Reorganize project folders#6

Merged

jayhawk-commits merged 824 commits into

Apr 30, 2025

This pull request is big! We're only showing the most recent 250 commits

Commits on Feb 25, 2025

Skip InProcess assembler for new archs

pjots
authored and
memmett
committed
Retrieve Architecture Information from the Machine Readable ISA

bethune-bryant
authored and
memmett
committed
Allow use of internal MRISA files when building rocRoller.

lawruble13
authored and
memmett
committed
Support building and testing on a machine with no GPU

awhittle3
authored and
memmett
committed
QA for OCP vs NANOO mode to Settings class

pjots
authored and
memmett
committed
Add FP6 and BF6 GEMM test with FP32 output

amd-chunxlin
authored and
memmett
committed
Added more expansive emulator test suite

pjots
authored and
memmett
committed
Create a ScaledMatrixMultiply test for one tile with fixed scale values

amd-chunxlin
authored and
memmett
committed
Add mixed precision matrix-multiply tests.
memmett
committed
Test Remaining 8/6/4 GEMM combinations

authored and
memmett
committed
Add scaled mixed-precision matrix multiplication tests

amd-chunxlin
authored and
memmett
committed
Test gfx950 new convert functions: test round-trip

pjots
authored and
memmett
committed
A separate suite for client mixed gemm tests

authored and
memmett
committed
Create unit tests for LDS transpose loads for 4, 6, 8, and 16-bit data types

jaopaulolc
authored and
memmett
committed
QA for scaled gemm tests

authored and
memmett
committed
Set gtest discovery_mode to pre_test

authored and
memmett
committed
Run different kernel options on emulator

jaopaulolc
authored and
memmett
committed
Fix inf/nan/clip for convert from FP32/FP16 to OCP FP8

authored and
memmett
committed
Merge branch 'master' into 'gfx950'.
memmett
committed
Add test to verify Settings::get is threadsafe

jaopaulolc
authored and
memmett
committed
QA for Fix inf/nan/clip for convert from FP32/FP16 to OCP FP8

amd-chunxlin
authored and
memmett
committed
Modify thread safety test.

bethune-bryant
authored and
memmett
committed
Update Rocm build & Remove hacks as assembler bugs were fixed

committed
Remove unused headers

authored and
memmett
committed
Merge commit 'e7610fc9' into gfx950
memmett
committed
Flatten template of PotentialObservers

awhittle3
authored and
memmett
committed
Update GEMM client: generate, validate, benchmark; support loading from assembly.
memmett
committed
Fix user-entrypoint to be able to restart a dev container.

bethune-bryant
authored and
memmett
committed
Add Pseudo random number generator

amd-chunxlin
authored and
memmett
committed
Merge commit '2584b5fe' into maemmett/update-gfx950
memmett
committed
Merge bug fix for settings race condition into master

jaopaulolc
authored and
memmett
committed
Merge commit '376eec09c072d6b755afd63d36564c5930b14d22' of github.com:ROCm/rocRoller into gfx950

bethune-bryant
authored and
memmett
committed
Merge commit '777bbb73' into maemmett/update-gfx950
memmett
committed
Update to rocm-6.2.2

awhittle3
authored and
memmett
committed
Merge commit 'e64a396b2b4cb01832238c0b211cb70e76fb86e3' of github.com:ROCm/rocRoller into gfx950

bethune-bryant
authored and
memmett
committed
Merge commit '83968475a9ffff796f60b9682a1fc9e19c6a104c' into awhittle/update-gfx950-6

awhittle3
authored and
memmett
committed
Ensure allocation is even aligned with odd chunks

jaopaulolc
authored and
memmett
committed
Merge commit '13e8f06a8b175d6b267c2713b36442feda856940' into gfx950

jaopaulolc
authored and
memmett
committed
Merge pull request from ROCm/brnelson/post_mrisa_950_merge

sdquiring
authored and
memmett
committed
Add unit tests trLoad F4/F6/F8/F16 for 32x32x64 mfma

authored and
memmett
committed
Add support for in/out operands for instructions

awhittle3
authored and
memmett
committed
Merge branch 'gfx950' into joao/update_gfx950

jaopaulolc
authored and
memmett
committed
Blockscale Transforms

authored and
memmett
committed
Add PERMLANE instructions to hazard observers

awhittle3
authored and
memmett
committed
Use data-generator for the GEMM inputs

liu-yiqian
authored and
memmett
committed
Merges 13e8f06 into gfx950.

