Index: Tensile-rocm-5.3.3/pytest.ini =================================================================== --- Tensile-rocm-5.3.3.orig/pytest.ini +++ Tensile-rocm-5.3.3/pytest.ini @@ -88,6 +88,7 @@ markers = xfail-gfx1011: architecture xfail-gfx1012: architecture xfail-gfx1030: architecture + xfail-gfx1031: architecture xfail-gfx1100: architecture xfail-gfx1101: architecture xfail-gfx1102: architecture @@ -100,6 +101,7 @@ markers = skip-gfx1011: architecture skip-gfx1012: architecture skip-gfx1030: architecture + skip-gfx1031: architecture skip-gfx1100: architecture skip-gfx1101: architecture skip-gfx1102: architecture Index: Tensile-rocm-5.3.3/Tensile/Common.py =================================================================== --- Tensile-rocm-5.3.3.orig/Tensile/Common.py +++ Tensile-rocm-5.3.3/Tensile/Common.py @@ -201,7 +201,7 @@ globalParameters["MergeFiles"] = True globalParameters["NumMergedFiles"] = 1 # The number of files that kernels should be split between when merging globalParameters["MaxFileName"] = 64 # If a file name would be longer than this, shorten it with a hash. -globalParameters["SupportedISA"] = [(8,0,3), (9,0,0), (9,0,6), (9,0,8), (9,0,10), (10,1,0), (10,1,1), (10,1,2), (10,3,0), (11,0,0), (11,0,1), (11,0,2)] # assembly kernels writer supports these architectures +globalParameters["SupportedISA"] = [(8,0,3), (9,0,0), (9,0,6), (9,0,8), (9,0,10), (10,1,0), (10,1,1), (10,1,2), (10,3,0), (10,3,1), (11,0,0), (11,0,1), (11,0,2)] # assembly kernels writer supports these architectures globalParameters["GenerateManifestAndExit"] = False # Output manifest file with list of expected library objects and exit globalParameters["NewClient"] = 2 # Old client deprecated: NewClient must be set to 2. @@ -271,7 +271,7 @@ architectureMap = { 'gfx906':'vega20', 'gfx906:xnack+':'vega20', 'gfx906:xnack-':'vega20', 'gfx908':'arcturus','gfx908:xnack+':'arcturus', 'gfx908:xnack-':'arcturus', 'gfx90a':'aldebaran', 'gfx90a:xnack+':'aldebaran', 'gfx90a:xnack-':'aldebaran', - 'gfx1010':'navi10', 'gfx1011':'navi12', 'gfx1012':'navi14', 'gfx1030':'navi21', + 'gfx1010':'navi10', 'gfx1011':'navi12', 'gfx1012':'navi14', 'gfx1030':'navi21', 'gfx1031':'navi22', 'gfx1100':'navi31', 'gfx1101':'navi32', 'gfx1102':'navi33' } Index: Tensile-rocm-5.3.3/Tensile/Source/CMakeLists.txt =================================================================== --- Tensile-rocm-5.3.3.orig/Tensile/Source/CMakeLists.txt +++ Tensile-rocm-5.3.3/Tensile/Source/CMakeLists.txt @@ -51,7 +51,7 @@ if(NOT DEFINED CXX_VERSION_STRING) endif() if(CMAKE_CXX_COMPILER STREQUAL "hipcc") - set(TENSILE_GPU_ARCHS gfx803 gfx900 gfx906:xnack- gfx908:xnack- gfx90a:xnack- gfx1010 gfx1011 gfx1012 gfx1030 gfx1100 gfx1101 gfx1102 CACHE STRING "GPU architectures") + set(TENSILE_GPU_ARCHS gfx803 gfx900 gfx906:xnack- gfx908:xnack- gfx90a:xnack- gfx1010 gfx1011 gfx1012 gfx1030 gfx1031 gfx1100 gfx1101 gfx1102 CACHE STRING "GPU architectures") else() set(TENSILE_GPU_ARCHS gfx803 gfx900 gfx906 gfx908 gfx90a gfx1010 gfx1011 gfx1012 gfx1030 gfx1100 gfx1101 gfx1102 CACHE STRING "GPU architectures") endif() Index: Tensile-rocm-5.3.3/Tensile/Source/lib/include/Tensile/AMDGPU.hpp =================================================================== --- Tensile-rocm-5.3.3.orig/Tensile/Source/lib/include/Tensile/AMDGPU.hpp +++ Tensile-rocm-5.3.3/Tensile/Source/lib/include/Tensile/AMDGPU.hpp @@ -63,6 +63,7 @@ namespace Tensile gfx1011 = 1011, gfx1012 = 1012, gfx1030 = 1030, + gfx1031 = 1031, gfx1100 = 1100, gfx1101 = 1101, gfx1102 = 1102 @@ -90,6 +91,8 @@ namespace Tensile return "gfx1012"; case AMDGPU::Processor::gfx1030: return "gfx1030"; + case AMDGPU::Processor::gfx1031: + return "gfx1032"; case AMDGPU::Processor::gfx1100: return "gfx1100"; case AMDGPU::Processor::gfx1101: Index: Tensile-rocm-5.3.3/Tensile/Source/lib/include/Tensile/Serialization/Predicates.hpp =================================================================== --- Tensile-rocm-5.3.3.orig/Tensile/Source/lib/include/Tensile/Serialization/Predicates.hpp +++ Tensile-rocm-5.3.3/Tensile/Source/lib/include/Tensile/Serialization/Predicates.hpp @@ -215,6 +215,7 @@ namespace Tensile iot::enumCase(io, value, "gfx1011", AMDGPU::Processor::gfx1011); iot::enumCase(io, value, "gfx1012", AMDGPU::Processor::gfx1012); iot::enumCase(io, value, "gfx1030", AMDGPU::Processor::gfx1030); + iot::enumCase(io, value, "gfx1031", AMDGPU::Processor::gfx1031); iot::enumCase(io, value, "gfx1100", AMDGPU::Processor::gfx1100); iot::enumCase(io, value, "gfx1101", AMDGPU::Processor::gfx1101); iot::enumCase(io, value, "gfx1102", AMDGPU::Processor::gfx1102);