diff --git a/Jenkinsfile b/Jenkinsfile
index 2b73508da0d3..78071fde4599 100755
--- a/Jenkinsfile
+++ b/Jenkinsfile
@@ -45,7 +45,7 @@
 // 'python3 jenkins/generate.py'
 // Note: This timestamp is here to ensure that updates to the Jenkinsfile are
 // always rebased on main before merging:
-// Generated at 2022-08-30T15:26:50.100067
+// Generated at 2022-09-01T11:52:42.195970
 
 import org.jenkinsci.plugins.pipeline.modeldefinition.Utils
 // NOTE: these lines are scanned by docker/dev_common.sh. Please update the regex as needed. -->
@@ -1538,7 +1538,7 @@ def shard_run_unittest_GPU_3_of_3() {
 }
 
 
-def shard_run_integration_CPU_1_of_10() {
+def shard_run_integration_CPU_1_of_4() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
@@ -1549,7 +1549,7 @@ def shard_run_integration_CPU_1_of_10() {
             withEnv([
               'PLATFORM=cpu',
               'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
+              'TVM_NUM_SHARDS=4',
               'TVM_SHARD_INDEX=0',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -1610,11 +1610,11 @@ def shard_run_integration_CPU_1_of_10() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: CPU 1 of 10')
+    Utils.markStageSkippedForConditional('integration: CPU 1 of 4')
   }
 }
 
-def shard_run_integration_CPU_2_of_10() {
+def shard_run_integration_CPU_2_of_4() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
@@ -1625,7 +1625,7 @@ def shard_run_integration_CPU_2_of_10() {
             withEnv([
               'PLATFORM=cpu',
               'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
+              'TVM_NUM_SHARDS=4',
               'TVM_SHARD_INDEX=1',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -1686,11 +1686,11 @@ def shard_run_integration_CPU_2_of_10() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: CPU 2 of 10')
+    Utils.markStageSkippedForConditional('integration: CPU 2 of 4')
   }
 }
 
-def shard_run_integration_CPU_3_of_10() {
+def shard_run_integration_CPU_3_of_4() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
@@ -1701,7 +1701,7 @@ def shard_run_integration_CPU_3_of_10() {
             withEnv([
               'PLATFORM=cpu',
               'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
+              'TVM_NUM_SHARDS=4',
               'TVM_SHARD_INDEX=2',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -1762,11 +1762,11 @@ def shard_run_integration_CPU_3_of_10() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: CPU 3 of 10')
+    Utils.markStageSkippedForConditional('integration: CPU 3 of 4')
   }
 }
 
-def shard_run_integration_CPU_4_of_10() {
+def shard_run_integration_CPU_4_of_4() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
@@ -1777,7 +1777,7 @@ def shard_run_integration_CPU_4_of_10() {
             withEnv([
               'PLATFORM=cpu',
               'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
+              'TVM_NUM_SHARDS=4',
               'TVM_SHARD_INDEX=3',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -1838,327 +1838,24 @@ def shard_run_integration_CPU_4_of_10() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: CPU 4 of 10')
+    Utils.markStageSkippedForConditional('integration: CPU 4 of 4')
   }
 }
 
-def shard_run_integration_CPU_5_of_10() {
-  if (!skip_ci && is_docs_only_build != 1) {
-    node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
-        try {
-          docker_init(ci_cpu)
-          init_git()
-          timeout(time: max_time, unit: 'MINUTES') {
-            withEnv([
-              'PLATFORM=cpu',
-              'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
-              'TVM_SHARD_INDEX=4',
-              "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
-              sh(
-                        script: """
-                          set -eux
-                          retry() {
-                            local max_retries=\$1
-                            shift
-                            local n=0
-                            local backoff_max=30
-                            until [ "\$n" -ge \$max_retries ]
-                            do
-                                "\$@" && break
-                                n=\$((n+1))
-                                if [ "\$n" -eq \$max_retries ]; then
-                                    echo "failed to update after attempt \$n / \$max_retries, giving up"
-                                    exit 1
-                                fi
-
-                                WAIT=\$(python3 -c 'import random; print(random.randint(10, 30))')
-                                echo "failed to update \$n / \$max_retries, waiting \$WAIT to try again"
-                                sleep \$WAIT
-                            done
-                          }
-
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_tsim.so build/libvta_tsim.so
-                          md5sum build/libvta_tsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm.so build/libtvm.so
-                          md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm_runtime.so build/libtvm_runtime.so
-                          md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/config.cmake build/config.cmake
-                          md5sum build/config.cmake
-                        """,
-                        label: 'Download artifacts from S3',
-                      )
-
-              ci_setup(ci_cpu)
-              sh (
-                script: "${docker_run} ${ci_cpu} ./tests/scripts/task_python_integration.sh",
-                label: 'Run CPU integration tests',
-              )
-            })
-          }
-        } finally {
-          sh(
-            script: """
-              set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_CPU --recursive
-            """,
-            label: 'Upload JUnits to S3',
-          )
-
-          junit 'build/pytest-results/*.xml'
-        }
-      }
-    }
-  } else {
-    Utils.markStageSkippedForConditional('integration: CPU 5 of 10')
-  }
-}
 