jaopaulolc
authored and
memmett
committed
Add Pseudo random number generator

amd-chunxlin
authored and
memmett
committed
Adds predicate functionality to CommandKernel & validates K size when K loop is unrolled

lawruble13
authored and
memmett
committed
Check whether archives for tested GPU exist

awhittle3
authored and
memmett
committed
LLVM YAML serialization and linking fix

awhittle3
authored and
memmett
committed
Fix issues caused by new amdclang compiler

jaopaulolc
authored and
memmett
committed
Migrate mxDataGeneratorTest to catch2

jaopaulolc
authored and
memmett
committed
Add gfx 12*

bethune-bryant
authored and
memmett
committed
Merge commit '651e2b5c76f20f0a37c1c66fc89769d999234153' into yiqialiu/datagen-gfx950

liu-yiqian
authored and
memmett
committed
FP32 to FP8/BF8 Conversion with non-changing stochastic value

authored and
memmett
committed
Merge branch 'gfx950' of github.com:ROCm/rocRoller into yiqialiu/datagen-gfx950

liu-yiqian
authored and
memmett
committed
Add codegen for transpose load from LDS

jaopaulolc
authored and
memmett
committed
Add Testing section to pull request template.

sdquiring
authored and
memmett
committed
Use data generator on gfx950

liu-yiqian
authored and
memmett
committed
Add BitFieldExtract expression

lawruble13
authored and
memmett
committed
Add support for large (> 16 bit) LDS offsets.

authored and
memmett
committed
Fix YAML serialization for small floating-point types.

sdquiring
authored and
memmett
committed
Add UInt8x4 datatype

pjots
authored and
memmett
committed
Split CMakeLists into chunks

awhittle3
authored and
memmett
committed
Fix codecov zip location

awhittle3
authored and
memmett
committed
Fix codecov zipping paths

awhittle3
authored and
memmett
committed
Expression Fixes

sdquiring
authored and
memmett
committed
Add a function containsSubexpression() for expressions

amd-chunxlin
authored and
memmett
committed
Merge commit '6726786ba38c99b74e4a54b168d66338cb7224cd' into gfx950

sdquiring
authored and
memmett
committed
Merge branch 'master' into gfx_950

bethune-bryant
authored and
memmett
committed
QA for Add BFE Expression

pjots
authored and
memmett
committed
Fix merge issues.

sdquiring
authored and
memmett
committed
Master Merge

sdquiring
authored and
memmett
committed
Fix include pathing

authored and
memmett
committed
Merge branch 'master' into gfx950

lawruble13
authored and
memmett
committed
Merge expression fixes into gfx950.

sdquiring
authored and
memmett
committed
Codecov fixes

bethune-bryant
authored and
memmett
committed
Merge branch 'master' into gfx950

lawruble13
authored and
memmett
committed
Check all source registers in VALUWriteReadlane94x observer.

amd-chunxlin
authored and
memmett
committed
Add yoonchoi

yoonseoch
authored and
memmett
committed
Fix merge issues

lawruble13
authored and
memmett
committed
Add padding to LDS for F6 transposes

jaopaulolc
authored and
memmett
committed
Test direct to lds capability

liu-yiqian
authored and
memmett
committed
Update gfx950 from master.

lawruble13
authored and
memmett
committed
Add K tail loops.

sdquiring
authored and
memmett
committed
Merge branch 'gfx950' of github.com:ROCm/rocRoller into update_gfx950

liu-yiqian
authored and
memmett
committed
More codecov cleanup.

bethune-bryant
authored and
memmett
committed
Update gfx950 from master.

lawruble13
authored and
memmett
committed
Support random number generation using Command

amd-chunxlin
authored and
memmett
committed
Use idiomatic address register updates for large offset support.

