Index: Tensile-rocm-5.0.1/pytest.ini
===================================================================
--- Tensile-rocm-5.0.1.orig/pytest.ini
+++ Tensile-rocm-5.0.1/pytest.ini
@@ -87,6 +87,7 @@ markers =
  xfail-gfx1011: architecture
  xfail-gfx1012: architecture
  xfail-gfx1030: architecture
+ xfail-gfx1031: architecture
  skip-gfx000:  architecture
  skip-gfx900:  architecture
  skip-gfx906:  architecture
@@ -96,3 +97,4 @@ markers =
  skip-gfx1011: architecture
  skip-gfx1012: architecture
  skip-gfx1030: architecture
+ skip-gfx1031: architecture
Index: Tensile-rocm-5.0.1/Tensile/Common.py
===================================================================
--- Tensile-rocm-5.0.1.orig/Tensile/Common.py
+++ Tensile-rocm-5.0.1/Tensile/Common.py
@@ -198,7 +198,7 @@ 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)] # 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)] # 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.
@@ -264,7 +264,7 @@ 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'
 }
 
 def getArchitectureName(gfxName):
Index: Tensile-rocm-5.0.1/Tensile/KernelWriterAssembly.py
===================================================================
--- Tensile-rocm-5.0.1.orig/Tensile/KernelWriterAssembly.py
+++ Tensile-rocm-5.0.1/Tensile/KernelWriterAssembly.py
@@ -2364,7 +2364,7 @@ class KernelWriterAssembly(KernelWriter)
 
     # integer i8x4
     elif kernel["ProblemType"]["DataType"].isInt8x4():
-      if self.version == (9,0,6) or self.version == (9,0,8) or self.version == (9,0,10) or self.version == (10,3,0):
+      if self.version == (9,0,6) or self.version == (9,0,8) or self.version == (9,0,10) or self.version == (10,3,0) or self.version == (10,3,1):
         for b in range(0, kernel["ThreadTile1"]):
           for a in range(0, kernel["ThreadTile0"]):
             for iui in range(0, innerUnroll):
Index: Tensile-rocm-5.0.1/Tensile/Source/CMakeLists.txt
===================================================================
--- Tensile-rocm-5.0.1.orig/Tensile/Source/CMakeLists.txt
+++ Tensile-rocm-5.0.1/Tensile/Source/CMakeLists.txt
@@ -47,7 +47,7 @@ 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 CACHE STRING "GPU architectures")
+  set(TENSILE_GPU_ARCHS gfx803 gfx900 gfx906:xnack- gfx908:xnack- gfx90a:xnack- gfx1010 gfx1011 gfx1012 gfx1030 gfx1031 CACHE STRING "GPU architectures")
 else()
   set(TENSILE_GPU_ARCHS gfx803 gfx900 gfx906 gfx908 gfx90a CACHE STRING "GPU architectures")
 endif()
Index: Tensile-rocm-5.0.1/Tensile/Source/lib/include/Tensile/AMDGPU.hpp
===================================================================
--- Tensile-rocm-5.0.1.orig/Tensile/Source/lib/include/Tensile/AMDGPU.hpp
+++ Tensile-rocm-5.0.1/Tensile/Source/lib/include/Tensile/AMDGPU.hpp
@@ -62,7 +62,8 @@ namespace Tensile
             gfx1010 = 1010,
             gfx1011 = 1011,
             gfx1012 = 1012,
-            gfx1030 = 1030
+            gfx1030 = 1030,
+            gfx1031 = 1031
         };
 
         AMDGPU();
Index: Tensile-rocm-5.0.1/Tensile/Source/lib/include/Tensile/Serialization/Predicates.hpp
===================================================================
--- Tensile-rocm-5.0.1.orig/Tensile/Source/lib/include/Tensile/Serialization/Predicates.hpp
+++ Tensile-rocm-5.0.1/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);
             }
         };
     } // namespace Serialization
Index: Tensile-rocm-5.0.1/Tensile/Source/lib/source/AMDGPU.cpp
===================================================================
--- Tensile-rocm-5.0.1.orig/Tensile/Source/lib/source/AMDGPU.cpp
+++ Tensile-rocm-5.0.1/Tensile/Source/lib/source/AMDGPU.cpp
@@ -82,6 +82,8 @@ namespace Tensile
             return stream << "gfx1012";
         case AMDGPU::Processor::gfx1030:
             return stream << "gfx1030";
+        case AMDGPU::Processor::gfx1031:
+            return stream << "gfx1031";
         }
         return stream;
     }