I'm not sure if this should be considered a bug in
copyImplicitOps or not, but implicit operands that are part
of the static instruction definition should not be copied.
llvm-svn: 280594
Summary:
This contains two changes that reduce the time spent in WQM, with the
intention of reducing bandwidth required by VMEM loads:
1. Sampling instructions by themselves don't need to run in WQM, only their
coordinate inputs need it (unless of course there is a dependent sampling
instruction). The initial scanInstructions step is modified accordingly.
2. When switching back from WQM to Exact, switch back as soon as possible.
This affects the logic in processBlock.
This should always be a win or at best neutral.
There are also some cleanups (e.g. remove unused ExecExports) and some new
debugging output.
Reviewers: arsenm, tstellarAMD, mareko
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: http://reviews.llvm.org/D22092
llvm-svn: 280590
Summary:
This fixes a rare bug in polygon stippling with non-monolithic pixel shaders.
The underlying problem is as follows: the prolog part contains the polygon
stippling sequence, i.e. a kill. The main part then enables WQM based on the
_reduced_ exec mask, effectively undoing most of the polygon stippling.
Since we cannot know whether polygon stippling will be used, the main part
of a non-monolithic shader must always return to exact mode to fix this
problem.
Reviewers: arsenm, tstellarAMD, mareko
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D23131
llvm-svn: 280589
readlane/writelane do not support using m0 as the output/input.
Constrain the register class of spill vregs to try to avoid this,
but also handle spilling of the physreg when necessary by inserting
an additional copy to a normal SGPR.
llvm-svn: 280584
Add runtime metdata for pointee alignment of pointer type kernel argument. The key is KeyArgPointeeAlign and the value is a 32 bit unsigned integer.
Differential Revision: https://reviews.llvm.org/D24145
llvm-svn: 280399
Summary:
Created a new td file MIMGInstructions.td which contains all definitions
of MIMG related instructions.
Reviewed by:
kzhuravl, vpykhtin
Differential Revision:
http://reviews.llvm.org/D24106
llvm-svn: 280385
Summary: This fixes some OpenCV tests that were broken by libclc commit r276443.
Reviewers: arsenm, jvesely
Subscribers: arsenm, wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D24051
llvm-svn: 280274
Summary:
Simply replace usage of aliases to functions with aliasee.
This came up when bitcode linking to builtin library and
calls to aliases not being resolved.
Also made minor improvements to existing test.
Reviewers: tstellarAMD, alex-t, vpykhtin
Subscribers: arsenm, wdng, rampitec
Differential Revision: https://reviews.llvm.org/D24023
llvm-svn: 280221
Move SDLoc initialization to comon place.
fall back to AMDGPU version in one place
Differential Revision: https://reviews.llvm.org/D23900
llvm-svn: 280030
Summary:
The SILoadStoreOptimizer can now look ahead more then one instruction when
looking for instructions to merge, which greatly improves the number of
loads/stores that we are able to merge.
Moving the pass before scheduling avoids increasing register pressure after
the scheduler, so that the scheduler's register pressure estimates will be
more accurate. It also gives more consistent results, since it is no longer
affected by minor scheduling changes.
Reviewers: arsenm
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: https://reviews.llvm.org/D23814
llvm-svn: 279991
Summary:
For shrinking SOPK instructions, we were creating a hint to tell the
register allocator to use the register allocated for src0 for the dst
operand as well. However, this seems to not work sometimes depending
on the order virtual registers are assigned physical registers.
To fix this, I've added a second allocation hint which does the reverse,
asks that the register allocated for dst is used for src0.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D23862
llvm-svn: 279968
Summary:
The SILoadStoreOptimizer will need to use AliasAnalysis here in order to
move it before scheduling.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D23813
llvm-svn: 279963
There's only one use of this for the convenience
of a pattern. I think v_mov_b64_pseudo should also be
moved, but SIFoldOperands does currently make use of it.
llvm-svn: 279901
It isn't used for anything, and is also misleading since
it could be spilled at the end of the block, so it can't be relied
on. There ends up being a verifier error about using an undefined
register since the spill kills the register.
llvm-svn: 279899
Summary:
If the scheduler clusters the loads, then the offsets will be sorted,
but it is possible for the scheduler to scheduler loads together
without out explicitly clustering them, which would give us non-sorted
offsets.
Also, we will want to do this if we move the load/store optimizer before
the scheduler.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D23776
llvm-svn: 279870
Summary:
There are a few different sgpr pressure sets, but we only care about
the one which covers all of the sgprs. We were using hard-coded
register pressure set names to determine the reg set id for the
biggest sgpr set. However, we were using the wrong name, and this
method is pretty fragile, since the reg pressure set names may
change.
The new method just looks for the pressure set that contains the most
reg units and sets that set as our SGPR pressure set. We've also
adopted the same technique for determining our VGPR pressure set.
Reviewers: arsenm
Subscribers: MatzeB, arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D23687
llvm-svn: 279867
MCContext already has many tasks, and separating CodeView out from it is
probably a good idea. The .cv_loc tracking was modelled on the DWARF
tracking which lived directly in MCContext.
Removes the inclusion of MCCodeView.h from MCContext.h, so now there are
only 10 build actions while I hack on CodeView support instead of 265.
llvm-svn: 279847
Summary:
This patch implements readlane/readfirstlane intrinsics.