bethune-bryant
authored and
memmett
committed
Add support for loading kernels via code-objects in the GEMM client.
memmett
committed
Add unit tests for long dword buffer_load and ds_store for blockscale

pjots
authored and
memmett
committed
Replace flat_load/store with global_load/store

amd-chunxlin
authored and
memmett
committed
Enable wider buffer load instructions for direct to LDS

liu-yiqian
authored and
memmett
committed
Merge remote-tracking branch 'origin/gfx950' into update_gfx950

jaopaulolc
authored and
memmett
committed
Add IdentifyParallelDimensions graph transformation.

sdquiring
authored and
memmett
committed
Fix pad-bytes for B matrix.
memmett
committed
Merge remote-tracking branch 'origin/master' into update_gfx950

jaopaulolc
authored and
memmett
committed
Merge branch 'gfx950' into update_gfx950

jaopaulolc
authored and
memmett
committed
Standardize RR includes to #include <rocRoller/...>

jaopaulolc
authored and
memmett
committed
Merge remote-tracking branch 'origin/gfx950' into update_gfx950

jaopaulolc
authored and
memmett
committed
Update gfx950 from master

jaopaulolc
authored and
memmett
committed
Fix Debug builds

jaopaulolc
authored and
memmett
committed
Update gfx950 from master

liu-yiqian
authored and
memmett
committed
Update READMEs

awhittle3
authored and
memmett
committed
QA add setCoord to set the initial value of K for the tailLoop ComputeIndex

pjots
authored and
memmett
committed
Merge remote-tracking branch 'origin/gfx950' into update_gfx950

jaopaulolc
authored and
memmett
committed
Prework for creating installation packages.

authored and
memmett
committed
Merge remote-tracking branch 'origin/gfx950' into update_gfx950

awhittle3
authored and
memmett
committed
Update gfx950 from master

jaopaulolc
authored and
memmett
committed
Move Convert expression data type from template parameter to data member

yoonseoch
authored and
memmett
committed
Merge pull request #1079 from ROCm/update_gfx950

awhittle3
authored and
memmett
committed
Resolving conflicts manually.

yoonseoch
authored and
memmett
committed
Reduce number of buffer descriptors.
memmett
committed
Add MX scaling support to GEMM client.
memmett
committed
Adding FP/BF6x16 and FP4x8 as valid destination types of convert expression

yoonseoch
authored and
memmett
committed
Merge remote-tracking branch 'origin/gfx950' into update_gfx950

awhittle3
authored and
memmett
committed
Merge remote-tracking branch 'origin/gfx950' into update_gfx950

yoonseoch
authored and
memmett
committed
Add nhenders to CODEOWNERS

ThanHenderson
authored and
memmett
committed
Update TensorDescriptor.hpp path

awhittle3
authored and
memmett
committed
Merge pull request from ROCm/update_gfx950

yoonseoch
authored and
memmett
committed
Small tweaks for HipBLASLt integration

awhittle3
authored and
memmett
committed
Merge branch 'gfx950' of github.com:ROCm/rocRoller into update_gfx950

liu-yiqian
authored and
memmett
committed
Merge pull request from ROCm/update_gfx950

awhittle3
authored and
memmett
committed
Add LoadTileDirect2LDS code generator

liu-yiqian
authored and
memmett
committed
Merge branch 'gfx950' of github.com:ROCm/rocRoller into update_gfx950

liu-yiqian
authored and
memmett
committed
Merge branch 'update_gfx950' of github.com:ROCm/rocRoller into update_gfx950

liu-yiqian
authored and
memmett
committed
Merge branch 'gfx950' into update_gfx950

liu-yiqian
authored and
memmett
committed
Update rocm build version

awhittle3
authored and
memmett
committed
Update gfx950 from master

liu-yiqian
authored and
memmett
committed
Enable Register::Value to represent a bitfield within a register

jaopaulolc
authored and
memmett
committed
Add Exchange control node and unit tests

pjots
authored and
memmett
committed
Merge branch gfx950 into update_gfx950

jaopaulolc
authored and
memmett
committed
Remove magic numbers. Client doesn't support single-scales.
memmett
committed
Enable Arithmetic tests on GFX10/12

jaopaulolc
authored and
memmett
committed
Improve the performance of finding nodes' orders in control graph

amd-chunxlin
authored and
memmett
committed
Enable creation of install packages.

