wouter/d4r
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Validate gfx1201 WMMA layout and compile first K kernel

Browse Source
This commit is contained in:
Codex committed 2026-09-29 19:50:50 +03:00
1 parent 64669c2c36
commit 6760294f60
10 files changed
+404 -10
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

CMakeLists.txt
+32 -1
View File
@@ -21,7 +21,28 @@ add_custom_command(OUTPUT "${probe_object}"
-o "${probe_object}"
DEPENDS tools/windows/probe_kernel.hip VERBATIM)
add_custom_target(d4r_probe_code_object ALL DEPENDS "${probe_object}")
foreach(probe hip_gfx1201_probe cuda_driver_probe d3d12_hip_interop_probe hip_stream_lifecycle_probe ngx_cuda_init_probe)
set(wmma_object "${CMAKE_CURRENT_BINARY_DIR}/wmma_gfx1201.hsaco")
add_custom_command(OUTPUT "${wmma_object}"
COMMAND "${D4R_HIP_CLANG}" -x hip --offload-arch=gfx1201
--offload-device-only --no-gpu-bundle-output -mno-wavefrontsize64
-nogpuinc -nogpulib -O2 "${CMAKE_CURRENT_SOURCE_DIR}/tools/windows/wmma_gfx1201.hip"
-o "${wmma_object}"
DEPENDS tools/windows/wmma_gfx1201.hip kernels/common/wmma_backend.h VERBATIM)
add_custom_target(d4r_wmma_code_object ALL DEPENDS "${wmma_object}")
# Compile-only K port smoke test. This object is intentionally not installed as
# a native override until its transformer outputs pass the K reference/replay.
set(k_enc1_object "${CMAKE_CURRENT_BINARY_DIR}/dltss_pwin_enc1_layer_gfx1201.hsaco")
add_custom_command(OUTPUT "${k_enc1_object}"
COMMAND "${D4R_HIP_CLANG}" -x hip -std=c++17 --offload-arch=gfx1201
--offload-device-only --no-gpu-bundle-output -mno-wavefrontsize64
-nogpuinc -nogpulib -O2 -DD4R_DEVICE_ONLY_MINIMAL
"${CMAKE_CURRENT_SOURCE_DIR}/kernels/k/dltss_pwin_enc1_layer.hip"
-o "${k_enc1_object}"
DEPENDS kernels/k/dltss_pwin_enc1_layer.hip kernels/k/pwin_layer.h
kernels/k/pwin_common.h kernels/common/wmma_backend.h
kernels/common/hip_device_minimal.h VERBATIM)
add_custom_target(d4r_k_gfx1201_compile ALL DEPENDS "${k_enc1_object}")
foreach(probe hip_gfx1201_probe gfx12_wmma_probe k_module_probe cuda_driver_probe d3d12_hip_interop_probe hip_stream_lifecycle_probe ngx_cuda_init_probe)
add_executable(d4r_${probe} tools/windows/${probe}.cpp)
target_compile_definitions(d4r_${probe} PRIVATE WIN32_LEAN_AND_MEAN NOMINMAX __HIP_PLATFORM_AMD__)
target_include_directories(d4r_${probe} PRIVATE "${D4R_HIP_ROOT}/include")
@@ -72,13 +93,23 @@ endif()
install(TARGETS d4r_nvapi_compat RUNTIME DESTINATION nvapi-compat)
target_link_libraries(d4r_d3d12_hip_interop_probe PRIVATE d3d12 dxgi dxguid)
add_dependencies(d4r_hip_gfx1201_probe d4r_probe_code_object)
add_dependencies(d4r_gfx12_wmma_probe d4r_wmma_code_object)
add_dependencies(d4r_k_module_probe d4r_k_gfx1201_compile)
install(FILES "${probe_object}" DESTINATION bin)
install(FILES "${wmma_object}" DESTINATION bin)
install(FILES "${k_enc1_object}" DESTINATION experimental/k)
install(FILES scripts/windows/test-windows-rdna4.ps1 DESTINATION .)
install(FILES docs/windows-rdna4-port.md DESTINATION docs)
include(CTest)
add_test(NAME hip_gfx1201 COMMAND d4r_hip_gfx1201_probe
--hip-root "${D4R_HIP_ROOT}" --module "${probe_object}")
set_tests_properties(hip_gfx1201 PROPERTIES TIMEOUT 90 LABELS "hardware;gfx1201")
add_test(NAME gfx12_wmma COMMAND d4r_gfx12_wmma_probe
--hip-root "${D4R_HIP_ROOT}" --module "${wmma_object}" --iterations 16)
set_tests_properties(gfx12_wmma PROPERTIES TIMEOUT 90 LABELS "hardware;gfx1201;wmma")
add_test(NAME k_enc1_module COMMAND d4r_k_module_probe
--hip-root "${D4R_HIP_ROOT}" --module "${k_enc1_object}")
set_tests_properties(k_enc1_module PROPERTIES TIMEOUT 90 LABELS "hardware;gfx1201;k_compile")
add_test(NAME d3d12_hip_interop COMMAND d4r_d3d12_hip_interop_probe
--hip-root "${D4R_HIP_ROOT}" --module "${probe_object}" --iterations 32)
set_tests_properties(d3d12_hip_interop PROPERTIES TIMEOUT 120 LABELS "hardware;interop")