-def shard_run_integration_CPU_6_of_10() {
+def shard_run_python_i386_1_of_3() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
-        try {
-          docker_init(ci_cpu)
-          init_git()
-          timeout(time: max_time, unit: 'MINUTES') {
-            withEnv([
-              'PLATFORM=cpu',
-              'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
-              'TVM_SHARD_INDEX=5',
-              "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
-              sh(
-                        script: """
-                          set -eux
-                          retry() {
-                            local max_retries=\$1
-                            shift
-                            local n=0
-                            local backoff_max=30
-                            until [ "\$n" -ge \$max_retries ]
-                            do
-                                "\$@" && break
-                                n=\$((n+1))
-                                if [ "\$n" -eq \$max_retries ]; then
-                                    echo "failed to update after attempt \$n / \$max_retries, giving up"
-                                    exit 1
-                                fi
-
-                                WAIT=\$(python3 -c 'import random; print(random.randint(10, 30))')
-                                echo "failed to update \$n / \$max_retries, waiting \$WAIT to try again"
-                                sleep \$WAIT
-                            done
-                          }
-
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_tsim.so build/libvta_tsim.so
-                          md5sum build/libvta_tsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm.so build/libtvm.so
-                          md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm_runtime.so build/libtvm_runtime.so
-                          md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/config.cmake build/config.cmake
-                          md5sum build/config.cmake
-                        """,
-                        label: 'Download artifacts from S3',
-                      )
-
-              ci_setup(ci_cpu)
-              sh (
-                script: "${docker_run} ${ci_cpu} ./tests/scripts/task_python_integration.sh",
-                label: 'Run CPU integration tests',
-              )
-            })
-          }
-        } finally {
-          sh(
-            script: """
-              set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_CPU --recursive
-            """,
-            label: 'Upload JUnits to S3',
-          )
-
-          junit 'build/pytest-results/*.xml'
-        }
-      }
-    }
-  } else {
-    Utils.markStageSkippedForConditional('integration: CPU 6 of 10')
-  }
-}
-
-def shard_run_integration_CPU_7_of_10() {
-  if (!skip_ci && is_docs_only_build != 1) {
-    node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
-        try {
-          docker_init(ci_cpu)
-          init_git()
-          timeout(time: max_time, unit: 'MINUTES') {
-            withEnv([
-              'PLATFORM=cpu',
-              'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
-              'TVM_SHARD_INDEX=6',
-              "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
-              sh(
-                        script: """
-                          set -eux
-                          retry() {
-                            local max_retries=\$1
-                            shift
-                            local n=0
-                            local backoff_max=30
-                            until [ "\$n" -ge \$max_retries ]
-                            do
-                                "\$@" && break
-                                n=\$((n+1))
-                                if [ "\$n" -eq \$max_retries ]; then
-                                    echo "failed to update after attempt \$n / \$max_retries, giving up"
-                                    exit 1
-                                fi
-
-                                WAIT=\$(python3 -c 'import random; print(random.randint(10, 30))')
-                                echo "failed to update \$n / \$max_retries, waiting \$WAIT to try again"
-                                sleep \$WAIT
-                            done
-                          }
-
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_tsim.so build/libvta_tsim.so
-                          md5sum build/libvta_tsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm.so build/libtvm.so
-                          md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm_runtime.so build/libtvm_runtime.so
-                          md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/config.cmake build/config.cmake
-                          md5sum build/config.cmake
-                        """,
-                        label: 'Download artifacts from S3',
-                      )
-
-              ci_setup(ci_cpu)
-              sh (
-                script: "${docker_run} ${ci_cpu} ./tests/scripts/task_python_integration.sh",
-                label: 'Run CPU integration tests',
-              )
-            })
-          }
-        } finally {
-          sh(
-            script: """
-              set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_CPU --recursive
-            """,
-            label: 'Upload JUnits to S3',
-          )
-
-          junit 'build/pytest-results/*.xml'
-        }
-      }
-    }
-  } else {
-    Utils.markStageSkippedForConditional('integration: CPU 7 of 10')
-  }
-}
-
-def shard_run_integration_CPU_8_of_10() {
-  if (!skip_ci && is_docs_only_build != 1) {
-    node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
-        try {
-          docker_init(ci_cpu)
-          init_git()
-          timeout(time: max_time, unit: 'MINUTES') {
-            withEnv([
-              'PLATFORM=cpu',
-              'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
-              'TVM_SHARD_INDEX=7',
-              "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
-              sh(
-                        script: """
-                          set -eux
-                          retry() {
-                            local max_retries=\$1
-                            shift
-                            local n=0
-                            local backoff_max=30
-                            until [ "\$n" -ge \$max_retries ]
-                            do
-                                "\$@" && break
-                                n=\$((n+1))
-                                if [ "\$n" -eq \$max_retries ]; then
-                                    echo "failed to update after attempt \$n / \$max_retries, giving up"
-                                    exit 1
-                                fi
-
-                                WAIT=\$(python3 -c 'import random; print(random.randint(10, 30))')
-                                echo "failed to update \$n / \$max_retries, waiting \$WAIT to try again"
-                                sleep \$WAIT
-                            done
-                          }
-
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_tsim.so build/libvta_tsim.so
-                          md5sum build/libvta_tsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm.so build/libtvm.so
-                          md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm_runtime.so build/libtvm_runtime.so
-                          md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/config.cmake build/config.cmake
-                          md5sum build/config.cmake
-                        """,
-                        label: 'Download artifacts from S3',
-                      )
-
-              ci_setup(ci_cpu)
-              sh (
-                script: "${docker_run} ${ci_cpu} ./tests/scripts/task_python_integration.sh",
-                label: 'Run CPU integration tests',
-              )
-            })
-          }
-        } finally {
-          sh(
-            script: """
-              set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_CPU --recursive
-            """,
-            label: 'Upload JUnits to S3',
-          )
-
-          junit 'build/pytest-results/*.xml'
-        }
-      }
-    }
-  } else {
-    Utils.markStageSkippedForConditional('integration: CPU 8 of 10')
-  }
-}
-
-def shard_run_integration_CPU_9_of_10() {
-  if (!skip_ci && is_docs_only_build != 1) {
-    node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-i386") {
         try {
-          docker_init(ci_cpu)
+          docker_init(ci_i386)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=cpu',
-              'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
-              'TVM_SHARD_INDEX=8',
+              'PLATFORM=i386',
+              'TEST_STEP_NAME=python: i386',
+              'TVM_NUM_SHARDS=3',
+              'TVM_SHARD_INDEX=0',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2183,24 +1880,24 @@ def shard_run_integration_CPU_9_of_10() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_tsim.so build/libvta_tsim.so
-                          md5sum build/libvta_tsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_fsim.so build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libvta_fsim.so build/libvta_fsim.so
                           md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_cpu)
+              ci_setup(ci_i386)
+              cpp_unittest(ci_i386)
+              python_unittest(ci_i386)
               sh (
-                script: "${docker_run} ${ci_cpu} ./tests/scripts/task_python_integration.sh",
-                label: 'Run CPU integration tests',
+                script: "${docker_run} ${ci_i386} ./tests/scripts/task_python_integration_i386only.sh",
+                label: 'Run i386 integration tests',
               )
             })
           }
@@ -2208,7 +1905,7 @@ def shard_run_integration_CPU_9_of_10() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_CPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/python_i386 --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -2218,23 +1915,23 @@ def shard_run_integration_CPU_9_of_10() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: CPU 9 of 10')
+    Utils.markStageSkippedForConditional('python: i386 1 of 3')
   }
 }
 
-def shard_run_integration_CPU_10_of_10() {
+def shard_run_python_i386_2_of_3() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-cpu") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-i386") {
         try {
-          docker_init(ci_cpu)
+          docker_init(ci_i386)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=cpu',
-              'TEST_STEP_NAME=integration: CPU',
-              'TVM_NUM_SHARDS=10',
-              'TVM_SHARD_INDEX=9',
+              'PLATFORM=i386',
+              'TEST_STEP_NAME=python: i386',
+              'TVM_NUM_SHARDS=3',
+              'TVM_SHARD_INDEX=1',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2259,32 +1956,32 @@ def shard_run_integration_CPU_10_of_10() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_tsim.so build/libvta_tsim.so
-                          md5sum build/libvta_tsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libvta_fsim.so build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libvta_fsim.so build/libvta_fsim.so
                           md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cpu/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_cpu)
+              ci_setup(ci_i386)
+              python_unittest(ci_i386)
               sh (
-                script: "${docker_run} ${ci_cpu} ./tests/scripts/task_python_integration.sh",
-                label: 'Run CPU integration tests',
+                script: "${docker_run} ${ci_i386} ./tests/scripts/task_python_integration_i386only.sh",
+                label: 'Run i386 integration tests',
               )
+              fsim_test(ci_i386)
             })
           }
         } finally {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_CPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/python_i386 --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -2294,12 +1991,11 @@ def shard_run_integration_CPU_10_of_10() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: CPU 10 of 10')
+    Utils.markStageSkippedForConditional('python: i386 2 of 3')
   }
 }
 
-
-def shard_run_python_i386_1_of_5() {
+def shard_run_python_i386_3_of_3() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-i386") {
@@ -2310,8 +2006,8 @@ def shard_run_python_i386_1_of_5() {
             withEnv([
               'PLATFORM=i386',
               'TEST_STEP_NAME=python: i386',
-              'TVM_NUM_SHARDS=5',
-              'TVM_SHARD_INDEX=0',
+              'TVM_NUM_SHARDS=3',
+              'TVM_SHARD_INDEX=2',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2349,7 +2045,6 @@ def shard_run_python_i386_1_of_5() {
                       )
 
               ci_setup(ci_i386)
-              cpp_unittest(ci_i386)
               python_unittest(ci_i386)
               sh (
                 script: "${docker_run} ${ci_i386} ./tests/scripts/task_python_integration_i386only.sh",
@@ -2371,23 +2066,24 @@ def shard_run_python_i386_1_of_5() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('python: i386 1 of 5')
+    Utils.markStageSkippedForConditional('python: i386 3 of 3')
   }
 }
 