lawruble13
authored and
memmett
committed
Split AddLDS transform into: an early AddLDS transform and a late AddPrefetch transform.
memmett
committed
Remove unnecessary fast arithmetic transformation which would be done in Expression::generate

amd-chunxlin
authored and
memmett
committed
Avoid adding redundant SetCoordinate nodes during unrollLoop

ThanHenderson
authored and
memmett
committed
Fix Performance Diff Rendering in Github Comment

bethune-bryant
authored and
memmett
committed
Merge branch 'master' into 'gfx950'.
memmett
committed
Merge branch 'gfx950' into update_gfx950

jaopaulolc
authored and
memmett
committed
Support GFX12 normal barriers

ThanHenderson
authored and
memmett
committed
Fix internal tile for conversion tests.
memmett
committed
Merge branch 'gfx950' into update_gfx950
memmett
committed
Only link libdivide and the GPUArchitecture_def library at build time.

lawruble13
authored and
memmett
committed
Remove debug code.
memmett
committed
Update gfx950 from master
memmett
committed
QA for Enable direct global to lds code generator

amd-chunxlin
authored and
memmett
committed
Add GFX12 support for buffer_load and buffer_store soffset

ThanHenderson
authored and
memmett
committed
Enable waitcnt generation on GFX12

jaopaulolc
authored and
memmett
committed
Fix LLVM YAML Backend

awhittle3
authored and
memmett
committed
Allow launchKernel to take a stream as an argument

amd-chunxlin
authored and
memmett
committed
Add Direct2LDS graph transform

liu-yiqian
authored and
memmett
committed
Update CodeQL and fix issues

awhittle3
authored and
memmett
committed
Merge branch 'master' into lwrubles/update_gfx950

lawruble13
authored and
memmett
committed
Update gfx950 from master 2025-01-27
memmett
committed
Add toString and output stream for all enums

amd-chunxlin
authored and
memmett
committed
Update gfx950 from master.

lawruble13
authored and
memmett
committed
Resolve merge conflicts

amd-chunxlin
authored and
memmett
committed
Update gfx950 from master

amd-chunxlin
authored and
memmett
committed
Restore Jenkins, Docker and README files.
memmett
committed
Remove Graupel.
memmett
committed

Commits on Mar 3, 2025

Store architecture information in repo

authored and
GitHub Enterprise
committed

Commits on Mar 7, 2025

Set PRs to use the parameter defaults from the job of the target branch

lawruble13
authored and
GitHub Enterprise
committed
Add formatting script for architecture YAML files

ThanHenderson
authored and
GitHub Enterprise
committed
Update files with license

awhittle3
authored and
GitHub Enterprise
committed
Add initial GFX12 WMMA write hazard and observers

ThanHenderson
authored and
GitHub Enterprise
committed

Commits on Mar 8, 2025

Add unit tests for colourByUnrollValue and RemoveDuplicates. Tidy RemoveDuplicates.

authored and
GitHub Enterprise
committed

Commits on Mar 11, 2025

Fix naming for code-coverage job (#1149)

lawruble13
authored and
GitHub Enterprise
committed
QA: Delete redundant SetCoordinate nodes

authored and
GitHub Enterprise
committed
Enable remaining non-GEMM tests on GFX12

jaopaulolc
authored and
GitHub Enterprise
committed
LWP100-1523 Get F6 kernels working again

authored and
GitHub Enterprise
committed

Commits on Mar 12, 2025

Extending BitwiseAnd generation for 64-bit literals

authored and
GitHub Enterprise
committed
Change from std's format to libfmt

awhittle3
authored and
GitHub Enterprise
committed
Allow configuration of mxDataGenerator source

lawruble13
authored and
GitHub Enterprise
committed

Commits on Mar 13, 2025

QA for Add MIT licenses to all source files

authored and
GitHub Enterprise
committed

Commits on Mar 14, 2025

Add support for ScaleMode::SingleScale. Allow A/B scale modes to be mixed.

