Introduced the convergent equivalent of the existing G_INTRINSIC opcodes:
- G_INTRINSIC_CONVERGENT
- G_INTRINSIC_CONVERGENT_W_SIDE_EFFECTS
Out of the targets that currently have some support for GlobalISel, the patch
assumes that the convergent intrinsics only relevant to SPIRV and AMDGPU.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D154766
Implement XCVbi intrinsics for CV32E40P according to the specification.
This commit is part of a patch-set to upstream the 7 vendor specific extensions of CV32E40P.
Contributors: @CharKeaney, @jeremybennett, @lewis-revill, @liaolucy, Nandni Jamnadas, @paolos, @simoncook, @xmj.
bf2ad26b4ff856aab9a62ad168e6bdefeedc374f originally commited.
e4777dc4b9cb371971523cc603e1b8a5c7255e7e reverted due to test failures caused by a merge conflict marker in llvm/test/CodeGen/RISCV/attributes that was accidentally checked in.
This commit removed the conflict marker and recommitted.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D154412
Implement XCVbi intrinsics for CV32E40P according to the specification.
This commit is part of a patch-set to upstream the 7 vendor specific extensions of CV32E40P.
Contributors: @CharKeaney, @jeremybennett, @lewis-revill, @liaolucy, Nandni Jamnadas, @PaoloS, @simoncook, @xmj.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D154412
Implement XCVsimd intrinsics for CV32E40P according to the specification.
This commit is part of a patch-set to upstream the 7 vendor specific extensions of CV32E40P.
Contributors: @CharKeaney, @jeremybennett, @lewis-revill, @liaolucy, Nandni Jamnadas, @PaoloS, @simoncook, @xmj.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D153721
According to the latest spec, Zvfbfwma requires Zvfbfmin and Zvfbfmin requires Zfbfmin, with FLH/FSH/FMV.H.X/HMV.X.H removed from Zvfbfwma.
Reviewed By: asb
Differential Revision: https://reviews.llvm.org/D155916
Implement XCValu intrinsics for CV32E40P according to the specification.
This is a commit of the patch-set to upstream the 7 vendor specific extensions of CV32E40P.
Contributors: @CharKeaney, Nandni Jamnadas, Serkan Muhcu, @jeremybennett, @lewis-revill, @liaolucy, @simoncook, @xmj
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D153748
The example describing bitcasts (and memory layout) involving
vector types was incorrect for little endian.
This was reported in https://reviews.llvm.org/D94964
Updated AggressiveInstCombine pass documentation because now AggressiveInstCombine can modify CFG.
Introduced in D154725.
Differential Revision: https://reviews.llvm.org/D156134
This patch is the first part of https://llvm.org/OpenProjects.html#llvm_patch_coverage.
We have first define a new variable LLVM_TEST_COVERAGE which when set, pass --per-test-coverage option to
llvm-lit which will help in setting a unique value to LLVM_PROFILE_FILE for each RUN. So for example
coverage data for test case llvm/test/Analysis/AliasSet/memtransfer.ll will be emitted as
build/test/Analysis/AliasSet/memtransfer0.profraw
Reviewed By: hnrklssn
Differential Revision: https://reviews.llvm.org/D154280
Currently, objcopy cannot set the new flag SHF_X86_64_LARGE. This change introduces the named flag "large" which translates to that section flag.
An "invalid argument" error is produced if a user attempts to set the flag on an architecture other than X86_64.
Reviewed By: jhenderson, MaskRay
Differential Revision: https://reviews.llvm.org/D153262
rocm-device-libs and llpc were avoiding using f64 sqrt
intrinsics in favor of their own expansions. Port the
expansion into the backend. Both of these users should be
updated to call the intrinsic instead.
The library and llpc expansions are slightly different.
llpc uses an ldexp to do the scale; the library uses a multiply.
Use ldexp to do the scale instead of the multiply.
I believe v_ldexp_f64 and v_mul_f64 are always the same number of
cycles, but it's cheaper to materialize the 32-bit integer constant
than the 64-bit double constant.
The libraries have another fast version of sqrt which will
be handled separately.
I am tempted to do this in an IR expansion instead. In the IR
we could take advantage of computeKnownFPClass to avoid
the 0-or-inf argument check.
On AIX, a libatomic supporting inline quadword atomic operations has been released, so that compatibility is not an issue now, we can enable quadword atomics by default.
Reviewed By: #powerpc, nemanjai
Differential Revision: https://reviews.llvm.org/D151312
When input to intrinsic is uniform value, reduced value is
same as input whereas if input value is divergent we need
to iterate over all active lanes of WaveFront to perform
the reduction.
The control flow for a `loop` has been set up, which
iterates over `only` active lanes to perform reduction.
Introduced WAVE_REDUCE_UMIN_PSEUDO_U32 and
WAVE_REDUCE_UMAX_PSEUDO_U32 Pseudos which
are lowered Post-ISel (in `EmitInstrWithCustomInserter `).
