19 Commits

Author SHA1 Message Date
Alex Voicu
c1fabd681f
[llvm][AMDGPU] Enable FWD_PROGRESS bit for GFX10+ (#128367)
From GFX10 onwards it is possible to employ benevolent scheduling of
waves. This patch unconditionally enables, for the `amdhsa` OS, the bit
which controls that capability, as it is beneficial for algorithms that
rely on more complex concurrent coordination and it is generally
performance neutral otherwise.
2025-03-17 23:17:46 +00:00
Stanislav Mekhanoshin
6c9a9d9fe2
[AMDGPU] Set inst_pref_size to maximum (#126981)
On gfx11 and gfx12 set initial instruction prefetch size to a
minimum of kernel size and maximum allowed value.

Fixes: SWDEV-513122
2025-03-03 10:40:31 -08:00
Stanislav Mekhanoshin
2479479285
[AMDGPU] Extend ComputePGMRSrc3 to gfx10+. NFCI. (#129289)
ComputePGMRSrc3 exists since gfx90a and gfx10+. Current code
only expects gfx90a. This is NFCI since we do not fill it on
gfx10+ yet.
2025-03-03 08:22:15 -08:00
Stanislav Mekhanoshin
d19187f5fe
[AMDGPU] Move into SIProgramInfo and cache getFunctionCodeSize. NFCI. (#127111)
This moves function as is, improvements to the estimate go into
a subseqent patch.
2025-02-17 18:22:48 -08:00
Stanislav Mekhanoshin
07405ca036
[AMDGPU] clang-format SIProgramInfo.h. NFC. (#127033) 2025-02-13 09:35:29 -08:00
Janek van Oirschot
17eaa23f7e
[AMDGPU] MCExpr-ify AMDGPU HSAMetadata (#94788)
Enables MCExpr for HSAMetadata, particularly, HSAMetadata's msgpack format.
2024-06-26 16:39:08 +01:00
Jonas Hahnfeld
52c5a81001
[AMDGPU] Fix unittest linking error with LLVM_LINK_LLVM_DYLIB (#91727)
Follow-up to https://github.com/llvm/llvm-project/pull/88257
2024-05-11 09:18:05 +02:00
Janek van Oirschot
d86b68afd7
MCExpr-ify SIProgramInfo (#88257)
Convert members in SIProgramInfo affected by variables provided by AMDGPUResourceUsageAnalysis into MCExprs.
2024-05-09 13:02:32 +01:00
Piotr Sobczak
fac093dd08
[AMDGPU] Update IEEE and DX10_CLAMP for GFX12 (#75030)
Co-authored-by: Stanislav Mekhanoshin <Stanislav.Mekhanoshin@amd.com>
2023-12-13 13:52:40 +01:00
David Stuttard
fc83f1de5d [AMDGPU] Add backend support for new PAL ELF Metadata 3.0
PAL Metadata 3.0 introduces an explicit structure in metadata for the
programmable registers written out by the compiler backend.
Rather than using opaque registers which can change between different
architectures and requires encoding the bitfield information in the backend,
which may change between versions.

This is the initial minimal implementation that enables the use of PAL Metadata
3.0.

The change itself should be NFC for non-PAL, although the way RSRC2 register is
handled has been changed slightly.

The test is fairly minimal, but checks that the metadata format looks as
expected and verifies a couple of special cases such as tgid_[xyz]_en handling
and PsInputAddr/Ena which also change to explicit fields.

Differential Revision: https://reviews.llvm.org/D147143
2023-04-14 09:57:13 +01:00
Vang Thao
67357739c6 [AMDGPU] Add remarks to output some resource usage
Add analyis remarks to output kernel name, register usage, occupancy,
scratch usage, spills, and LDS information.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D123878
2022-07-15 11:01:53 -07:00
Stanislav Mekhanoshin
a8d9d50762 [AMDGPU] gfx90a support
Differential Revision: https://reviews.llvm.org/D96906
2021-02-17 16:01:32 -08:00
Sebastian Neubauer
1124bf4ab7 [AMDGPU] Set rsrc1 flags for graphics shaders
Before they were only set for compute kernels and compute shaders but
not for other shaders.

Differential Revision: https://reviews.llvm.org/D89399
2020-11-04 12:25:41 +01:00
Stanislav Mekhanoshin
075bc48a7f [AMDGPU] separate accounting for agprs
Account and report agprs separately on gfx908. Other targets
do not change the reporting.

Differential Revision: https://reviews.llvm.org/D68307

llvm-svn: 373411
2019-10-02 00:26:58 +00:00
Stanislav Mekhanoshin
2594fa8593 [AMDGPU] Fix high occupancy calculation and print it
We had couple places which still return 10 as a maximum
occupancy. Fixed.

Also print comment about occupancy as compiler see it.

Differential Revision: https://reviews.llvm.org/D65423

llvm-svn: 367381
2019-07-31 01:07:10 +00:00
Stanislav Mekhanoshin
41bbe101a2 [AMDGPU] gfx1010 s_code_end generation
Also add some missing metadata in the streamer.

Differential Revision: https://reviews.llvm.org/D61531

llvm-svn: 359937
2019-05-03 21:26:39 +00:00
Matt Arsenault
aa6fb4c45e AMDGPU: Remove debugger related subtarget features
As far as I know these aren't needed anymore.

llvm-svn: 354634
2019-02-21 23:27:46 +00:00
Chandler Carruth
2946cd7010 Update the file headers across all of the LLVM projects in the monorepo
to reflect the new license.

We understand that people may be surprised that we're moving the header
entirely to discuss the new license. We checked this carefully with the
Foundation's lawyer and we believe this is the correct approach.

Essentially, all code in the project is now made available by the LLVM
project under our new license, so you will see that the license headers
include that license only. Some of our contributors have contributed
code under our old license, and accordingly, we have retained a copy of
our old license notice in the top-level files in each project and
repository.

llvm-svn: 351636
2019-01-19 08:50:56 +00:00
Scott Linder
2ad2c18b82 [AMDGPU] Refactor HSAMetadataStream::emitKernel (NFC)
Move all metadata construction into AMDGPUHSAMetadataStreamer.

Differential Revision: https://reviews.llvm.org/D48176

llvm-svn: 336707
2018-07-10 17:31:32 +00:00