valid replacement for DefRC:DefSubReg. Making
the default implementation to check for simple copies breaks
a variety of ARM and x86 tests by producing illegal subregister uses.
The ARM tests are not actually changed since it should still be using
the same sharesSameRegisterFile implementation, this just relaxes
them to not check for specific registers.
llvm-svn: 248478
Instead of always inserting a copy in case
the super register is itself a subregister,
only extract to the super reg class if this is
actually the case.
This shouldn't really change codegen, but
makes looking at the output of SIFixSGPRCopies
easier to read.
llvm-svn: 248467
We may have subregister defs which are unused but not discovered and
cleaned up prior to liveness analysis. This creates multiple connected
components in the resulting live range which are forbidden in the
MachineVerifier because they would unnecesarily constrain the register
allocator. Rewrite those dead definitions to define a newly created
virtual register.
Differential Revision: http://reviews.llvm.org/D13035
llvm-svn: 248335
This patch adds support for combining patterns such as (FMUL(FADD(1.0, x), y)) and (FMUL(FSUB(x, 1.0), y)) to their FMA equivalents.
This is useful in particular for linear interpolation cases such as (FADD(FMUL(x, t), FMUL(y, FSUB(1.0, t))))
Differential Revision: http://reviews.llvm.org/D13003
llvm-svn: 248210
If storing multiple FP constants, some subset of the stores
would be replaced with integers due to visit order, so
MergeConsecutiveStores would only partially merge
these.
llvm-svn: 248169
Summary:
This helps mostly when we use add instructions for address calculations
that contain immediates.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D12256
llvm-svn: 247157
Currently this hits an assert that extload should
always be supported, which assumes integer extloads.
This moves a hack out of SI's argument lowering and
is covered by existing tests.
llvm-svn: 247113
sub C, x - > add (sub 0, x), C for DS offsets.
This is mostly to fix regressions that show up when
SeparateConstOffsetFromGEP is enabled.
llvm-svn: 247054
The VOP3 encoding of these allows any SGPR pair for the i1
output, but this was forced before to always use vcc.
This doesn't yet try to use this, but does add the operand
to the definitions so the main change is adding vcc to the
output of the VOP2 encoding.
llvm-svn: 246358
Without a memory operand, mayLoad or mayStore instructions
are treated as hasUnorderedMemRef, which results in much worse
scheduling.
We really should have a verifier check that any
non-side effecting mayLoad or mayStore has a memory operand.
There are a few instructions (interp and images) which I'm
not sure what / where to add these.
llvm-svn: 246356
As a follow-up to r246098, require `DISubprogram` definitions
(`isDefinition: true`) to be 'distinct'. Specifically, add an assembler
check, a verifier check, and bitcode upgrading logic to combat testcase
bitrot after the `DIBuilder` change.
While working on the testcases, I realized that
test/Linker/subprogram-linkonce-weak-odr.ll isn't relevant anymore. Its
purpose was to check for a corner case in PR22792 where two subprogram
definitions match exactly and share the same metadata node. The new
verifier check, requiring that subprogram definitions are 'distinct',
precludes that possibility.
I updated almost all the IR with the following script:
git grep -l -E -e '= !DISubprogram\(.* isDefinition: true' |
grep -v test/Bitcode |
xargs sed -i '' -e 's/= \(!DISubprogram(.*, isDefinition: true\)/= distinct \1/'
Likely some variant of would work for out-of-tree testcases.
llvm-svn: 246327
We can wait on either VM, EXP or LGKM.
The waits are independent.
Without this patch, a wait inserted because of one of them
would also wait for all the previous others.
This patch makes s_wait only wait for the ones we need for the next
instruction.
