[CK Tile] Enable hardware OOB buffer load offset trick by default (#7466)
## Summary
Enables `CK_TILE_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK`
inside `config.hpp`.
### Background
When loading from global memory with out-of-bound (OOB) check, CK Tile
must suppress invalid lanes. The previous default used a software
branch:
```cpp
// Old path (oob_conditional_check, no trick)
if(!src_thread_element_valid) { return zeros; }
return amd_buffer_load_impl(...);
```
This generates divergent control flow, the compiler emits exec-mask save/restore and per-lane comparison SALU instructions one set per buffer load that touches a padded dimension.
### Change
With the trick enabled, invalid lanes are suppressed entirely in hardware:
```cpp
// New path (trick enabled)
uint32_t shift = src_thread_element_valid ? 0 : 0x80000000;
return amd_buffer_load_impl(resource, shift + offset, 0);
```
The `0x80000000` offset overflows the buffer descriptor's declared size, causing the hardware to silently return zero for that lane - no branch, no exec mask manipulation. This matches the behavior of old CK XDL kernels, which use an unconditional load followed by a `v_cndmask` select.
### Expected impact
Eliminates ALU overhead from OOB validity branches which reduces the kernel execution time, especially for memory-bound cases.
---------
Co-authored-by: Bartłomiej Kocot <barkocot@amd.com>
[CK_TILE, CK_BUILDER] Add bwd data to CK Tile profiler (#5516)
## Motivation
We want close the performance gap between old CK and CK Tile for bwd
data convolutions. To achieve this, we need tow things
- Configurations for the old CK kernel instances such that we can map
them into CK Tile instances.
- Support in CK profiler to run the CK Tile instance with the same API
as for old CK instances.
## Technical Details
Extracted kernel configurations from old CK. The codegen python script
for CK Tile convs is extended to support also bwd data. The generated
instances are added to the CMake build (target
`device_grouped_conv_bwd_data_tile_instances`).
A new profiler op (`grouped_conv_bwd_data_tile`) has been added to the
CK Profiler. The API is same as for old CK's profiler op
`grouped_conv_bwd_data`.
---------
Co-authored-by: Ville Pietilä <>
[CK][CK Tile] Move grouped conv cpp instances to build dir (#5609)
## Motivation
Move grouped conv .cpp instances to build dir. Fix generate instances
script.
## Technical Details
Avoid CI problem when instances in experimental directory are not
removed
## Test Plan
test_grouped_convnd_*_tile
## Test Result
Pending
## Submission Checklist
- [x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
[CK_TILE] Add CK Tile bwd weight profiler (#4797)
## Motivation
To compare old CK and CK Tile, we need to extend the current CK profiler
to support running also CK Tile instance with the same API. In order to
have the same instance coverage in CK Tile compared to the old CK, I've
added code generation from old CK configurations to CK Tile instances
using the CK Builder.
## Technical Details
- The codegen python script for CK Tile fwd convs is extended to support
also bwd weight and bwd data.
- The generated instances are added to the CMake build (target
`device_grouped_conv_bwd_weight_tile_instance`s).
- A new profiler op (`grouped_conv_bwd_weight_tile`) has been added to
the CK Profiler.
---------
Co-authored-by: Ville Pietilä <>
Co-authored-by: Bartlomiej Kocot <barkocot@amd.com>
[CK][CK TILE] Add has hot loop check for pipeline v1
## Motivation
Add has hot loop check for pipeline v1 (v1 basic and v1 basic async).
Enable more tests which have been fixed by this change.
## Technical Details
Hot loop has been executed without num loop check.
## Test Plan
test_grouped_convnd_fwd_tile
## Test Result
Passed
## Submission Checklist
- [x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
AICK-651
AICK-663