memmett
authored and
GitHub Enterprise
committed
Coordinate Graph changes to enable WMMA MatrixMultiply on GFX12

authored and
GitHub Enterprise
committed

Commits on Mar 17, 2025

Change log level of execute kernel

bnemanich
authored and
GitHub Enterprise
committed
Include Load/store packing info in control graph.

sdquiring
authored and
GitHub Enterprise
committed

Commits on Mar 18, 2025

Add a flag to observe the buffer_load lds instruction.

liu-yiqian
authored and
GitHub Enterprise
committed
Add Buffer Store Hazard Observer

authored and
GitHub Enterprise
committed

Commits on Mar 19, 2025

Fix lifetime of SCRATCH buffer for client StreamK runs and re-enable rrperf StreamK runs.

memmett
authored and
GitHub Enterprise
committed
Load consecutive data when direct-2-LDS is used, and enable longer buffer load instructions.

liu-yiqian
authored and
GitHub Enterprise
committed
Widen expression to 64bit

yoonseoch
authored and
GitHub Enterprise
committed
64bit BitwiseAnd generation over non-64bit operands

authored and
GitHub Enterprise
committed

Commits on Mar 20, 2025

libfmt as a fetched dependency

awhittle3
authored and
GitHub Enterprise
committed
Add WMMA Basic GEMM tests with f32 accumulation

ThanHenderson
authored and
GitHub Enterprise
committed

Commits on Mar 21, 2025

Changes to build with newer amdclang

jaopaulolc
authored and
GitHub Enterprise
committed
Fix BFE expression generation & add tests

lawruble13
authored and
GitHub Enterprise
committed
Cleanup tmp files left by SubprocessAssembler

awhittle3
authored and
GitHub Enterprise
committed

Commits on Mar 22, 2025

Allow direct-from-global loads (for MX) to be used with prefetching.

memmett
authored and
GitHub Enterprise
committed
Parameterise Direct2LDS tests in GEMMTest

memmett
authored and
GitHub Enterprise
committed

Commits on Mar 23, 2025

Add a test that checks no redundant NOPs is generated

liu-yiqian
authored and
GitHub Enterprise
committed
Add mixed WMMA MatrixMultiply tests

ThanHenderson
authored and
GitHub Enterprise
committed
Update CODEOWNERS to refer to team

ThanHenderson
authored and
GitHub Enterprise
committed

Commits on Mar 24, 2025

Skip GPU assert tests on gfx908

liu-yiqian
authored and
GitHub Enterprise
committed
Enable buffer_load/ds_read dword and swizzle for block scale

pjots
authored and
GitHub Enterprise
committed

Commits on Mar 25, 2025

Fix packaging when included as a subdirectory (e.g. via FetchContent)

lawruble13
authored and
GitHub Enterprise
committed
Additional WWMA Hazard Observers

authored and
GitHub Enterprise
committed

Commits on Mar 26, 2025

Remove nMacY from StreamK SCRATCH tile

ThanHenderson
authored and
GitHub Enterprise
committed
Remove old benchmarks and old guideposts

awhittle3
authored and
GitHub Enterprise
committed

Commits on Mar 27, 2025

Move tests in CMakeLists.txt to go with tests

awhittle3
authored and
GitHub Enterprise
committed
Hotfixes for HipBLASLt Integration

awhittle3
authored and
GitHub Enterprise
committed

Commits on Mar 28, 2025

Link Boost publically

awhittle3
authored and
GitHub Enterprise
committed
Revert "Link Boost publicly"

bnemanich
authored and
GitHub Enterprise
committed

Commits on Apr 1, 2025

Change `add` in `FuseTernary` to match the pattern of other fusions when fusing to `AddShiftL`.

lawruble13
authored and
GitHub Enterprise
committed
Link Boost publicly

bnemanich
authored and
GitHub Enterprise
committed
Ignore boost in CodeQL report

jaopaulolc
authored and
GitHub Enterprise
committed

Commits on Apr 2, 2025

Use/emit expcnt on supported architectures. Move WMMA tests to catch2.

jaopaulolc
authored and
GitHub Enterprise
committed
Add new GEMM tests with large macro tiles to profile kernel generation time.