-def shard_run_python_i386_2_of_5() {
+
+def shard_run_test_Hexagon_1_of_8() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-i386") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
         try {
-          docker_init(ci_i386)
+          docker_init(ci_hexagon)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=i386',
-              'TEST_STEP_NAME=python: i386',
-              'TVM_NUM_SHARDS=5',
-              'TVM_SHARD_INDEX=1',
+              'PLATFORM=hexagon',
+              'TEST_STEP_NAME=test: Hexagon',
+              'TVM_NUM_SHARDS=8',
+              'TVM_SHARD_INDEX=0',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2412,32 +2108,31 @@ def shard_run_python_i386_2_of_5() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/hexagon_api_output build/hexagon_api_output --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_i386)
-              python_unittest(ci_i386)
+              add_hexagon_permissions()
+              ci_setup(ci_hexagon)
+              cpp_unittest(ci_hexagon)
               sh (
-                script: "${docker_run} ${ci_i386} ./tests/scripts/task_python_integration_i386only.sh",
-                label: 'Run i386 integration tests',
+                script: "${docker_run} ${ci_hexagon} ./tests/scripts/task_python_hexagon.sh",
+                label: 'Run Hexagon tests',
               )
-              fsim_test(ci_i386)
             })
           }
         } finally {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/python_i386 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Hexagon --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -2447,23 +2142,23 @@ def shard_run_python_i386_2_of_5() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('python: i386 2 of 5')
+    Utils.markStageSkippedForConditional('test: Hexagon 1 of 8')
   }
 }
 
-def shard_run_python_i386_3_of_5() {
+def shard_run_test_Hexagon_2_of_8() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-i386") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
         try {
-          docker_init(ci_i386)
+          docker_init(ci_hexagon)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=i386',
-              'TEST_STEP_NAME=python: i386',
-              'TVM_NUM_SHARDS=5',
-              'TVM_SHARD_INDEX=2',
+              'PLATFORM=hexagon',
+              'TEST_STEP_NAME=test: Hexagon',
+              'TVM_NUM_SHARDS=8',
+              'TVM_SHARD_INDEX=1',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2488,23 +2183,22 @@ def shard_run_python_i386_3_of_5() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/hexagon_api_output build/hexagon_api_output --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_i386)
-              python_unittest(ci_i386)
+              add_hexagon_permissions()
+              ci_setup(ci_hexagon)
               sh (
-                script: "${docker_run} ${ci_i386} ./tests/scripts/task_python_integration_i386only.sh",
-                label: 'Run i386 integration tests',
+                script: "${docker_run} ${ci_hexagon} ./tests/scripts/task_python_hexagon.sh",
+                label: 'Run Hexagon tests',
               )
             })
           }
@@ -2512,7 +2206,7 @@ def shard_run_python_i386_3_of_5() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/python_i386 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Hexagon --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -2522,23 +2216,23 @@ def shard_run_python_i386_3_of_5() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('python: i386 3 of 5')
+    Utils.markStageSkippedForConditional('test: Hexagon 2 of 8')
   }
 }
 
-def shard_run_python_i386_4_of_5() {
+def shard_run_test_Hexagon_3_of_8() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-i386") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
         try {
-          docker_init(ci_i386)
+          docker_init(ci_hexagon)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=i386',
-              'TEST_STEP_NAME=python: i386',
-              'TVM_NUM_SHARDS=5',
-              'TVM_SHARD_INDEX=3',
+              'PLATFORM=hexagon',
+              'TEST_STEP_NAME=test: Hexagon',
+              'TVM_NUM_SHARDS=8',
+              'TVM_SHARD_INDEX=2',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2563,23 +2257,22 @@ def shard_run_python_i386_4_of_5() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/hexagon_api_output build/hexagon_api_output --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_i386)
-              python_unittest(ci_i386)
+              add_hexagon_permissions()
+              ci_setup(ci_hexagon)
               sh (
-                script: "${docker_run} ${ci_i386} ./tests/scripts/task_python_integration_i386only.sh",
-                label: 'Run i386 integration tests',
+                script: "${docker_run} ${ci_hexagon} ./tests/scripts/task_python_hexagon.sh",
+                label: 'Run Hexagon tests',
               )
             })
           }
@@ -2587,7 +2280,7 @@ def shard_run_python_i386_4_of_5() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/python_i386 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Hexagon --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -2597,23 +2290,23 @@ def shard_run_python_i386_4_of_5() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('python: i386 4 of 5')
+    Utils.markStageSkippedForConditional('test: Hexagon 3 of 8')
   }
 }
 
-def shard_run_python_i386_5_of_5() {
+def shard_run_test_Hexagon_4_of_8() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/integration-python-i386") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
         try {
-          docker_init(ci_i386)
+          docker_init(ci_hexagon)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=i386',
-              'TEST_STEP_NAME=python: i386',
-              'TVM_NUM_SHARDS=5',
-              'TVM_SHARD_INDEX=4',
+              'PLATFORM=hexagon',
+              'TEST_STEP_NAME=test: Hexagon',
+              'TVM_NUM_SHARDS=8',
+              'TVM_SHARD_INDEX=3',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2638,23 +2331,22 @@ def shard_run_python_i386_5_of_5() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/i386/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/hexagon_api_output build/hexagon_api_output --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_i386)
-              python_unittest(ci_i386)
+              add_hexagon_permissions()
+              ci_setup(ci_hexagon)
               sh (
-                script: "${docker_run} ${ci_i386} ./tests/scripts/task_python_integration_i386only.sh",
-                label: 'Run i386 integration tests',
+                script: "${docker_run} ${ci_hexagon} ./tests/scripts/task_python_hexagon.sh",
+                label: 'Run Hexagon tests',
               )
             })
           }
@@ -2662,7 +2354,7 @@ def shard_run_python_i386_5_of_5() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/python_i386 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Hexagon --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -2672,12 +2364,11 @@ def shard_run_python_i386_5_of_5() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('python: i386 5 of 5')
+    Utils.markStageSkippedForConditional('test: Hexagon 4 of 8')
   }
 }
 
-
-def shard_run_test_Hexagon_1_of_7() {
+def shard_run_test_Hexagon_5_of_8() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
@@ -2688,8 +2379,8 @@ def shard_run_test_Hexagon_1_of_7() {
             withEnv([
               'PLATFORM=hexagon',
               'TEST_STEP_NAME=test: Hexagon',
-              'TVM_NUM_SHARDS=7',
-              'TVM_SHARD_INDEX=0',
+              'TVM_NUM_SHARDS=8',
+              'TVM_SHARD_INDEX=4',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2727,7 +2418,6 @@ def shard_run_test_Hexagon_1_of_7() {
 
               add_hexagon_permissions()
               ci_setup(ci_hexagon)
-              cpp_unittest(ci_hexagon)
               sh (
                 script: "${docker_run} ${ci_hexagon} ./tests/scripts/task_python_hexagon.sh",
                 label: 'Run Hexagon tests',
@@ -2748,11 +2438,11 @@ def shard_run_test_Hexagon_1_of_7() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Hexagon 1 of 7')
+    Utils.markStageSkippedForConditional('test: Hexagon 5 of 8')
   }
 }
 
-def shard_run_test_Hexagon_2_of_7() {
+def shard_run_test_Hexagon_6_of_8() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
@@ -2763,8 +2453,8 @@ def shard_run_test_Hexagon_2_of_7() {
             withEnv([
               'PLATFORM=hexagon',
               'TEST_STEP_NAME=test: Hexagon',
-              'TVM_NUM_SHARDS=7',
-              'TVM_SHARD_INDEX=1',
+              'TVM_NUM_SHARDS=8',
+              'TVM_SHARD_INDEX=5',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2822,11 +2512,11 @@ def shard_run_test_Hexagon_2_of_7() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Hexagon 2 of 7')
+    Utils.markStageSkippedForConditional('test: Hexagon 6 of 8')
   }
 }
 
-def shard_run_test_Hexagon_3_of_7() {
+def shard_run_test_Hexagon_7_of_8() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
@@ -2837,8 +2527,8 @@ def shard_run_test_Hexagon_3_of_7() {
             withEnv([
               'PLATFORM=hexagon',
               'TEST_STEP_NAME=test: Hexagon',
-              'TVM_NUM_SHARDS=7',
-              'TVM_SHARD_INDEX=2',
+              'TVM_NUM_SHARDS=8',
+              'TVM_SHARD_INDEX=6',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2896,11 +2586,11 @@ def shard_run_test_Hexagon_3_of_7() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Hexagon 3 of 7')
+    Utils.markStageSkippedForConditional('test: Hexagon 7 of 8')
   }
 }
 