TODO: need to define a new register class to consider the case
that the source could be a vector register or M0.
Reviewed by:
arsenm and tstellarAMD
Differential Revision:
http://reviews.llvm.org/D22489
llvm-svn: 279660
Re-apply this patch, hopefully I will get away without any warnings
in the constructor now.
This patch removes the MachineFunctionAnalysis. Instead we keep a
map from IR Function to MachineFunction in the MachineModuleInfo.
This allows the insertion of ModulePasses into the codegen pipeline
without breaking it because the MachineFunctionAnalysis gets dropped
before a module pass.
Peak memory should stay unchanged without a ModulePass in the codegen
pipeline: Previously the MachineFunction was freed at the end of a codegen
function pipeline because the MachineFunctionAnalysis was dropped; With
this patch the MachineFunction is freed after the AsmPrinter has
finished.
Differential Revision: http://reviews.llvm.org/D23736
llvm-svn: 279602
I want to compute the SSA property of .mir files automatically in
upcoming patches. The problem with this is that some inputs will be
reported as static single assignment with some passes claiming not to
support SSA form. In reality though those passes do not support PHI
instructions => Track the presence of PHI instructions separate from the
SSA property.
Differential Revision: https://reviews.llvm.org/D22719
llvm-svn: 279573
Re-apply this commit with the deletion of a MachineFunction delegated to
a separate pass to avoid use after free when doing this directly in
AsmPrinter.
This patch removes the MachineFunctionAnalysis. Instead we keep a
map from IR Function to MachineFunction in the MachineModuleInfo.
This allows the insertion of ModulePasses into the codegen pipeline
without breaking it because the MachineFunctionAnalysis gets dropped
before a module pass.
Peak memory should stay unchanged without a ModulePass in the codegen
pipeline: Previously the MachineFunction was freed at the end of a codegen
function pipeline because the MachineFunctionAnalysis was dropped; With
this patch the MachineFunction is freed after the AsmPrinter has
finished.
Differential Revision: http://reviews.llvm.org/D23736
llvm-svn: 279564
This patch removes the MachineFunctionAnalysis. Instead we keep a
map from IR Function to MachineFunction in the MachineModuleInfo.
This allows the insertion of ModulePasses into the codegen pipeline
without breaking it because the MachineFunctionAnalysis gets dropped
before a module pass.
Peak memory should stay unchanged without a ModulePass in the codegen
pipeline: Previously the MachineFunction was freed at the end of a codegen
function pipeline because the MachineFunctionAnalysis was dropped; With
this patch the MachineFunction is freed after the AsmPrinter has
finished.
Differential Revision: http://reviews.llvm.org/D23736
llvm-svn: 279502
Do most of the lowering in a pre-RA pass. Keep the skip jump
insertion late, plus a few other things that require more
work to move out.
One concern I have is now there may be COPY instructions
which do not have the necessary implicit exec uses
if they will be lowered to v_mov_b32.
This has a positive effect on SGPR usage in shader-db.
llvm-svn: 279464
Currently nodes_iterator may dereference to a NodeType* or a NodeType&. Make them all dereference to NodeType*, which is NodeRef later.
Differential Revision: https://reviews.llvm.org/D23704
Differential Revision: https://reviews.llvm.org/D23705
llvm-svn: 279326
The names of the tablegen defs now match the names of the ISD nodes.
This makes the world a slightly saner place, as previously "fround" matched
ISD::FP_ROUND and not ISD::FROUND.
Differential Revision: https://reviews.llvm.org/D23597
llvm-svn: 279129
This is a mechanical change of comments in switches like fallthrough,
fall-through, or fall-thru to use the LLVM_FALLTHROUGH macro instead.
llvm-svn: 278902
minimal and boring form than the old pass manager's version.
This pass does the very minimal amount of work necessary to inline
functions declared as always-inline. It doesn't support a wide array of
things that the legacy pass manager did support, but is alse ... about
20 lines of code. So it has that going for it. Notably things this
doesn't support:
- Array alloca merging
- To support the above, bottom-up inlining with careful history
tracking and call graph updates
- DCE of the functions that become dead after this inlining.
- Inlining through call instructions with the always_inline attribute.
Instead, it focuses on inlining functions with that attribute.
The first I've omitted because I'm hoping to just turn it off for the
primary pass manager. If that doesn't pan out, I can add it here but it
will be reasonably expensive to do so.
The second should really be handled by running global-dce after the
inliner. I don't want to re-implement the non-trivial logic necessary to
do comdat-correct DCE of functions. This means the -O0 pipeline will
have to be at least 'always-inline,global-dce', but that seems
reasonable to me. If others are seriously worried about this I'd like to
hear about it and understand why. Again, this is all solveable by
factoring that logic into a utility and calling it here, but I'd like to
wait to do that until there is a clear reason why the existing
pass-based factoring won't work.
The final point is a serious one. I can fairly easily add support for
this, but it seems both costly and a confusing construct for the use
case of the always inliner running at -O0. This attribute can of course
still impact the normal inliner easily (although I find that
a questionable re-use of the same attribute). I've started a discussion
to sort out what semantics we want here and based on that can figure out
if it makes sense ta have this complexity at O0 or not.
One other advantage of this design is that it should be quite a bit
faster due to checking for whether the function is a viable candidate
for inlining exactly once per function instead of doing it for each call
site.
Anyways, hopefully a reasonable starting point for this pass.