Reviewed By: arsenm, #amdgpu
Differential Revision: https://reviews.llvm.org/D154858
Extends the new frexp scaled reciprocal to the general case. The
reciprocal case is just the same thing when frexp of 1 is constant
folded. Could probably clean up the code to rely on that constant
folding.
Improves results for the IEEE path for the default OpenCL division. We
used to only emit the fdiv.fast intrinsic with a 2.5 ulp accuracy
threshold with DAZ, which uses explicit range checks. This gives us a
better fast option with the default IEEE behavior.
The highlight change is a new denormal safe 1ulp lowering which uses
rcp after using frexp to perform input scaling. This saves 2
instructions compared to other implementations which performed an
explicit denormal range change. This improves the OpenCL default, and
requires a flag for HIP. I don't believe there's any flag wired up for
OpenMP to emit the necessary fpmath metadata.
This provides several improvements and changes that were hard to
separate without regressing one case or another. Disturbingly the
OpenCL conformance test seems to have the reciprocal test commented
out. I locally hacked it back in to test this.
Starts introducing f32 rsq intrinsics in AMDGPUCodeGenPrepare. Like
the rcp case, we could do this in codegen if !fpmath were preserved
(although we would lose some computeKnownFPClass tricks). Start
requiring contract flags to form rsq. The rsq fusion actually improves
the result from ~2ulp to ~1ulp. We have some older fusion in codegen
which only keys off unsafe math which should be refined.
Expand rsq patterns by checking for denormal inputs and pre/post
multiplying like the current library code does. We also take advantage
of computeKnownFPClass to avoid the scaling when we can statically
prove the input cannot be a denormal. We could do the same for the rcp
case, but unlike rsq a large input can underflow to denormal. We need
additional upper bound exponent checks on the input in order to do the
same for rcp.
This rsq handling also now starts handling the negated case. We
introduce rsq with an fneg. In the case the fneg doesn't fold into its
user, it's a neutral change but provides improvement if it is foldable
as a source modifier.
Also starts respecting the arcp attribute properly, and more strictly
interprets afn. We were previously interpreting afn as implying you
could do the reciprocal expansion of an fdiv. The codegen handling of
these also needs to be revisited.
This also effectively introduces the optimization
combineRepeatedFPDivisors enables, just done in the IR instead (and
only for f32).
This is almost across the board better. The one minor regression is
for gfx6/buggy frexp case where for multiple reciprocals, we could
previously reuse rematerialized constants per instance (it's neutral
for a single rcp).
The fdiv.fast and sqrt handling need to be revisited next.
https://reviews.llvm.org/D155593
This patch is the first part of https://llvm.org/OpenProjects.html#llvm_patch_coverage.
We have first define a new variable LLVM_TEST_COVERAGE which when set, pass --emit-coverage option to
llvm-lit which will help in setting a unique value to LLVM_PROFILE_FILE for each RUN. So for example
coverage data for test case llvm/test/Analysis/AliasSet/memtransfer.ll will be emitted as
build/test/Analysis/AliasSet/memtransfer.profraw
Reviewed By: hnrklssn
Differential Revision: https://reviews.llvm.org/D154280
We provide a way to specify arguments in the form of `name=value`
so that we don't have to specify all optional arguments before the
one we'd like to change. Required arguments can alse be specified
in this way.
Note that the argument can only be specified once regardless of
the way (named or positional) to specify and positional arguments
should be put before named arguments.
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D152998
This option had originally been added in D83069 to allow disabling the
check that something is going to get run at all when a specific test name
is used on the command-line. Since we now use getTestsForPath() (from D151664)
to get the tests to run for a specific path, we don't need a specific check
for this anymore -- Lit will produce the same complaint it would produce if
you provided a directory with no tests.
If one needs to run a specific test on the command-line and the Lit
configuration would normally not include that test, the configuration
should be set up as a "standalone" configuration or it should be fixed
to allow for that test to be found (i.e. probably fix the allowed test
suffixes).
Differential Revision: https://reviews.llvm.org/D153967
With the new behaviour, the /MD or similar options aren't added to
e.g. CMAKE_CXX_FLAGS_RELEASE, but are added separately by CMake.
They can be changed by the cmake variable
CMAKE_MSVC_RUNTIME_LIBRARY or with the target property
MSVC_RUNTIME_LIBRARY.
LLVM has had its own custom CMake flags, e.g. LLVM_USE_CRT_RELEASE,
which affects which CRT is used for release mode builds. Deprecate
these and direct users to use CMAKE_MSVC_RUNTIME_LIBRARY directly
instead (and do a best effort attempt at setting CMAKE_MSVC_RUNTIME_LIBRARY
based on the existing LLVM_USE_CRT_ flags). This only handles the
simple cases, it doesn't handle multi-config generators with
different LLVM_USE_CRT_* variables for different configs though,
but that's probably fine - we should move over to the new upstream
CMake mechanism anyway, and push users towards that.
Change code in compiler-rt, that previously tried to override the
CRT choice to /MT, to set CMAKE_MSVC_RUNTIME_LIBRARY instead of
meddling in the old variables.