Here's an example of subtle perf reduction this patch solves:
This is without the patch:
buffer_load_format_xyzw v[8:11], v0, s[44:47], 0 idxen
buffer_load_format_xyzw v[12:15], v0, s[48:51], 0 idxen
s_load_dwordx4 s[44:47], s[8:9], 0xc
s_waitcnt lgkmcnt(0)
buffer_load_format_xyzw v[16:19], v0, s[52:55], 0 idxen
s_load_dwordx4 s[48:51], s[8:9], 0x10
s_waitcnt vmcnt(1)
buffer_load_format_xyzw v[20:23], v0, s[44:47], 0 idxen
The s_waitcnt vmcnt(1) is useless.
The reason it is added is because the last
buffer_load_format_xyzw needs s[44:47], which was issued
by the first s_load_dwordx4. It waits for all VM
before that call to have finished.
Internally after every instruction, 3 counters (for VM, EXP and LGTM)
are updated after every instruction. For example buffer_load_format_xyzw
will
increase the VM counter, and s_load_dwordx4 the LGKM one.
Without the patch, for every defined register,
the current 3 counters are stored, and are used to know
how long to wait when an instruction needs the register.
Because of that, the s[44:47] counter includes that to use the register
you need to wait for the previous buffer_load_format_xyzw.
Instead this patch stores only the counters that matter for the
register,
and puts zero for the other ones, since we don't need any wait for them.
Patch by: Axel Davy
Differential Revision: http://reviews.llvm.org/D11883
llvm-svn: 245755
The pass adds new kernel arguments for image attributes, and
resolves calls to dummy attribute and resource id getter functions.
Patch by: Zoltan Gilian
llvm-svn: 244372
Summary: This allows us to consolidate several of the TableGen patterns.
Reviewers: arsenm
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11602
llvm-svn: 244253
This commit fixes a bug in the class 'SIInstrInfo' where the implicit register
machine operands were added to a machine instruction in an incorrect order -
the implicit uses were added before the implicit defs.
I found this bug while working on moving the implicit register operand
verification code from the MIR parser to the machine verifier.
This commit also makes the method 'addImplicitDefUseOperands' in the machine
instruction class public so that it can be reused in the 'SIInstrInfo' class.
Reviewers: Matt Arsenault
Differential Revision: http://reviews.llvm.org/D11689
llvm-svn: 243799
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.
llvm-svn: 243462
Author: Dave Airlie <airlied@redhat.com>
In order to implement indirect sampler loads, we don't
want to match on a VGPR load but an SGPR one for constants,
as we cannot feed VGPRs to the sampler only SGPRs.
this should be applicable for llvm 3.7 as well.
llvm-svn: 243294
Summary:
The MUBUF addr64 bit has been removed on VI, so we must use FLAT
instructions when the pointer is stored in VGPRs.
Reviewers: arsenm
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D11067
llvm-svn: 242673
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.
llvm-svn: 242177
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.
llvm-svn: 242174
The two-address instruction pass will convert these back to v_mad_f32
if necessary.
Differential Revision: http://reviews.llvm.org/D11060
llvm-svn: 242038
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.
llvm-svn: 241948
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.
llvm-svn: 241462
A call to removeEmptySubranges() is necessary after every operation that
potentially removes all segments from a subregister range; this case in
the register coalescer was missing.
llvm-svn: 241027
Summary: We need to set MTYPE = 2 for VI shaders when targeting the HSA runtime.
Reviewers: arsenm
Differential Revision: http://reviews.llvm.org/D10777
llvm-svn: 240841
Summary:
This way the function symbol points to the start of amd_kernel_code_t
rather than the start of the function.
Reviewers: arsenm
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10705
llvm-svn: 240829
If pseudoToMCOpcode failed, we would return the original opcode, so operands
would be swapped, but the instruction would remain the same.
It resulted in LSHLREV a, b ---> LSHLREV b, a.
This fixes Glamor text rendering and
piglit/arb_sample_shading-builtin-gl-sample-mask on VI.
This is a candidate for stable branches.
v2: the test was simplified by Tom Stellard
llvm-svn: 240824
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.
llvm-svn: 239825