Differential Revision: https://reviews.llvm.org/D23299
llvm-svn: 278896
The structs ImmOp and RegOp are in AArch64AsmParser.cpp (inside
anonymous namespace).
This diff changes the order of fields and removes the excessive padding
(8 bytes).
Patch by Alexander Shaposhnikov
llvm-svn: 278844
Recall that MSVC always gives enums the type 'int', nothing else. MSVC
2015 does not appear to have this problem anymore.
Clang-cl -Wmicrosoft-enum-value flags this, FWIW, so now I have a true
positive for my warning. :)
llvm-svn: 278762
Remove all ilist_iterator to pointer casts. There were two reasons for
casts:
- Checking for an uninitialized (i.e., null) iterator. I added
MachineInstrBundleIterator::isValid() to check for that case.
- Comparing an iterator against the underlying pointer value while
avoiding converting the pointer value to an iterator. This is
occasionally necessary in MachineInstrBundleIterator, since there is
an assertion in the constructors that the underlying MachineInstr is
not bundled (but we don't care about that if we're just checking for
pointer equality).
To support the latter case, I rewrote the == and != operators for
ilist_iterator and MachineInstrBundleIterator.
- The implicit constructors now use enable_if to exclude
const-iterator => non-const-iterator conversions from overload
resolution (previously it was a compiler error on instantiation, now
it's SFINAE).
- The == and != operators are now global (friends), and are not
templated.
- MachineInstrBundleIterator has overloads to compare against both
const_pointer and const_reference. This avoids the implicit
conversions to MachineInstrBundleIterator that assert, instead just
checking the address (and I added unit tests to confirm this).
Notably, the only remaining uses of ilist_iterator::getNodePtrUnchecked
are in ilist.h, and no code outside of ilist*.h directly relies on this
UB end-iterator-to-pointer conversion anymore. It's still needed for
ilist_*sentinel_traits, but I'll clean that up soon.
llvm-svn: 278478
Summary:
This patch define and implement amdgcn image intrinsics with sampler.
1. define vdata type to be llvm_anyfloat_ty, address type to be llvm_anyfloat_ty,
and rsrc type to be llvm_anyint_ty. As a result, we expect the intrinsics name
to have three suffixes to overload each of these three types;
2. D128 as well as two other flags are implied in the three types, for example,
if you use v8i32 as resource type, then r128 is 0!
3. don't expose TFE flag, and other flags are exposed in the instruction order:
unrm, glc, slc, lwe and da.
Differential Revision: http://reviews.llvm.org/D22838
Reviewed by:
arsenm and tstellarAMD
llvm-svn: 278291
Insert before the skip branch if one is created.
This is a somewhat more natural placement relative
to the skip branches, and makes it possible to implement
analyzeBranch for skip blocks.
The test changes are mostly due to a quirk where
the block label is not emitted if there is a terminator
that is not also a branch.
llvm-svn: 278273
Summary:
This is the setting of the Vulkan closed source driver.
It decreases the max wave count from 10 to 8.
26010 shaders in 14650 tests
Totals:
VGPRS: 829593 -> 808440 (-2.55 %)
Spilled SGPRs: 81878 -> 42226 (-48.43 %)
Spilled VGPRs: 367 -> 358 (-2.45 %)
Scratch VGPRs: 1764 -> 1748 (-0.91 %) dwords per thread
Code Size: 36677864 -> 35923932 (-2.06 %) bytes
There is a massive decrease in SGPR spilling in general and -7.4% spilled
VGPRs for DiRT Showdown (= SGPRs spilled to scratch?)
Reviewers: arsenm, tstellarAMD, nhaehnle
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D23034
llvm-svn: 277867
Adding missing tests for OCL type names for half, float, double, char, short, long, and unknown.
Patch by Aaron En Ye Shi.
Differential Revision: https://reviews.llvm.org/D22964
llvm-svn: 277759
Summary:
TargetBaseAlign is no longer required since LSV checks if target allows misaligned accesses.
A constant defining a base alignment is still needed for stack accesses where alignment can be adjusted.
Previous patch (D22936) was reverted because tests were failing. This patch also fixes the cause of those failures:
- x86 failing tests either did not have the right target, or the right alignment.
- NVPTX failing tests did not have the right alignment.
- AMDGPU failing test (merge-stores) should allow vectorization with the given alignment but the target info
considers <3xi32> a non-standard type and gives up early. This patch removes the condition and only checks
for a maximum size allowed and relies on the next condition checking for %4 for correctness.
This should be revisited to include 3xi32 as a MVT type (on arsenm's non-immediate todo list).
Note that checking the sizeInBits for a MVT is undefined (leads to an assertion failure),
so we need to create an EVT, hence the interface change in allowsMisaligned to include the Context.
Reviewers: arsenm, jlebar, tstellarAMD
Subscribers: jholewinski, arsenm, mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D23068
llvm-svn: 277735
On modern Intel processors hardware SQRT in many cases is faster than RSQRT
followed by Newton-Raphson refinement. The patch introduces a simple heuristic
to choose between hardware SQRT instruction and Newton-Raphson software
estimation.
The patch treats scalars and vectors differently. The heuristic is that for
scalars the compiler should optimize for latency while for vectors it should
optimize for throughput. It is based on the assumption that throughput bound
code is likely to be vectorized.