-def shard_run_test_Hexagon_4_of_7() {
+def shard_run_test_Hexagon_8_of_8() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
@@ -2911,8 +2601,8 @@ def shard_run_test_Hexagon_4_of_7() {
             withEnv([
               'PLATFORM=hexagon',
               'TEST_STEP_NAME=test: Hexagon',
-              'TVM_NUM_SHARDS=7',
-              'TVM_SHARD_INDEX=3',
+              'TVM_NUM_SHARDS=8',
+              'TVM_SHARD_INDEX=7',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -2970,23 +2660,24 @@ def shard_run_test_Hexagon_4_of_7() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Hexagon 4 of 7')
+    Utils.markStageSkippedForConditional('test: Hexagon 8 of 8')
   }
 }
 
-def shard_run_test_Hexagon_5_of_7() {
+
+def shard_run_integration_aarch64_1_of_4() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
+    node('ARM-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
         try {
-          docker_init(ci_hexagon)
+          docker_init(ci_arm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=hexagon',
-              'TEST_STEP_NAME=test: Hexagon',
-              'TVM_NUM_SHARDS=7',
-              'TVM_SHARD_INDEX=4',
+              'PLATFORM=arm',
+              'TEST_STEP_NAME=integration: aarch64',
+              'TVM_NUM_SHARDS=4',
+              'TVM_SHARD_INDEX=0',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3011,22 +2702,23 @@ def shard_run_test_Hexagon_5_of_7() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
+                          md5sum build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/hexagon_api_output build/hexagon_api_output --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              add_hexagon_permissions()
-              ci_setup(ci_hexagon)
+              ci_setup(ci_arm)
+              python_unittest(ci_arm)
               sh (
-                script: "${docker_run} ${ci_hexagon} ./tests/scripts/task_python_hexagon.sh",
-                label: 'Run Hexagon tests',
+                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_integration.sh",
+                label: 'Run CPU integration tests',
               )
             })
           }
@@ -3034,7 +2726,7 @@ def shard_run_test_Hexagon_5_of_7() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Hexagon --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_aarch64 --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3044,23 +2736,23 @@ def shard_run_test_Hexagon_5_of_7() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Hexagon 5 of 7')
+    Utils.markStageSkippedForConditional('integration: aarch64 1 of 4')
   }
 }
 
-def shard_run_test_Hexagon_6_of_7() {
+def shard_run_integration_aarch64_2_of_4() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
+    node('ARM-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
         try {
-          docker_init(ci_hexagon)
+          docker_init(ci_arm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=hexagon',
-              'TEST_STEP_NAME=test: Hexagon',
-              'TVM_NUM_SHARDS=7',
-              'TVM_SHARD_INDEX=5',
+              'PLATFORM=arm',
+              'TEST_STEP_NAME=integration: aarch64',
+              'TVM_NUM_SHARDS=4',
+              'TVM_SHARD_INDEX=1',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3085,22 +2777,23 @@ def shard_run_test_Hexagon_6_of_7() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
+                          md5sum build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/hexagon_api_output build/hexagon_api_output --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              add_hexagon_permissions()
-              ci_setup(ci_hexagon)
+              ci_setup(ci_arm)
+              python_unittest(ci_arm)
               sh (
-                script: "${docker_run} ${ci_hexagon} ./tests/scripts/task_python_hexagon.sh",
-                label: 'Run Hexagon tests',
+                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_integration.sh",
+                label: 'Run CPU integration tests',
               )
             })
           }
@@ -3108,7 +2801,7 @@ def shard_run_test_Hexagon_6_of_7() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Hexagon --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_aarch64 --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3118,23 +2811,23 @@ def shard_run_test_Hexagon_6_of_7() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Hexagon 6 of 7')
+    Utils.markStageSkippedForConditional('integration: aarch64 2 of 4')
   }
 }
 
-def shard_run_test_Hexagon_7_of_7() {
+def shard_run_integration_aarch64_3_of_4() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('CPU-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-hexagon") {
+    node('ARM-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
         try {
-          docker_init(ci_hexagon)
+          docker_init(ci_arm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=hexagon',
-              'TEST_STEP_NAME=test: Hexagon',
-              'TVM_NUM_SHARDS=7',
-              'TVM_SHARD_INDEX=6',
+              'PLATFORM=arm',
+              'TEST_STEP_NAME=integration: aarch64',
+              'TVM_NUM_SHARDS=4',
+              'TVM_SHARD_INDEX=2',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3159,22 +2852,23 @@ def shard_run_test_Hexagon_7_of_7() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
+                          md5sum build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/hexagon/build/hexagon_api_output build/hexagon_api_output --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              add_hexagon_permissions()
-              ci_setup(ci_hexagon)
+              ci_setup(ci_arm)
+              python_unittest(ci_arm)
               sh (
-                script: "${docker_run} ${ci_hexagon} ./tests/scripts/task_python_hexagon.sh",
-                label: 'Run Hexagon tests',
+                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_integration.sh",
+                label: 'Run CPU integration tests',
               )
             })
           }
@@ -3182,7 +2876,7 @@ def shard_run_test_Hexagon_7_of_7() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Hexagon --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_aarch64 --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3192,12 +2886,11 @@ def shard_run_test_Hexagon_7_of_7() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Hexagon 7 of 7')
+    Utils.markStageSkippedForConditional('integration: aarch64 3 of 4')
   }
 }
 
-
-def shard_run_integration_aarch64_1_of_4() {
+def shard_run_integration_aarch64_4_of_4() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('ARM-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
@@ -3209,7 +2902,7 @@ def shard_run_integration_aarch64_1_of_4() {
               'PLATFORM=arm',
               'TEST_STEP_NAME=integration: aarch64',
               'TVM_NUM_SHARDS=4',
-              'TVM_SHARD_INDEX=0',
+              'TVM_SHARD_INDEX=3',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3268,23 +2961,24 @@ def shard_run_integration_aarch64_1_of_4() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: aarch64 1 of 4')
+    Utils.markStageSkippedForConditional('integration: aarch64 4 of 4')
   }
 }
 
-def shard_run_integration_aarch64_2_of_4() {
+
+def shard_run_topi_GPU_1_of_3() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('ARM-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
+    node('GPU') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/topi-python-gpu") {
         try {
-          docker_init(ci_arm)
+          docker_init(ci_gpu)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=arm',
-              'TEST_STEP_NAME=integration: aarch64',
-              'TVM_NUM_SHARDS=4',
-              'TVM_SHARD_INDEX=1',
+              'PLATFORM=gpu',
+              'TEST_STEP_NAME=topi: GPU',
+              'TVM_NUM_SHARDS=3',
+              'TVM_SHARD_INDEX=0',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3309,23 +3003,22 @@ def shard_run_integration_aarch64_2_of_4() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libvta_fsim.so build/libvta_fsim.so
                           md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_arm)
-              python_unittest(ci_arm)
+              ci_setup(ci_gpu)
               sh (
-                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_integration.sh",
-                label: 'Run CPU integration tests',
+                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_topi.sh",
+                label: 'Run TOPI tests',
               )
             })
           }
@@ -3333,7 +3026,7 @@ def shard_run_integration_aarch64_2_of_4() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_aarch64 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_GPU --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3343,23 +3036,23 @@ def shard_run_integration_aarch64_2_of_4() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: aarch64 2 of 4')
+    Utils.markStageSkippedForConditional('topi: GPU 1 of 3')
   }
 }
 
