llvm-project.git/llvm/test/CodeGen/AMDGPU/lower-kernel-lds-constexpr.ll, branch users/nico/python-2 Unnamed repository; edit this file 'description' to name the repository. AMDGPU: Use poison instead of undef in module lds pass 2023-09-02T15:33:26+00:00 Matt Arsenault Matthew.Arsenault@amd.com 2023-09-02T15:13:18+00:00 1f52060000f1fead314f7173f4c62bc58b9ca7c3 






[AMDGPU] Convert some tests to opaque pointers (NFC)
2022-12-19T11:41:13+00:00

Nikita Popov
npopov@redhat.com

2022-12-19T11:39:01+00:00

bdf2fbba9cee60b4b260ff17e4f44c475c11e715












[amdgpu] Reimplement LDS lowering
2022-12-07T22:02:54+00:00

Jon Chesterfield
jonathanchesterfield@gmail.com

2022-12-07T22:02:53+00:00

d77ae7f2513504655e555cd326208598093d66e2

Renames the current lowering scheme to "module" and introduces two new
ones, "kernel" and "table", plus a "hybrid" that chooses between those three
on a per-variable basis.

Unit tests are set up to pass with the default lowering of "module" or "hybrid"
with this patch defaulting to "module", which will be a less dramatic codegen
change relative to the current. This reflects the sparsity of test coverage for
the table lowering method. Hybrid is better than module in every respect and
will be default in a subsequent patch.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D139433





Renames the current lowering scheme to "module" and introduces two new
ones, "kernel" and "table", plus a "hybrid" that chooses between those three
on a per-variable basis.

Unit tests are set up to pass with the default lowering of "module" or "hybrid"
with this patch defaulting to "module", which will be a less dramatic codegen
change relative to the current. This reflects the sparsity of test coverage for
the table lowering method. Hybrid is better than module in every respect and
will be default in a subsequent patch.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D139433







Revert "[amdgpu] Reimplement LDS lowering"
2022-12-06T17:01:36+00:00

Nico Weber
thakis@chromium.org

2022-12-06T17:01:36+00:00

a862d09a92cac6a93bf8aac9c2c13197661f9c8e

This reverts commit 982017240d7f25a8a6969b8b73dc51f9ac5b93ed.
Breaks check-llvm, see https://reviews.llvm.org/D139433#3974862





This reverts commit 982017240d7f25a8a6969b8b73dc51f9ac5b93ed.
Breaks check-llvm, see https://reviews.llvm.org/D139433#3974862







[amdgpu] Reimplement LDS lowering
2022-12-06T16:28:15+00:00

Jon Chesterfield
jonathanchesterfield@gmail.com

2022-12-06T16:10:42+00:00

982017240d7f25a8a6969b8b73dc51f9ac5b93ed

Renames the current lowering scheme to "module" and introduces two new
ones, "kernel" and "table", plus a "hybrid" that chooses between those three
on a per-variable basis.

Unit tests are set up to pass with the default lowering of "module" or "hybrid"
with this patch defaulting to "module", which will be a less dramatic codegen
change relative to the current. This reflects the sparsity of test coverage for
the table lowering method. Hybrid is better than module in every respect and
will be default in a subsequent patch.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D139433





Renames the current lowering scheme to "module" and introduces two new
ones, "kernel" and "table", plus a "hybrid" that chooses between those three
on a per-variable basis.

Unit tests are set up to pass with the default lowering of "module" or "hybrid"
with this patch defaulting to "module", which will be a less dramatic codegen
change relative to the current. This reflects the sparsity of test coverage for
the table lowering method. Hybrid is better than module in every respect and
will be default in a subsequent patch.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D139433







[amdgpu] Expand all ConstantExpr users of LDS variables in instructions
2022-09-14T06:55:46+00:00

Jon Chesterfield
jonathanchesterfield@gmail.com

2022-09-14T06:55:44+00:00

cdb9738963a1584d6530bcf7b102423f54a6bdb1

Bug noted in D112717 can be sidestepped with this change.

Expanding all ConstantExpr involved with LDS up front makes the variable specialisation simpler. Excludes ConstantExpr that don't access LDS to avoid disturbing codegen elsewhere.

Reviewed By: rampitec

Differential Revision: https://reviews.llvm.org/D133422





Bug noted in D112717 can be sidestepped with this change.

Expanding all ConstantExpr involved with LDS up front makes the variable specialisation simpler. Excludes ConstantExpr that don't access LDS to avoid disturbing codegen elsewhere.

Reviewed By: rampitec

Differential Revision: https://reviews.llvm.org/D133422







[IR] Replace *all* uses of a constant expression by corresponding instruction
2021-11-02T04:31:46+00:00

hsmahesha
mahesha.comp@gmail.com

2021-11-02T04:31:36+00:00

e9ea99249623f19563ec0c4c8d6a70fb5c4b670f

When a constant expression CE is being converted into a corresponding instruction I,
CE is supposed to be replaced by I. However, it is possible that CE is being used multiple
times within a parent instruction PI. Make sure that *all* the uses of CE within PI are
replaced by I.

Reviewed By: rampitec, arsenm

Differential Revision: https://reviews.llvm.org/D112717





When a constant expression CE is being converted into a corresponding instruction I,
CE is supposed to be replaced by I. However, it is possible that CE is being used multiple
times within a parent instruction PI. Make sure that *all* the uses of CE within PI are
replaced by I.

Reviewed By: rampitec, arsenm

Differential Revision: https://reviews.llvm.org/D112717







[AMDGPU] Propagate LDS align into to instructions
2021-06-23T07:57:16+00:00

Stanislav Mekhanoshin
Stanislav.Mekhanoshin@amd.com

2021-06-15T00:01:54+00:00

2b43209ee37ce27cede17cf5424e7655adfe3ac2

Differential Revision: https://reviews.llvm.org/D104316





Differential Revision: https://reviews.llvm.org/D104316







[AMDGPU] Fixed constexpr expansion to handle multiple uses
2021-06-16T23:57:41+00:00

Stanislav Mekhanoshin
Stanislav.Mekhanoshin@amd.com

2021-06-16T22:38:39+00:00

0a07343e34fc84052e6cc54e55654412b0ed01f4

Recently added convertConstantExprsToInstructions() does not handle
a case when a same ConstantExpr used multiple times in the same
instruction. A first use is replaced and the rest of the uses in the
instruction are replaced as well with the replaceUsesOfWith(). Then
function attempts to replace a constant already destroyed.

So far this interface is only used by the AMDGPU BE.

Differential Revision: https://reviews.llvm.org/D104425





Recently added convertConstantExprsToInstructions() does not handle
a case when a same ConstantExpr used multiple times in the same
instruction. A first use is replaced and the rest of the uses in the
instruction are replaced as well with the replaceUsesOfWith(). Then
function attempts to replace a constant already destroyed.

So far this interface is only used by the AMDGPU BE.

Differential Revision: https://reviews.llvm.org/D104425







[AMDGPU] Handle constant LDS uses from different kernels
2021-06-07T22:39:08+00:00

Stanislav Mekhanoshin
Stanislav.Mekhanoshin@amd.com

2021-06-07T22:30:06+00:00

05289dfb62461b73654c5fd64b1ef044a88ee027

This allows to lower an LDS variable into a kernel structure
even if there is a constant expression used from different
kernels.

Differential Revision: https://reviews.llvm.org/D103655





This allows to lower an LDS variable into a kernel structure
even if there is a constant expression used from different
kernels.

Differential Revision: https://reviews.llvm.org/D103655