This builtin allows the creation of custom scheduling pipelines on a per-region
basis. Like the sched_barrier builtin this is intended to be used either for
testing, in situations where the default scheduler heuristics cannot be
improved, or in critical kernels where users are trying to get performance that
is close to handwritten assembly. Obviously using these builtins will require
extra work from the kernel writer to maintain the desired behavior.
The builtin can be used to create groups of instructions called "scheduling
groups" where ordering between the groups is enforced by the scheduler.
__builtin_amdgcn_sched_group_barrier takes three parameters. The first parameter
is a mask that determines the types of instructions that you would like to
synchronize around and add to a scheduling group. These instructions will be
selected from the bottom up starting from the sched_group_barrier's location
during instruction scheduling. The second parameter is the number of matching
instructions that will be associated with this sched_group_barrier. The third
parameter is an identifier which is used to describe what other
sched_group_barriers should be synchronized with. Note that multiple
sched_group_barriers must be added in order for them to be useful since they
only synchronize with other sched_group_barriers. Only "scheduling groups" with
a matching third parameter will have any enforced ordering between them.
As an example, the code below tries to create a pipeline of 1 VMEM_READ
instruction followed by 1 VALU instruction followed by 5 MFMA instructions...
// 1 VMEM_READ
__builtin_amdgcn_sched_group_barrier(32, 1, 0)
// 1 VALU
__builtin_amdgcn_sched_group_barrier(2, 1, 0)
// 5 MFMA
__builtin_amdgcn_sched_group_barrier(8, 5, 0)
// 1 VMEM_READ
__builtin_amdgcn_sched_group_barrier(32, 1, 0)
// 3 VALU
__builtin_amdgcn_sched_group_barrier(2, 3, 0)
// 2 VMEM_WRITE
__builtin_amdgcn_sched_group_barrier(64, 2, 0)
Reviewed By: jrbyrnes
Differential Revision: https://reviews.llvm.org/D128158
Adds an intrinsic/builtin that can be used to fine tune scheduler behavior. If
there is a need to have highly optimized codegen and kernel developers have
knowledge of inter-wave runtime behavior which is unknown to the compiler this
builtin can be used to tune scheduling.
This intrinsic creates a barrier between scheduling regions. The immediate
parameter is a mask to determine the types of instructions that should be
prevented from crossing the sched_barrier. In this initial patch, there are only
two variations. A mask of 0 means that no instructions may be scheduled across
the sched_barrier. A mask of 1 means that non-memory, non-side-effect inducing
instructions may cross the sched_barrier.
Note that this intrinsic is only meant to work with the scheduling passes. Any
other transformations that may move code will not be impacted in the ways
described above.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D124700
This adds -no-opaque-pointers to clang tests whose output will
change when opaque pointers are enabled by default. This is
intended to be part of the migration approach described in
https://discourse.llvm.org/t/enabling-opaque-pointers-by-default/61322/9.
The patch has been produced by replacing %clang_cc1 with
%clang_cc1 -no-opaque-pointers for tests that fail with opaque
pointers enabled. Worth noting that this doesn't cover all tests,
there's a remaining ~40 tests not using %clang_cc1 that will need
a followup change.
Differential Revision: https://reviews.llvm.org/D123115
This reverts commit ab98f2c712 and 98eea392cd.
It includes a fix for the clang test which triggered the revert. I failed to notice this one because there was another AMDGPU llvm test with a similiar name and the exact same text in the error message. Odd. Since only one build bot reported the clang test, I didn't notice that one.
[AMDGPU] Add __builtin_amdgcn_grid_size
Similar to D76772, loads the data from the dispatch pointer. Marked invariant.
Patch also updates the openmp devicertl to use this builtin.
Reviewed By: yaxunl
Differential Revision: https://reviews.llvm.org/D90251
The main purpose of introducing these builtins is to add a range
metadata [1, 1025) on the work group size loaded from dispatch
ptr, which cannot be done by source code.
Differential Revision: https://reviews.llvm.org/D76772
The way address space declarations for builtins currently work
is nearly useless. The code assumes the address spaces used for
builtins is a confusingly named "target address space" from user
code using __attribute__((address_space(N))) that matches
the builtin declaration. There's no way to use this to declare
a builtin that returns a language specific address space.
The terminology used is highly cofusing since it has nothing
to do with the the address space selected by the target to use
for a language address space.
This feature is essentially unused as-is. AMDGPU and NVPTX
are the only in-tree targets attempting to use this. The AMDGPU
builtins certainly do not behave as intended (i.e. all of the
builtins returning pointers can never compile because the numbered
address space never matches the expected named address space).
The NVPTX builtins are missing tests for some, and the others
seem to rely on an implicit addrspacecast.
Change the used address space for builtins based on a target
hook to allow using a language address space for a builtin.
This allows the same builtin declaration to be used for multiple
languages with similarly purposed address spaces (e.g. the same
AMDGPU builtin can be used in OpenCL and CUDA even though the
constant address spaces are arbitarily different).
This breaks the possibility of using arbitrary numbered
address spaces alongside the named address spaces for builtins.
If this is an issue we probably need to introduce another builtin
declaration character to distinguish language address spaces from
so-called "target address spaces".
llvm-svn: 338707
A recent change requires opencl triple environment for compiling OpenCL
program, which causes regressions in libclc.
This patch fixes that. Instead of deducing language based on triple
environment, it checks LangOptions.
Differential Revision: https://reviews.llvm.org/D33445
llvm-svn: 303644
Alloca always returns a pointer in alloca address space, which may
be different from the type defined by the language. For example,
in C++ the auto variables are in the default address space. Therefore
cast alloca to the expected address space when necessary.
Differential Revision: https://reviews.llvm.org/D32248
llvm-svn: 303370
The wave barrier represents the discardable barrier. Its main purpose is to
carry convergent attribute, thus preventing illegal CFG optimizations. All lanes
in a wave come to convergence point simultaneously with SIMT, thus no special
instruction is needed in the ISA. The barrier is discarded during code generation.
Differential Revision: https://reviews.llvm.org/D26584
llvm-svn: 287006
Summary:
int __builtin_amdgcn_ds_swizzle (int a, int imm);
while imm is a constant.
Differential Revision:
http://reviews.llvm.org/D23682
llvm-svn: 279165
Austin Kerbow