-def shard_run_integration_aarch64_3_of_4() {
+def shard_run_topi_GPU_2_of_3() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('ARM-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
+    node('GPU') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/topi-python-gpu") {
         try {
-          docker_init(ci_arm)
+          docker_init(ci_gpu)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=arm',
-              'TEST_STEP_NAME=integration: aarch64',
-              'TVM_NUM_SHARDS=4',
-              'TVM_SHARD_INDEX=2',
+              'PLATFORM=gpu',
+              'TEST_STEP_NAME=topi: GPU',
+              'TVM_NUM_SHARDS=3',
+              'TVM_SHARD_INDEX=1',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3384,23 +3077,22 @@ def shard_run_integration_aarch64_3_of_4() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libvta_fsim.so build/libvta_fsim.so
                           md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_arm)
-              python_unittest(ci_arm)
+              ci_setup(ci_gpu)
               sh (
-                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_integration.sh",
-                label: 'Run CPU integration tests',
+                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_topi.sh",
+                label: 'Run TOPI tests',
               )
             })
           }
@@ -3408,7 +3100,7 @@ def shard_run_integration_aarch64_3_of_4() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_aarch64 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_GPU --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3418,23 +3110,23 @@ def shard_run_integration_aarch64_3_of_4() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: aarch64 3 of 4')
+    Utils.markStageSkippedForConditional('topi: GPU 2 of 3')
   }
 }
 
-def shard_run_integration_aarch64_4_of_4() {
+def shard_run_topi_GPU_3_of_3() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('ARM-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
+    node('GPU') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/topi-python-gpu") {
         try {
-          docker_init(ci_arm)
+          docker_init(ci_gpu)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=arm',
-              'TEST_STEP_NAME=integration: aarch64',
-              'TVM_NUM_SHARDS=4',
-              'TVM_SHARD_INDEX=3',
+              'PLATFORM=gpu',
+              'TEST_STEP_NAME=topi: GPU',
+              'TVM_NUM_SHARDS=3',
+              'TVM_SHARD_INDEX=2',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3459,23 +3151,22 @@ def shard_run_integration_aarch64_4_of_4() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libvta_fsim.so build/libvta_fsim.so
                           md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_arm)
-              python_unittest(ci_arm)
+              ci_setup(ci_gpu)
               sh (
-                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_integration.sh",
-                label: 'Run CPU integration tests',
+                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_topi.sh",
+                label: 'Run TOPI tests',
               )
             })
           }
@@ -3483,7 +3174,7 @@ def shard_run_integration_aarch64_4_of_4() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/integration_aarch64 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_GPU --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3493,23 +3184,23 @@ def shard_run_integration_aarch64_4_of_4() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('integration: aarch64 4 of 4')
+    Utils.markStageSkippedForConditional('topi: GPU 3 of 3')
   }
 }
 
 
-def shard_run_topi_GPU_1_of_4() {
+def shard_run_frontend_GPU_1_of_6() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('GPU') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/topi-python-gpu") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
         try {
           docker_init(ci_gpu)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
               'PLATFORM=gpu',
-              'TEST_STEP_NAME=topi: GPU',
-              'TVM_NUM_SHARDS=4',
+              'TEST_STEP_NAME=frontend: GPU',
+              'TVM_NUM_SHARDS=6',
               'TVM_SHARD_INDEX=0',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -3549,8 +3240,8 @@ def shard_run_topi_GPU_1_of_4() {
 
               ci_setup(ci_gpu)
               sh (
-                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_topi.sh",
-                label: 'Run TOPI tests',
+                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_frontend.sh",
+                label: 'Run Python frontend tests',
               )
             })
           }
@@ -3558,7 +3249,7 @@ def shard_run_topi_GPU_1_of_4() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_GPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_GPU --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3568,22 +3259,22 @@ def shard_run_topi_GPU_1_of_4() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('topi: GPU 1 of 4')
+    Utils.markStageSkippedForConditional('frontend: GPU 1 of 6')
   }
 }
 
-def shard_run_topi_GPU_2_of_4() {
+def shard_run_frontend_GPU_2_of_6() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('GPU') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/topi-python-gpu") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
         try {
           docker_init(ci_gpu)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
               'PLATFORM=gpu',
-              'TEST_STEP_NAME=topi: GPU',
-              'TVM_NUM_SHARDS=4',
+              'TEST_STEP_NAME=frontend: GPU',
+              'TVM_NUM_SHARDS=6',
               'TVM_SHARD_INDEX=1',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -3623,8 +3314,8 @@ def shard_run_topi_GPU_2_of_4() {
 
               ci_setup(ci_gpu)
               sh (
-                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_topi.sh",
-                label: 'Run TOPI tests',
+                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_frontend.sh",
+                label: 'Run Python frontend tests',
               )
             })
           }
@@ -3632,7 +3323,7 @@ def shard_run_topi_GPU_2_of_4() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_GPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_GPU --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3642,22 +3333,22 @@ def shard_run_topi_GPU_2_of_4() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('topi: GPU 2 of 4')
+    Utils.markStageSkippedForConditional('frontend: GPU 2 of 6')
   }
 }
 
-def shard_run_topi_GPU_3_of_4() {
+def shard_run_frontend_GPU_3_of_6() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('GPU') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/topi-python-gpu") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
         try {
           docker_init(ci_gpu)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
               'PLATFORM=gpu',
-              'TEST_STEP_NAME=topi: GPU',
-              'TVM_NUM_SHARDS=4',
+              'TEST_STEP_NAME=frontend: GPU',
+              'TVM_NUM_SHARDS=6',
               'TVM_SHARD_INDEX=2',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -3697,8 +3388,8 @@ def shard_run_topi_GPU_3_of_4() {
 
               ci_setup(ci_gpu)
               sh (
-                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_topi.sh",
-                label: 'Run TOPI tests',
+                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_frontend.sh",
+                label: 'Run Python frontend tests',
               )
             })
           }
@@ -3706,7 +3397,7 @@ def shard_run_topi_GPU_3_of_4() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_GPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_GPU --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3716,22 +3407,22 @@ def shard_run_topi_GPU_3_of_4() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('topi: GPU 3 of 4')
+    Utils.markStageSkippedForConditional('frontend: GPU 3 of 6')
   }
 }
 
-def shard_run_topi_GPU_4_of_4() {
+def shard_run_frontend_GPU_4_of_6() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('GPU') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/topi-python-gpu") {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
         try {
           docker_init(ci_gpu)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
               'PLATFORM=gpu',
-              'TEST_STEP_NAME=topi: GPU',
-              'TVM_NUM_SHARDS=4',
+              'TEST_STEP_NAME=frontend: GPU',
+              'TVM_NUM_SHARDS=6',
               'TVM_SHARD_INDEX=3',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -3771,8 +3462,8 @@ def shard_run_topi_GPU_4_of_4() {
 
               ci_setup(ci_gpu)
               sh (
-                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_topi.sh",
-                label: 'Run TOPI tests',
+                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_frontend.sh",
+                label: 'Run Python frontend tests',
               )
             })
           }
@@ -3780,7 +3471,7 @@ def shard_run_topi_GPU_4_of_4() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_GPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_GPU --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -3790,12 +3481,11 @@ def shard_run_topi_GPU_4_of_4() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('topi: GPU 4 of 4')
+    Utils.markStageSkippedForConditional('frontend: GPU 4 of 6')
   }
 }
 
-
-def shard_run_frontend_GPU_1_of_6() {
+def shard_run_frontend_GPU_5_of_6() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('GPU') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
@@ -3807,7 +3497,7 @@ def shard_run_frontend_GPU_1_of_6() {
               'PLATFORM=gpu',
               'TEST_STEP_NAME=frontend: GPU',
               'TVM_NUM_SHARDS=6',
-              'TVM_SHARD_INDEX=0',
+              'TVM_SHARD_INDEX=4',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3865,11 +3555,11 @@ def shard_run_frontend_GPU_1_of_6() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('frontend: GPU 1 of 6')
+    Utils.markStageSkippedForConditional('frontend: GPU 5 of 6')
   }
 }
 
-def shard_run_frontend_GPU_2_of_6() {
+def shard_run_frontend_GPU_6_of_6() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('GPU') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
@@ -3881,7 +3571,7 @@ def shard_run_frontend_GPU_2_of_6() {
               'PLATFORM=gpu',
               'TEST_STEP_NAME=frontend: GPU',
               'TVM_NUM_SHARDS=6',
-              'TVM_SHARD_INDEX=1',
+              'TVM_SHARD_INDEX=5',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3939,23 +3629,24 @@ def shard_run_frontend_GPU_2_of_6() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('frontend: GPU 2 of 6')
+    Utils.markStageSkippedForConditional('frontend: GPU 6 of 6')
   }
 }
 