This resolves the policy issue in
https://github.com/llvm/llvm-project/issues/63286, and should
handle the issues that were observed originally when the
minimum CMake version was bumped, in
https://github.com/llvm/llvm-project/issues/62719 and
https://github.com/llvm/llvm-project/issues/62739.
Differential Revision: https://reviews.llvm.org/D155233
The next sections in GettingStarted assume you are still in the root
directory llvm-project when using ninja.
Make the `cmake --build` command match it as well.
Note: I am a new cmake user and this confused me.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D153727
According the discuss on D154953, we need to make the LangRef change
before the optimization relied on the new behaviour:
vscale_range implies vscale is a power-of-two value, parse of the
attribute to reject values that are not a power-of-two.
Thanks nikic for the wonderful summary of discussing on D154953:
To provide a bit more context here. We would like to have power of two vscale exposed in a target-independent way, so we can make use of this in places like ValueTracking, just like we currently do the vscale range. Some options that have been discussed are:
- Remove support for non-power-of-two vscales entirely. (This is my personal preference, but this is hard to undo if it turns out someone does need them.)
- Add an extra attribute vscale_pow2, or a data layout property.
- Make vscale_range imply power-of-two vscale, as a compromise solution (what this patch does). This would be relatively easy to turn into one of the two above at a later point.
Reviewed By: paulwalker-arm, nikic, efriedma
Differential Revision: https://reviews.llvm.org/D155193
This is a mostly NFC change cleaning up and clarifying components of the
in-tree CORE-V (xcv*) extensions following discussions on the remaining
extensions.
This makes the following changes to the xcbitmanip and xcvmac support:
1. Add missing extensions from RISCVISAInfo, such that they can be
supported in clang's -march option.
2. Clarify the extension version number is 1.0.0 in documentation.
3. Clarify the extensions are by OpenHW Group, and the capitilization
of the CORE-V extension family.
4. Add CORE-V to extension name in RISCVFeatures, both to be consistent
with other vendors, and also better distinguish e.g. CORE-V bit
manipulation vs RISC-V's standard Zb extensions.
Differential Revision: https://reviews.llvm.org/D155283
The test migration to opaque pointers has finished, so we can finally
drop typed pointer support from LLVM \o/
This removes the ability to disable typed pointers, as well as the
-opaque-pointers option, but otherwise doesn't yet touch any API
surface. I'll leave deprecation/removal of compatibility APIs to
future changes.
This also drops a few tests: These are either testing errors that
only occur with typed pointers, or type linking behavior that, to
the best of my knowledge, only applies to typed pointers.
Note that this will break some tests in the experimental SPIRV
backend, because the maintainers have failed to update their tests
in a reasonable time-frame, despite multiple warnings. In accordance
with our experimental target policy, this is not a blocking concern.
This issue is tracked at https://github.com/llvm/llvm-project/issues/60133.
Differential Revision: https://reviews.llvm.org/D155079
Do the LDS frame calculation once, in the IR pass, instead of repeating the work in the backend.
Prior to this patch:
The IR lowering pass sets up a per-kernel LDS frame and annotates the variables with absolute_symbol
metadata so that the assembler can build lookup tables out of it. There is a fragile association between
kernel functions and named structs which is used to recompute the frame layout in the backend, with
fatal_errors catching inconsistencies in the second calculation.
After this patch:
The IR lowering pass additionally sets a frame size attribute on kernels. The backend uses the same
absolute_symbol metadata that the assembler uses to place objects within that frame size.
Deleted the now dead allocation code from the backend. Left for a later cleanup:
- enabling lowering for anonymous functions
- removing the elide-module-lds attribute (test churn, it's not used by llc any more)
- adjusting the dynamic alignment check to not use symbol names
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D155190
Extend D127824 to the 32-bit Power architecture.
AFAICT GNU objdump -d dumps all instructions for 32-bit as well.
Reviewed By: #powerpc, nemanjai
Differential Revision: https://reviews.llvm.org/D155114
This is a reboot of the original design and implementation by
Nicolai Haehnle <nicolai.haehnle@amd.com>:
https://reviews.llvm.org/D85603
This change also obsoletes an earlier attempt at restarting the work on
convergence tokens:
https://reviews.llvm.org/D104504
Changes relative to D85603:
1. Clean up the definition of a "convergent operation", a convergent
call and convergent function.
2. Clean up the relationship between dynamic instances, sets of threads and
convergence tokens.
3. Redistribute the formal rules into the definitions of the convergence
intrinsics.
4. Expand on the semantics of entering a function from outside LLVM,
and the environment-defined outcome of the entry intrinsic.
5. Replace the term "cycle" with "closed path". The static rules are defined
in terms of closed paths, and then a relation is established with cycles.
6. Specify that if a function contains a controlled convergent operation, then
all convergent operations in that function must be controlled.
7. Describe an optional procedure to infer tokens for uncontrolled convergent
operations.
8. Introduce controlled maximal convergence-before and controlled m-converged
property as an update to the original properties in UniformityAnalysis.
9. Additional constraint that a cycle heart can only occur in the header of a
reducible cycle (natural loop).
Reviewed By: nhaehnle
Differential Revision: https://reviews.llvm.org/D147116