Basically, the patch disables scalar NR for big cores and disables NR completely
for Skylake. Firstly, scalar SQRT has shorter latency than NR code in big cores.
Secondly, vector SQRT has been greatly improved in Skylake and has better
throughput compared to NR.
Differential Revision: https://reviews.llvm.org/D21379
llvm-svn: 277725
Summary:
Two types of stores are possible in pixel shaders: stores to memory that are
explicitly requested at the API level, and stores that are an implementation
detail of register spilling or lowering of arrays.
For the first kind of store, we must ensure that helper pixels have no effect
and hence WQM must be disabled. The second kind of store must always be
executed, because the written value may be loaded again in a way that is
relevant for helper pixels as well -- and there are no externally visible
effects anyway.
This is a candidate for the 3.9 release branch.
Reviewers: arsenm, tstellarAMD, mareko
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: https://reviews.llvm.org/D22675
llvm-svn: 277504
Summary:
There are cases where uniform branch conditions are computed in VGPRs, and
we didn't correctly mark those as WQM.
The stray change in basic-branch.ll is because invoking the LiveIntervals
analysis leads to the detection of a dead register that would otherwise not
be seen at -O0.
This is a candidate for the 3.9 branch, as it fixes a possible hang.
Reviewers: arsenm, tstellarAMD, mareko
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D22673
llvm-svn: 277500
This adds a target hook getInstSizeInBytes to TargetInstrInfo that a lot of
subclasses already implement.
Differential Revision: https://reviews.llvm.org/D22885
llvm-svn: 277126
Summary:
We were using reserved VGPRs for SGPR spilling and this was causing
some programs with a workgroup size of 1024 to use more than 64
registers, which is illegal.
Reviewers: arsenm, mareko, nhaehnle
Subscribers: nhaehnle, arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D22032
llvm-svn: 276980
Summary:
SI_ELSE is lowered into two parts:
s_or_saveexec_b64 dst, src (at the start of the basic block)
s_xor_b64 exec, exec, dst (at the end of the basic block)
The idea is that dst contains the exec mask of the preceding IF block. It can
happen that SIWholeQuadMode decides to switch from WQM to Exact mode inside
the basic block that contains SI_ELSE, in which case it introduces an instruction
s_and_b64 exec, exec, s[...]
which masks out bits that can correspond to both the IF and the ELSE paths.
So the resulting sequence must be:
s_or_savexec_b64 dst, src
s_and_b64 exec, exec, s[...] <-- added by SIWholeQuadMode
s_and_b64 dst, dst, exec <-- added by SILowerControlFlow
s_xor_b64 exec, exec, dst
Whether to add the additional s_and_b64 dst, dst, exec is currently determined
via the ExecModified tracking. With this change, it is instead determined by
an additional flag on SI_ELSE which is set by SIWholeQuadMode.
Finally: It also occured to me that an alternative approach for the long run
is for SILowerControlFlow to unconditionally emit
s_or_saveexec_b64 dst, src
...
s_and_b64 dst, dst, exec
s_xor_b64 exec, exec, dst
and have a pass that detects and cleans up the "redundant AND with exec"
pattern where possible. This could be useful anyway, because we also add
instructions
s_and_b64 vcc, exec, vcc
before s_cbranch_scc (in moveToALU), and those are often redundant. I have
some pending changes to how KILL is lowered that could also benefit from
such a cleanup pass.
In any case, this current patch could help in the short term with the whole
ExecModified business.
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D22846
llvm-svn: 276972
TargetOptions wants the ExceptionHandling enum. Move that to
MCTargetOptions.h to avoid transitively including Dwarf.h everywhere in
clang. Now you can add a DWARF tag without a full rebuild of clang
semantic analysis.
llvm-svn: 276883
ABIArgOffset is a problem because properly fsetting the
KernArgSize requires that the reserved area before the
real kernel arguments be correctly aligned, which requires
fixing clover.
llvm-svn: 276766
Some targets, notably AArch64 for ILP32, have different relocation encodings
based upon the ABI. This is an enabling change, so a future patch can use the
ABIName from MCTargetOptions to chose which relocations to use. Tested using
check-llvm.
The corresponding change to clang is in: http://reviews.llvm.org/D16538
Patch by: Joel Jones
Differential Revision: https://reviews.llvm.org/D16213
llvm-svn: 276654
This reverts commit r276298.
Data stored in .rodata can have a negative offset from .text, but we
don't support negative values in relocations yet.
This caused a regression in one of the amp conformance tests:
5_Data_Cont/5_2_a_v/5_2_3_m/Assignment/Test.02.01
llvm-svn: 276498
This adds the actual MachineLegalizeHelper to do the work and a trivial pass
wrapper that legalizes all instructions in a MachineFunction. Currently the
only transformation supported is splitting up a vector G_ADD into one acting on
smaller vectors.
llvm-svn: 276461
The size can exceed s_movk_i32's limit, and we don't
want to use it this early since it inhibits optimizations.
This should probably be merged to the release branch.
llvm-svn: 276438
If 2.5 ulp is acceptable, denormals are not required, and
isn't a reciprocal which will already be handled, replace
with a faster fdiv.
Simplify the lowering tests by using per function
subtarget features.
llvm-svn: 276051
Only if the value is negative or positive is what matters,
so use a constant that doesn't require an instruction to
materialize.