-def shard_run_frontend_GPU_3_of_6() {
+
+def shard_run_topi_aarch64_1_of_2() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('GPU') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
+    node('ARM-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
         try {
-          docker_init(ci_gpu)
+          docker_init(ci_arm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=gpu',
-              'TEST_STEP_NAME=frontend: GPU',
-              'TVM_NUM_SHARDS=6',
-              'TVM_SHARD_INDEX=2',
+              'PLATFORM=arm',
+              'TEST_STEP_NAME=topi: aarch64',
+              'TVM_NUM_SHARDS=2',
+              'TVM_SHARD_INDEX=0',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -3980,22 +3671,27 @@ def shard_run_frontend_GPU_3_of_6() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libvta_fsim.so build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
                           md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_gpu)
+              ci_setup(ci_arm)
+              cpp_unittest(ci_arm)
               sh (
-                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_frontend.sh",
-                label: 'Run Python frontend tests',
+                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_arm_compute_library.sh",
+                label: 'Run test_arm_compute_lib test',
+              )
+              sh (
+                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_topi.sh",
+                label: 'Run TOPI tests',
               )
             })
           }
@@ -4003,7 +3699,7 @@ def shard_run_frontend_GPU_3_of_6() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_GPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_aarch64 --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -4013,23 +3709,23 @@ def shard_run_frontend_GPU_3_of_6() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('frontend: GPU 3 of 6')
+    Utils.markStageSkippedForConditional('topi: aarch64 1 of 2')
   }
 }
 
-def shard_run_frontend_GPU_4_of_6() {
+def shard_run_topi_aarch64_2_of_2() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('GPU') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
+    node('ARM-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
         try {
-          docker_init(ci_gpu)
+          docker_init(ci_arm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=gpu',
-              'TEST_STEP_NAME=frontend: GPU',
-              'TVM_NUM_SHARDS=6',
-              'TVM_SHARD_INDEX=3',
+              'PLATFORM=arm',
+              'TEST_STEP_NAME=topi: aarch64',
+              'TVM_NUM_SHARDS=2',
+              'TVM_SHARD_INDEX=1',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4054,22 +3750,26 @@ def shard_run_frontend_GPU_4_of_6() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libvta_fsim.so build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
                           md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_gpu)
+              ci_setup(ci_arm)
+              sh (
+                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_arm_compute_library.sh",
+                label: 'Run test_arm_compute_lib test',
+              )
               sh (
-                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_frontend.sh",
-                label: 'Run Python frontend tests',
+                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_topi.sh",
+                label: 'Run TOPI tests',
               )
             })
           }
@@ -4077,7 +3777,7 @@ def shard_run_frontend_GPU_4_of_6() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_GPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_aarch64 --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -4087,23 +3787,24 @@ def shard_run_frontend_GPU_4_of_6() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('frontend: GPU 4 of 6')
+    Utils.markStageSkippedForConditional('topi: aarch64 2 of 2')
   }
 }
 
-def shard_run_frontend_GPU_5_of_6() {
+
+def shard_run_frontend_aarch64_1_of_2() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('GPU') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
+    node('ARM-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-arm") {
         try {
-          docker_init(ci_gpu)
+          docker_init(ci_arm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=gpu',
-              'TEST_STEP_NAME=frontend: GPU',
-              'TVM_NUM_SHARDS=6',
-              'TVM_SHARD_INDEX=4',
+              'PLATFORM=arm',
+              'TEST_STEP_NAME=frontend: aarch64',
+              'TVM_NUM_SHARDS=2',
+              'TVM_SHARD_INDEX=0',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4128,21 +3829,21 @@ def shard_run_frontend_GPU_5_of_6() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libvta_fsim.so build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
                           md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_gpu)
+              ci_setup(ci_arm)
               sh (
-                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_frontend.sh",
+                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_frontend_cpu.sh",
                 label: 'Run Python frontend tests',
               )
             })
@@ -4151,7 +3852,7 @@ def shard_run_frontend_GPU_5_of_6() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_GPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_aarch64 --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -4161,23 +3862,23 @@ def shard_run_frontend_GPU_5_of_6() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('frontend: GPU 5 of 6')
+    Utils.markStageSkippedForConditional('frontend: aarch64 1 of 2')
   }
 }
 
-def shard_run_frontend_GPU_6_of_6() {
+def shard_run_frontend_aarch64_2_of_2() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('GPU') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-gpu") {
+    node('ARM-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-arm") {
         try {
-          docker_init(ci_gpu)
+          docker_init(ci_arm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=gpu',
-              'TEST_STEP_NAME=frontend: GPU',
-              'TVM_NUM_SHARDS=6',
-              'TVM_SHARD_INDEX=5',
+              'PLATFORM=arm',
+              'TEST_STEP_NAME=frontend: aarch64',
+              'TVM_NUM_SHARDS=2',
+              'TVM_SHARD_INDEX=1',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4202,21 +3903,21 @@ def shard_run_frontend_GPU_6_of_6() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libvta_fsim.so build/libvta_fsim.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
                           md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/gpu/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_gpu)
+              ci_setup(ci_arm)
               sh (
-                script: "${docker_run} ${ci_gpu} ./tests/scripts/task_python_frontend.sh",
+                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_frontend_cpu.sh",
                 label: 'Run Python frontend tests',
               )
             })
@@ -4225,7 +3926,7 @@ def shard_run_frontend_GPU_6_of_6() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_GPU --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_aarch64 --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -4235,23 +3936,23 @@ def shard_run_frontend_GPU_6_of_6() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('frontend: GPU 6 of 6')
+    Utils.markStageSkippedForConditional('frontend: aarch64 2 of 2')
   }
 }
 
 
-def shard_run_topi_aarch64_1_of_2() {
+def shard_run_test_Cortex_M_1_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('ARM-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
+    node('CPU-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
         try {
-          docker_init(ci_arm)
+          docker_init(ci_cortexm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=arm',
-              'TEST_STEP_NAME=topi: aarch64',
-              'TVM_NUM_SHARDS=2',
+              'PLATFORM=cortexm',
+              'TEST_STEP_NAME=test: Cortex-M',
+              'TVM_NUM_SHARDS=12',
               'TVM_SHARD_INDEX=0',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -4277,27 +3978,27 @@ def shard_run_topi_aarch64_1_of_2() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/microtvm_template_projects build/microtvm_template_projects --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_arm)
-              cpp_unittest(ci_arm)
+              add_microtvm_permissions()
+              ci_setup(ci_cortexm)
+              cpp_unittest(ci_cortexm)
               sh (
-                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_arm_compute_library.sh",
-                label: 'Run test_arm_compute_lib test',
+                script: "${docker_run} ${ci_cortexm} ./tests/scripts/task_demo_microtvm.sh",
+                label: 'Run microTVM demos',
               )
               sh (
-                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_topi.sh",
-                label: 'Run TOPI tests',
+                script: "${docker_run} ${ci_cortexm} ./tests/scripts/task_python_microtvm.sh",
+                label: 'Run microTVM tests',
               )
             })
           }
@@ -4305,7 +4006,7 @@ def shard_run_topi_aarch64_1_of_2() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_aarch64 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Cortex_M --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -4315,22 +4016,22 @@ def shard_run_topi_aarch64_1_of_2() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('topi: aarch64 1 of 2')
+    Utils.markStageSkippedForConditional('test: Cortex-M 1 of 12')
   }
 }
 