amd-chunxlin
authored and
GitHub Enterprise
committed
Add client SwizzleScale and assertions.

liu-yiqian
authored and
GitHub Enterprise
committed

Commits on Apr 3, 2025

Fix performance job parameter to skip building against the target branch.

lawruble13
authored and
GitHub Enterprise
committed
Remove KernelGraph dependency from CommandSolution.hpp

bnemanich
authored and
GitHub Enterprise
committed
Remove references to gfx940 and gfx941

bnemanich
authored and
GitHub Enterprise
committed

Commits on Apr 4, 2025

Add new tests for BFE generator

amd-chunxlin
authored and
GitHub Enterprise
committed

Commits on Apr 7, 2025

Link yaml-cpp and boost statically

awhittle3
authored and
GitHub Enterprise
committed
Disable unused globals flake8 checks. Adjust assembly check.

jaopaulolc
authored and
GitHub Enterprise
committed

Commits on Apr 9, 2025

Enable DataFlowTags to alias each other to reduce register usage.

sdquiring
authored and
GitHub Enterprise
committed
Implement new compareNodes function without using orderCache

authored and
GitHub Enterprise
committed

Commits on Apr 10, 2025

Update Boost version and only build multi_index.

awhittle3
authored and
GitHub Enterprise
committed
Add ScopeManagerTest and remove scale-size magic number

amd-chunxlin
authored and
GitHub Enterprise
committed

Commits on Apr 11, 2025

Enable GFX12 F16 accumulation matrix instructions

ThanHenderson
authored and
GitHub Enterprise
committed
Fix gemm client config loading (types, transposes, scaling).

memmett
authored and
GitHub Enterprise
committed

Commits on Apr 14, 2025

Fix SPR scheduler-dependency locking

ThanHenderson
authored and
GitHub Enterprise
committed

Commits on Apr 15, 2025

Refine LoadPacked test case

ThanHenderson
authored and
GitHub Enterprise
committed

Commits on Apr 16, 2025

Better guard WMMA tests with more specific capabilities

jaopaulolc
authored and
GitHub Enterprise
committed

Commits on Apr 17, 2025

Properly guard with HasWMMA_F16_ACC

ThanHenderson
authored and
GitHub Enterprise
committed
Fix hazard observer for VCC/SGPR reads

awhittle3
authored and
GitHub Enterprise
committed
Fix codecov skip_zip argument

awhittle3
authored and
GitHub Enterprise
committed
Fix unit tests that set env vars

awhittle3
authored and
GitHub Enterprise
committed

Commits on Apr 18, 2025

Move Prefetch graph transform to a separate file.

liu-yiqian
authored and
GitHub Enterprise
committed

Commits on Apr 22, 2025

Prefer SGPR carry for addition when possible

awhittle3
authored and
GitHub Enterprise
committed
Restore isLoadLDSForExchange logic in AddPrefetch.
memmett
committed
Change F8F6F4 CT to account for wave32

jaopaulolc
authored and
GitHub Enterprise
committed

Commits on Apr 23, 2025

Remove _deps from build dir in performance job

jaopaulolc
authored and
GitHub Enterprise
committed
Fix bug in AliasDataFlowTags related to the order in which we add aliases.

sdquiring
authored and
GitHub Enterprise
committed

Commits on Apr 24, 2025

Add PiecewiseAffineJoin edge and ConnectWorkgroup helper routines.

memmett
authored and
GitHub Enterprise
committed
Enable merging of scale loads across K unrolls in the swizzle pass.

pjots
authored and
GitHub Enterprise
committed
Hypergraph breadth first visit upstream

awhittle3
authored and
GitHub Enterprise
committed

Commits on Apr 25, 2025

Replace kernel args

awhittle3
authored and
GitHub Enterprise
committed
Make msgpack dependency private

bnemanich
authored and
GitHub Enterprise
committed
Consolidate Convert SubExpressions

awhittle3
authored and
GitHub Enterprise
committed
Add address sanitizer to rocRoller builds

bnemanich
authored and
GitHub Enterprise
committed
Use set in cmake for sanitize

bnemanich
authored and
GitHub Enterprise
committed

Commits on Apr 30, 2025