These should really just emit the write exec directly,
but for stick with the kill pseudo-terminator.
llvm-svn: 275988
This is to help moveSILowerControlFlow to before regalloc.
There are a couple of tradeoffs with this. The complete CFG
is visible to more passes, the loop body avoids an extra copy of m0,
vcc isn't required, and immediate offsets can be shrunk into s_movk_i32.
The disadvantage is the register allocator doesn't understand that
the single lane's vector is dead within the loop body, so an extra
register is used to outlive the loop block when expanding the
VGPR -> m0 loop. This also now results in worse waitcnt insertion
before the loop instead of after for pending operations at the point
of the indexing, but that should be fixed by future improvements to
cross block waitcnt insertion.
v_movreld_b32's operands are now modeled more correctly since vdst
is not a true output. This is kind of a hack to treat vdst as a
use operand. Extra checking is required in the verifier since
I can't seem to get tablegen to emit an implicit operand for a
virtual register.
llvm-svn: 275934
This is currently only called with GEP users. A direct
alloca would only happen with current typed pointers
for arrays which are a perverse case.
Also fix crashes on 0 x and 1 x arrays.
llvm-svn: 275869
Summary:
The work item intrinsics are not available for the shader
calling conventions. And even if we did hook them up most
shader stages haves some extra restrictions on the amount
of available LDS.
Reviewers: tstellarAMD, arsenm
Subscribers: nhaehnle, arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D20728
llvm-svn: 275779
In this situation:
%VGPR2<def> = BUFFER_LOAD_DWORD_OFFSET %SGPR8_SGPR9_SGPR10_SGPR11,
%VGPR7<def,tied3> = V_MAC_F32_e32 %VGPR0<undef>, %VGPR1<kill>, %VGPR7<kill,tied0>, %EXEC<imp-use>
%VGPR3_VGPR4_VGPR5_VGPR6<def> = COPY %VGPR0_VGPR1_VGPR2_VGPR3
%VGPR4<def> = COPY %VGPR2
The copy for VGPR1 -> VGPR4 was an error from reading undefined VGPR1,
but VGPR4 is defined immediately after this copy.
llvm-svn: 275635
Summary:
Instead, we take a single flags arg (a bitset).
Also add a default 0 alignment, and change the order of arguments so the
alignment comes before the flags.
This greatly simplifies many callsites, and fixes a bug in
AMDGPUISelLowering, wherein the order of the args to getLoad was
inverted. It also greatly simplifies the process of adding another flag
to getLoad.
Reviewers: chandlerc, tstellarAMD
Subscribers: jholewinski, arsenm, jyknight, dsanders, nemanjai, llvm-commits
Differential Revision: http://reviews.llvm.org/D22249
llvm-svn: 275592
Added emitting metadata to elf for runtime.
Runtime requires certain information (metadata) about kernels to be able to execute and query them. Such information is emitted to an elf section as a key-value pair stream.
Differential Revision: https://reviews.llvm.org/D21849
llvm-svn: 275566
Also stop trying to insert skip blocks at end_cf. This
was inserting them at the end of the block which doesn't make
sense. The skip should be inserted at the beginning of the block
right after the end cf. Just remove this for now since no tests
seem to stress this and I think this can be handled more generally
later.
Fixes bug 28550
llvm-svn: 275510
Summary:
v2: don't count SGPRs spilled to scratch twice
I think this is sufficient. It doesn't count private memory usage, which
happens often and uses scratch but isn't technically a spill. The private
memory usage can be computed by:
[scratch_per_thread - vgpr_spills - a random multiple of SGPR spills].
The fact SGPR spills add very high numbers to the scratch size make that
computation a guessing game, but I don't have a solution to that.
Reviewers: tstellarAMD
Subscribers: arsenm, kzhuravl
Differential Revision: http://reviews.llvm.org/D22197
llvm-svn: 275288
- Add new TTI instruction checks
- Don't use const for blocks that are mutated.
- Checking isBranch and isTerminator should be redundant
llvm-svn: 275252
Summary:
Previously, constant index insertelements would be turned into SI_INDIRECT_DST,
which is bound to prevent some optimization opportunities. Worse, it mislead
the heuristic that decides whether immediates should be lowered to S_MOV_B32
or V_MOV_B32 in a way that resulted in unnecessary v_readfirstlanes.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: http://reviews.llvm.org/D22217
llvm-svn: 275160
Summary:
Setting MIMG to 0 has a bunch of unexpected side effects, including that
isVMEM returns false which leads to incorrect treatment in the hazard
recognizer. The reason I noticed it is that it also leads to incorrect
treatment in VGPR-to-SGPR copies, which is one cause of the referenced bug.
The only reason why MIMG was set to 0 is to signal the special handling of
dmasks, but that can be checked differently.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96877
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: http://reviews.llvm.org/D22210
llvm-svn: 275113
Summary:
The main bug fix here is using the 32-bit encoding of V_ADD_I32 in
materializeFrameBaseRegister and resolveFrameIndex, so that arbitrary
immediates work.
The second part is that we may now require the SegmentWaveByteOffset
even when there are initially no stack objects and VGPR spilling isn't
enabled, for stack slots that are allocated later. This means that some
bits become effectively dead and can be cleaned up.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96602
Tested-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: http://reviews.llvm.org/D21551
llvm-svn: 275108
Remove remaining implicit conversions from MachineInstrBundleIterator to
MachineInstr* from the AMDGPU backend. In most cases, I made them less
attractive by preferring MachineInstr& or using a ranged-based for loop.