-def shard_run_topi_aarch64_2_of_2() {
+def shard_run_test_Cortex_M_2_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('ARM-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/ut-python-arm") {
+    node('CPU-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
         try {
-          docker_init(ci_arm)
+          docker_init(ci_cortexm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=arm',
-              'TEST_STEP_NAME=topi: aarch64',
-              'TVM_NUM_SHARDS=2',
+              'PLATFORM=cortexm',
+              'TEST_STEP_NAME=test: Cortex-M',
+              'TVM_NUM_SHARDS=12',
               'TVM_SHARD_INDEX=1',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
@@ -4356,26 +4057,22 @@ def shard_run_topi_aarch64_2_of_2() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/microtvm_template_projects build/microtvm_template_projects --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_arm)
-              sh (
-                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_arm_compute_library.sh",
-                label: 'Run test_arm_compute_lib test',
-              )
+              add_microtvm_permissions()
+              ci_setup(ci_cortexm)
               sh (
-                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_topi.sh",
-                label: 'Run TOPI tests',
+                script: "${docker_run} ${ci_cortexm} ./tests/scripts/task_python_microtvm.sh",
+                label: 'Run microTVM tests',
               )
             })
           }
@@ -4383,7 +4080,7 @@ def shard_run_topi_aarch64_2_of_2() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/topi_aarch64 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Cortex_M --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -4393,24 +4090,23 @@ def shard_run_topi_aarch64_2_of_2() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('topi: aarch64 2 of 2')
+    Utils.markStageSkippedForConditional('test: Cortex-M 2 of 12')
   }
 }
 
-
-def shard_run_frontend_aarch64_1_of_2() {
+def shard_run_test_Cortex_M_3_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('ARM-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-arm") {
+    node('CPU-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
         try {
-          docker_init(ci_arm)
+          docker_init(ci_cortexm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=arm',
-              'TEST_STEP_NAME=frontend: aarch64',
-              'TVM_NUM_SHARDS=2',
-              'TVM_SHARD_INDEX=0',
+              'PLATFORM=cortexm',
+              'TEST_STEP_NAME=test: Cortex-M',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=2',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4435,22 +4131,22 @@ def shard_run_frontend_aarch64_1_of_2() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/microtvm_template_projects build/microtvm_template_projects --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_arm)
+              add_microtvm_permissions()
+              ci_setup(ci_cortexm)
               sh (
-                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_frontend_cpu.sh",
-                label: 'Run Python frontend tests',
+                script: "${docker_run} ${ci_cortexm} ./tests/scripts/task_python_microtvm.sh",
+                label: 'Run microTVM tests',
               )
             })
           }
@@ -4458,7 +4154,7 @@ def shard_run_frontend_aarch64_1_of_2() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_aarch64 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Cortex_M --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -4468,23 +4164,23 @@ def shard_run_frontend_aarch64_1_of_2() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('frontend: aarch64 1 of 2')
+    Utils.markStageSkippedForConditional('test: Cortex-M 3 of 12')
   }
 }
 
-def shard_run_frontend_aarch64_2_of_2() {
+def shard_run_test_Cortex_M_4_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
-    node('ARM-SMALL') {
-      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/frontend-python-arm") {
+    node('CPU-SMALL') {
+      ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
         try {
-          docker_init(ci_arm)
+          docker_init(ci_cortexm)
           init_git()
           timeout(time: max_time, unit: 'MINUTES') {
             withEnv([
-              'PLATFORM=arm',
-              'TEST_STEP_NAME=frontend: aarch64',
-              'TVM_NUM_SHARDS=2',
-              'TVM_SHARD_INDEX=1',
+              'PLATFORM=cortexm',
+              'TEST_STEP_NAME=test: Cortex-M',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=3',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4509,22 +4205,22 @@ def shard_run_frontend_aarch64_2_of_2() {
                             done
                           }
 
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm.so build/libtvm.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/libtvm.so build/libtvm.so
                           md5sum build/libtvm.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libvta_fsim.so build/libvta_fsim.so
-                          md5sum build/libvta_fsim.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/libtvm_runtime.so build/libtvm_runtime.so
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/libtvm_runtime.so build/libtvm_runtime.so
                           md5sum build/libtvm_runtime.so
-                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/arm/build/config.cmake build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/config.cmake build/config.cmake
                           md5sum build/config.cmake
+                          retry 3 aws s3 cp --no-progress s3://${s3_prefix}/cortexm/build/microtvm_template_projects build/microtvm_template_projects --recursive
                         """,
                         label: 'Download artifacts from S3',
                       )
 
-              ci_setup(ci_arm)
+              add_microtvm_permissions()
+              ci_setup(ci_cortexm)
               sh (
-                script: "${docker_run} ${ci_arm} ./tests/scripts/task_python_frontend_cpu.sh",
-                label: 'Run Python frontend tests',
+                script: "${docker_run} ${ci_cortexm} ./tests/scripts/task_python_microtvm.sh",
+                label: 'Run microTVM tests',
               )
             })
           }
@@ -4532,7 +4228,7 @@ def shard_run_frontend_aarch64_2_of_2() {
           sh(
             script: """
               set -eux
-              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/frontend_aarch64 --recursive
+              aws s3 cp --no-progress build/pytest-results s3://${s3_prefix}/pytest-results/test_Cortex_M --recursive
             """,
             label: 'Upload JUnits to S3',
           )
@@ -4542,12 +4238,11 @@ def shard_run_frontend_aarch64_2_of_2() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('frontend: aarch64 2 of 2')
+    Utils.markStageSkippedForConditional('test: Cortex-M 4 of 12')
   }
 }
 
-
-def shard_run_test_Cortex_M_1_of_8() {
+def shard_run_test_Cortex_M_5_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
@@ -4558,8 +4253,8 @@ def shard_run_test_Cortex_M_1_of_8() {
             withEnv([
               'PLATFORM=cortexm',
               'TEST_STEP_NAME=test: Cortex-M',
-              'TVM_NUM_SHARDS=8',
-              'TVM_SHARD_INDEX=0',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=4',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4597,11 +4292,6 @@ def shard_run_test_Cortex_M_1_of_8() {
 
               add_microtvm_permissions()
               ci_setup(ci_cortexm)
-              cpp_unittest(ci_cortexm)
-              sh (
-                script: "${docker_run} ${ci_cortexm} ./tests/scripts/task_demo_microtvm.sh",
-                label: 'Run microTVM demos',
-              )
               sh (
                 script: "${docker_run} ${ci_cortexm} ./tests/scripts/task_python_microtvm.sh",
                 label: 'Run microTVM tests',
@@ -4622,11 +4312,11 @@ def shard_run_test_Cortex_M_1_of_8() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Cortex-M 1 of 8')
+    Utils.markStageSkippedForConditional('test: Cortex-M 5 of 12')
   }
 }
 
-def shard_run_test_Cortex_M_2_of_8() {
+def shard_run_test_Cortex_M_6_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
@@ -4637,8 +4327,8 @@ def shard_run_test_Cortex_M_2_of_8() {
             withEnv([
               'PLATFORM=cortexm',
               'TEST_STEP_NAME=test: Cortex-M',
-              'TVM_NUM_SHARDS=8',
-              'TVM_SHARD_INDEX=1',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=5',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4696,11 +4386,11 @@ def shard_run_test_Cortex_M_2_of_8() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Cortex-M 2 of 8')
+    Utils.markStageSkippedForConditional('test: Cortex-M 6 of 12')
   }
 }
 
-def shard_run_test_Cortex_M_3_of_8() {
+def shard_run_test_Cortex_M_7_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
@@ -4711,8 +4401,8 @@ def shard_run_test_Cortex_M_3_of_8() {
             withEnv([
               'PLATFORM=cortexm',
               'TEST_STEP_NAME=test: Cortex-M',
-              'TVM_NUM_SHARDS=8',
-              'TVM_SHARD_INDEX=2',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=6',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4770,11 +4460,11 @@ def shard_run_test_Cortex_M_3_of_8() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Cortex-M 3 of 8')
+    Utils.markStageSkippedForConditional('test: Cortex-M 7 of 12')
   }
 }
 
-def shard_run_test_Cortex_M_4_of_8() {
+def shard_run_test_Cortex_M_8_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
@@ -4785,8 +4475,8 @@ def shard_run_test_Cortex_M_4_of_8() {
             withEnv([
               'PLATFORM=cortexm',
               'TEST_STEP_NAME=test: Cortex-M',
-              'TVM_NUM_SHARDS=8',
-              'TVM_SHARD_INDEX=3',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=7',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4844,11 +4534,11 @@ def shard_run_test_Cortex_M_4_of_8() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Cortex-M 4 of 8')
+    Utils.markStageSkippedForConditional('test: Cortex-M 8 of 12')
   }
 }
 