Once all the backends are fixed I'll make the operator explicit so that
this doesn't bitrot back.
llvm-svn: 274906
Change a while loop that was checking for nullptr on an
iterator-to-pointer conversion to an infinite for loop. Now it's clear
that the condition doesn't terminate.
The only change in behaviour is if an invalid iterator (holding nullptr)
was passed into AMDGPUCFGStructurizer::reversePredicateSetter. There
are only two callers, and they both dereference the iterator before
sending it in, so rather than adding an early return to avoid the loop
I've just asserted (using a static_cast, to avoid an implicit conversion
to pointer).
llvm-svn: 274902
The commit message is inaccurate, modifiesRegister
will check for partial defs of exec.
We currently don't ever emit partial defs of exec,
so it doesn't really matter.
llvm-svn: 274886
The cost model should not assume vector casts get completely scalarized, since
on targets that have vector support, the common case is a partial split up to
the legal vector size. So, when a vector cast gets split, the resulting casts
end up legal and cheap.
Instead of pessimistically assuming scalarization, base TTI can use the costs
the concrete TTI provides for the split vector, plus a fudge factor to account
for the cost of the split itself. This fudge factor is currently 1 by default,
except on AMDGPU where inserts and extracts are considered free.
Differential Revision: http://reviews.llvm.org/D21251
llvm-svn: 274642
Summary:
Since "AMDGPU: Fix verifier errors in SILowerControlFlow", the logic that
ensures that a non-void-returning shader falls off the end of the last
basic block was effectively disabled, since SI_RETURN is now used.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96731
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: http://reviews.llvm.org/D21975
llvm-svn: 274612
Registers are printed a lot, so don't create temporary
std::strings. Using char instead of a string to an ostream
saves a function call.
llvm-svn: 274581
Because of the special immediate operand, the constant
bus is already used so SGPRs are never useful.
r263212 changed the name of the immediate operand, which
broke the verifier check for the restriction.
llvm-svn: 274564
Summary:
These have been replaced with TableGen code (except for isConstantLoad,
which is still used for R600). The queries were broken for cases
where MemOperand was a PseudoSourceValue.
Reviewers: arsenm
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: http://reviews.llvm.org/D21684
llvm-svn: 274561
Summary:
The isGlobalLoad() query was returning true for constant address space loads
with memory types less than 32-bits, which is wrong. This logic has been
replaced with PatFrag in the TableGen files, to provide the same functionality.
Reviewers: arsenm
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: http://reviews.llvm.org/D21696
llvm-svn: 274521
Due to visit order problems, in the case of an unaligned copy
the legalized DAG fails to eliminate extra instructions introduced
by the expansion of both unaligned parts.
llvm-svn: 274397
There was a combine before to handle the simple copy case.
Split this into handling loads and stores separately.
We might want to change how this handles some of the vector
extloads, since this can result in large code size increases.
llvm-svn: 274394
Enable testing different scheduling variants if sgpr usage
is very high. It was previously disabled because of a bug
in handleMove, but it has been fixed since.
Patch by Axel Davy
llvm-svn: 274372
This was reverted in r268740 because of problems with corresponding Clang change.
Clang change was updated and resubmitted in r274220.
Check calling convention in AMDGPUMachineFunction::isKernel
This will be used for AMDGPU_HSA_KERNEL symbol type in output ELF.
Also, in the future unused non-kernels may be optimized.
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm, joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D19917
llvm-svn: 274341
Summary: dst_sel and dst_unused disabled for VOPC as they have no effect on result
Reviewers: artem.tamazov, tstellarAMD, vpykhtin
Subscribers: arsenm, kzhuravl
Differential Revision: http://reviews.llvm.org/D21376
llvm-svn: 274340
TargetSubtargetInfo::overrideSchedPolicy takes two MachineInstr*
arguments (begin and end) that invite implicit conversions from
MachineInstrBundleIterator. One option would be to change their type to
an iterator, but since they don't seem to have been used since the API
was added in 2010, I'm deleting the dead code.
llvm-svn: 274304
This is a mechanical change to make TargetLowering API take MachineInstr&
(instead of MachineInstr*), since the argument is expected to be a valid
MachineInstr. In one case, changed a parameter from MachineInstr* to
MachineBasicBlock::iterator, since it was used as an insertion point.
As a side effect, this removes a bunch of MachineInstr* to
MachineBasicBlock::iterator implicit conversions, a necessary step
toward fixing PR26753.
llvm-svn: 274287
This is mostly a mechanical change to make TargetInstrInfo API take
MachineInstr& (instead of MachineInstr* or MachineBasicBlock::iterator)
when the argument is expected to be a valid MachineInstr. This is a
general API improvement.
Although it would be possible to do this one function at a time, that
would demand a quadratic amount of churn since many of these functions
call each other. Instead I've done everything as a block and just
updated what was necessary.
This is mostly mechanical fixes: adding and removing `*` and `&`
operators. The only non-mechanical change is to split
ARMBaseInstrInfo::getOperandLatencyImpl out from
ARMBaseInstrInfo::getOperandLatency. Previously, the latter took a
`MachineInstr*` which it updated to the instruction bundle leader; now,
the latter calls the former either with the same `MachineInstr&` or the
bundle leader.
As a side effect, this removes a bunch of MachineInstr* to
MachineBasicBlock::iterator implicit conversions, a necessary step
toward fixing PR26753.
Note: I updated WebAssembly, Lanai, and AVR (despite being
off-by-default) since it turned out to be easy. I couldn't run tests
for AVR since llc doesn't link with it turned on.
llvm-svn: 274189
Don't cast GV expression to MCSymbolRefExpr. r272705 changed GV to binary
expressions by including offset even if the offset it 0
(we haven't hit this sooner since tested workloads don't include static offsets)
We don't really care about the type of expression, so set it directly.
Fixes: r272705
Consider section relative relocations. Since all const as data is in one boffer section relative is equivalent to abs32.
Fixes: r273166
Differential Revision: http://reviews.llvm.org/D21633
llvm-svn: 273785
Debugger prologue is emitted if -mattr=+amdgpu-debugger-emit-prologue.
Debugger prologue writes work group IDs and work item IDs to scratch memory at fixed location in the following format:
- offset 0: work group ID x
- offset 4: work group ID y
- offset 8: work group ID z
- offset 16: work item ID x
- offset 20: work item ID y
- offset 24: work item ID z
Set
- amd_kernel_code_t::debug_wavefront_private_segment_offset_sgpr to scratch wave offset reg
- amd_kernel_code_t::debug_private_segment_buffer_sgpr to scratch rsrc reg
- amd_kernel_code_t::is_debug_supported to true if all debugger features are enabled
Differential Revision: http://reviews.llvm.org/D20335
llvm-svn: 273769
Summary:
Offset folding only works if you are emitting relocations, and we don't
emit relocations for local address space globals.
Reviewers: arsenm, nhaustov
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: http://reviews.llvm.org/D21647
llvm-svn: 273765
COPY was lacking a scheduling class, define it to avoid regressions in
the upcoming change to the bidirectional MachineScheduler. Approved by
tstellar on IRC.
Differential Revision: http://reviews.llvm.org/D21540
llvm-svn: 273751
The only real reason to use it is for testing, so replace
it with a command line option instead of a potentially function
dependent feature.
llvm-svn: 273653
Split AMDGPUSubtarget into amdgcn/r600 specific subclasses.
This removes most of the static_casting of the basic codegen
classes everywhere, and tries to restrict the features
visible on the wrong target.
llvm-svn: 273652
Summary:
We will start generating this in a future patch.
Reviewers: arsenm, kzhuravl, rafael, ruiu, tony-tye
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: http://reviews.llvm.org/D21482
llvm-svn: 273628
The exit-on-error flag was necessary in order to avoid an assertion when
handling DYNAMIC_STACKALLOC nodes in SelectionDAGLegalize.
We can avoid the assertion by creating some dummy nodes. This enables us to
remove the exit-on-error flag on the first 2 run lines (SI), but on the third
run line (R600) we would run into another assertion when trying to reserve
indirect registers. This patch also replaces that assertion with an early exit
from the function.
Fixes PR27761.
Differential Revision: http://reviews.llvm.org/D20852
llvm-svn: 273550
The main sin this was committing was using terminator
instructions in the middle of the block, and then
not updating the block successors / predecessors.
Split the blocks up to avoid this and introduce new
pseudo instructions for branches taken with exec masking.
Also use a pseudo instead of emitting s_endpgm and erasing
it in the special case of a non-void return.
llvm-svn: 273467
The implicit operand is added by the initial instruction construction,
so this was adding an additional vcc use. The original one
was missing the undef flag the original condition had,
so the verifier would complain.
llvm-svn: 273182
This will help sneak undefs past GVN into the DAG for
some tests.
Also add missing intrinsic for rsq_legacy, even though the node
was already selected to the instruction. Also start passing
the debug location to intrinsic errors.
llvm-svn: 273181
Don't use AllocateStack because kernel arguments have nothing
to do with the stack. The ensureMaxAlignment call was still
changing the stack alignment.
llvm-svn: 273080
This should select to s_trap, but that requires
additonal work to setup and enable the trap handler.
For now emit s_endpgm so bugpoint stops getting stuck
on the unsupported call to abort.
Emit a warning that this will only terminate the wave and
not really trap.
llvm-svn: 273062
Summary:
This fixes two related bugs. First, the generic optimization passes
unfortunately generate negative constant offsets but the hardware treats
SOffset as an unsigned value.
Second, there is a hardware bug on SI and CI, where address clamping in MUBUF
instructions does not work correctly when SOffset is larger than the buffer
size. This patch works around this bug by never using SOffset.
An alternative workaround would be to do the clamping manually when SOffset
is too large, but generating the required code sequence during instruction
selection would be rather involved, and in any case the resulting code would
probably be worse.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96360
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: http://reviews.llvm.org/D21326
llvm-svn: 272761
Summary:
We we have an MCConstantExpr, we can encode it directly into the instruction
instead of emitting fixups.
Reviewers: artem.tamazov, vpykhtin, SamWot, nhaustov, arsenm
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: http://reviews.llvm.org/D21236
Change-Id: I88b3edf288d48e65c5d705fc4850d281f8e36948
llvm-svn: 272750
Summary:
We can now reference symbols directly in operands, like this:
s_mov_b32 s0, global
Reviewers: artem.tamazov, vpykhtin, SamWot, nhaustov
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: http://reviews.llvm.org/D21038
llvm-svn: 272748
If a local_unnamed_addr attribute is attached to a global, the address
is known to be insignificant within the module. It is distinct from the
existing unnamed_addr attribute in that it only describes a local property
of the module rather than a global property of the symbol.
This attribute is intended to be used by the code generator and LTO to allow
the linker to decide whether the global needs to be in the symbol table. It is
possible to exclude a global from the symbol table if three things are true:
- This attribute is present on every instance of the global (which means that
the normal rule that the global must have a unique address can be broken without
being observable by the program by performing comparisons against the global's
address)
- The global has linkonce_odr linkage (which means that each linkage unit must have
its own copy of the global if it requires one, and the copy in each linkage unit
must be the same)
- It is a constant or a function (which means that the program cannot observe that
the unique-address rule has been broken by writing to the global)
Although this attribute could in principle be computed from the module
contents, LTO clients (i.e. linkers) will normally need to be able to compute
this property as part of symbol resolution, and it would be inefficient to
materialize every module just to compute it.
See:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20160509/356401.htmlhttp://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20160516/356738.html
for earlier discussion.
Part of the fix for PR27553.
Differential Revision: http://reviews.llvm.org/D20348
llvm-svn: 272709
Summary:
We now use a standard fixup type applying the pc-relative address of
constant address space variables, and we have the GlobalAddress lowering
code add the required 4 byte offset to the global address rather than
doing it as part of the fixup.
This refactoring will make it easier to use the same code for global
address space variables and also simplifies the code.
Re-commit this after fixing a bug where we were trying to use a
reference to a Triple object that had already been destroyed.
Reviewers: arsenm, kzhuravl
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: http://reviews.llvm.org/D21154
llvm-svn: 272705
Summary:
We now use a standard fixup type applying the pc-relative address of
constant address space variables, and we have the GlobalAddress lowering
code add the required 4 byte offset to the global address rather than
doing it as part of the fixup.
This refactoring will make it easier to use the same code for global
address space variables and also simplifies the code.
Reviewers: arsenm, kzhuravl
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: http://reviews.llvm.org/D21154
llvm-svn: 272675
The feature allows for conditional assembly etc.
TODO: make those symbols read-only.
Test added.
Differential Revision: http://reviews.llvm.org/D21238
llvm-svn: 272673
Summary:
Mesa and other users must set this to enable coalescing:
- STRIDE = 0
- SWIZZLE_ENABLE = 1
This makes one particular compute shader 8x faster.
Reviewers: tstellarAMD, arsenm
Subscribers: arsenm, kzhuravl
Differential Revision: http://reviews.llvm.org/D21136
llvm-svn: 272556
The condition reg of the cndmask_b64 expansion can't be killed by
the first one, and the implicit super register implicit def is needed.
llvm-svn: 272554
This used to be free, copying and moving DebugLocs became expensive
after the metadata rewrite. Passing by reference eliminates a ton of
track/untrack operations. No functionality change intended.
llvm-svn: 272512
Summary:
We need to set the fixup type to FK_Data_4 for the
SCRATCH_RSRC_DWORD[01] symbols, since these require absolute
relocations, and fixup_si_rodata is for relative relocations.
Reviewers: arsenm, kzhuravl
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: http://reviews.llvm.org/D21153
llvm-svn: 272417
Summary:
sext() modifier is supported in SDWA instructions only for integer operands. Spec is unclear should integer operands support abs and neg modifiers with sext - for now they are not supported.
Renamed InputModsWithNoDefault to FloatInputMods. Added SextInputMods for operands that support sext() modifier.
Added AMDGPUOperand::Modifier struct to handle register and immediate modifiers.
Code cleaning in AMDGPUOperand class: organize method in groups (render-, predicate-methods...).
Reviewers: vpykhtin, artem.tamazov, tstellarAMD
Subscribers: arsenm, kzhuravl
Differential Revision: http://reviews.llvm.org/D20968
llvm-svn: 272384
Summary:
This fixes a bug with ds_*permute instructions where if it was passed a
constant address, then the offset operand would get assigned a register
operand instead of an immediate.
Reviewers: scchan, arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19994
llvm-svn: 272349
This was using extract_subreg sub0 to extract the low register
of the result instead of sub0_sub1, producing an invalid copy.
There doesn't seem to be a way to use the compound subreg indices
in tablegen since those are generated, so manually select it.
llvm-svn: 272344
Summary:
The presence of this attribute indicates that VGPR outputs should be computed
in whole quad mode. This will be used by Mesa for prolog pixel shaders, so
that derivatives can be taken of shader inputs computed by the prolog, fixing
a bug.
The generated code could certainly be improved: if a prolog pixel shader is
used (which isn't common in modern OpenGL - they're used for gl_Color, polygon
stipples, and forcing per-sample interpolation), Mesa will use this attribute
unconditionally, because it has to be conservative. So WQM may be used in the
prolog when it isn't really needed, and furthermore a silly back-and-forth
switch is likely to happen at the boundary between prolog and main shader
parts.
Fixing this is a bit involved: we'd first have to add a mechanism by which
LLVM writes the WQM-related input requirements to the main shader part binary,
and then Mesa specializes the prolog part accordingly. At that point, we may
as well just compile a monolithic shader...
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95130
Reviewers: arsenm, tstellarAMD, mareko
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: http://reviews.llvm.org/D20839
llvm-svn: 272063
Matt Arsenault