-def shard_run_test_Cortex_M_5_of_8() {
+def shard_run_test_Cortex_M_9_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
@@ -4859,8 +4549,8 @@ def shard_run_test_Cortex_M_5_of_8() {
             withEnv([
               'PLATFORM=cortexm',
               'TEST_STEP_NAME=test: Cortex-M',
-              'TVM_NUM_SHARDS=8',
-              'TVM_SHARD_INDEX=4',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=8',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4918,11 +4608,11 @@ def shard_run_test_Cortex_M_5_of_8() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Cortex-M 5 of 8')
+    Utils.markStageSkippedForConditional('test: Cortex-M 9 of 12')
   }
 }
 
-def shard_run_test_Cortex_M_6_of_8() {
+def shard_run_test_Cortex_M_10_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
@@ -4933,8 +4623,8 @@ def shard_run_test_Cortex_M_6_of_8() {
             withEnv([
               'PLATFORM=cortexm',
               'TEST_STEP_NAME=test: Cortex-M',
-              'TVM_NUM_SHARDS=8',
-              'TVM_SHARD_INDEX=5',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=9',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -4992,11 +4682,11 @@ def shard_run_test_Cortex_M_6_of_8() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Cortex-M 6 of 8')
+    Utils.markStageSkippedForConditional('test: Cortex-M 10 of 12')
   }
 }
 
-def shard_run_test_Cortex_M_7_of_8() {
+def shard_run_test_Cortex_M_11_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
@@ -5007,8 +4697,8 @@ def shard_run_test_Cortex_M_7_of_8() {
             withEnv([
               'PLATFORM=cortexm',
               'TEST_STEP_NAME=test: Cortex-M',
-              'TVM_NUM_SHARDS=8',
-              'TVM_SHARD_INDEX=6',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=10',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -5066,11 +4756,11 @@ def shard_run_test_Cortex_M_7_of_8() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Cortex-M 7 of 8')
+    Utils.markStageSkippedForConditional('test: Cortex-M 11 of 12')
   }
 }
 
-def shard_run_test_Cortex_M_8_of_8() {
+def shard_run_test_Cortex_M_12_of_12() {
   if (!skip_ci && is_docs_only_build != 1) {
     node('CPU-SMALL') {
       ws("workspace/exec_${env.EXECUTOR_NUMBER}/tvm/test-cortexm") {
@@ -5081,8 +4771,8 @@ def shard_run_test_Cortex_M_8_of_8() {
             withEnv([
               'PLATFORM=cortexm',
               'TEST_STEP_NAME=test: Cortex-M',
-              'TVM_NUM_SHARDS=8',
-              'TVM_SHARD_INDEX=7',
+              'TVM_NUM_SHARDS=12',
+              'TVM_SHARD_INDEX=11',
               "SKIP_SLOW_TESTS=${skip_slow_tests}"], {
               sh(
                         script: """
@@ -5140,7 +4830,7 @@ def shard_run_test_Cortex_M_8_of_8() {
       }
     }
   } else {
-    Utils.markStageSkippedForConditional('test: Cortex-M 8 of 8')
+    Utils.markStageSkippedForConditional('test: Cortex-M 12 of 12')
   }
 }
 
@@ -5300,71 +4990,50 @@ stage('Test') {
   'unittest: GPU 3 of 3': {
     shard_run_unittest_GPU_3_of_3()
   },
-  'integration: CPU 1 of 10': {
-    shard_run_integration_CPU_1_of_10()
-  },
-  'integration: CPU 2 of 10': {
-    shard_run_integration_CPU_2_of_10()
-  },
-  'integration: CPU 3 of 10': {
-    shard_run_integration_CPU_3_of_10()
-  },
-  'integration: CPU 4 of 10': {
-    shard_run_integration_CPU_4_of_10()
-  },
-  'integration: CPU 5 of 10': {
-    shard_run_integration_CPU_5_of_10()
-  },
-  'integration: CPU 6 of 10': {
-    shard_run_integration_CPU_6_of_10()
+  'integration: CPU 1 of 4': {
+    shard_run_integration_CPU_1_of_4()
   },
-  'integration: CPU 7 of 10': {
-    shard_run_integration_CPU_7_of_10()
+  'integration: CPU 2 of 4': {
+    shard_run_integration_CPU_2_of_4()
   },
-  'integration: CPU 8 of 10': {
-    shard_run_integration_CPU_8_of_10()
+  'integration: CPU 3 of 4': {
+    shard_run_integration_CPU_3_of_4()
   },
-  'integration: CPU 9 of 10': {
-    shard_run_integration_CPU_9_of_10()
+  'integration: CPU 4 of 4': {
+    shard_run_integration_CPU_4_of_4()
   },
-  'integration: CPU 10 of 10': {
-    shard_run_integration_CPU_10_of_10()
+  'python: i386 1 of 3': {
+    shard_run_python_i386_1_of_3()
   },
-  'python: i386 1 of 5': {
-    shard_run_python_i386_1_of_5()
+  'python: i386 2 of 3': {
+    shard_run_python_i386_2_of_3()
   },
-  'python: i386 2 of 5': {
-    shard_run_python_i386_2_of_5()
+  'python: i386 3 of 3': {
+    shard_run_python_i386_3_of_3()
   },
-  'python: i386 3 of 5': {
-    shard_run_python_i386_3_of_5()
+  'test: Hexagon 1 of 8': {
+    shard_run_test_Hexagon_1_of_8()
   },
-  'python: i386 4 of 5': {
-    shard_run_python_i386_4_of_5()
+  'test: Hexagon 2 of 8': {
+    shard_run_test_Hexagon_2_of_8()
   },
-  'python: i386 5 of 5': {
-    shard_run_python_i386_5_of_5()
+  'test: Hexagon 3 of 8': {
+    shard_run_test_Hexagon_3_of_8()
   },
-  'test: Hexagon 1 of 7': {
-    shard_run_test_Hexagon_1_of_7()
+  'test: Hexagon 4 of 8': {
+    shard_run_test_Hexagon_4_of_8()
   },
-  'test: Hexagon 2 of 7': {
-    shard_run_test_Hexagon_2_of_7()
+  'test: Hexagon 5 of 8': {
+    shard_run_test_Hexagon_5_of_8()
   },
-  'test: Hexagon 3 of 7': {
-    shard_run_test_Hexagon_3_of_7()
+  'test: Hexagon 6 of 8': {
+    shard_run_test_Hexagon_6_of_8()
   },
-  'test: Hexagon 4 of 7': {
-    shard_run_test_Hexagon_4_of_7()
+  'test: Hexagon 7 of 8': {
+    shard_run_test_Hexagon_7_of_8()
   },
-  'test: Hexagon 5 of 7': {
-    shard_run_test_Hexagon_5_of_7()
-  },
-  'test: Hexagon 6 of 7': {
-    shard_run_test_Hexagon_6_of_7()
-  },
-  'test: Hexagon 7 of 7': {
-    shard_run_test_Hexagon_7_of_7()
+  'test: Hexagon 8 of 8': {
+    shard_run_test_Hexagon_8_of_8()
   },
   'integration: aarch64 1 of 4': {
     shard_run_integration_aarch64_1_of_4()
@@ -5378,17 +5047,14 @@ stage('Test') {
   'integration: aarch64 4 of 4': {
     shard_run_integration_aarch64_4_of_4()
   },
-  'topi: GPU 1 of 4': {
-    shard_run_topi_GPU_1_of_4()
+  'topi: GPU 1 of 3': {
+    shard_run_topi_GPU_1_of_3()
   },
-  'topi: GPU 2 of 4': {
-    shard_run_topi_GPU_2_of_4()
+  'topi: GPU 2 of 3': {
+    shard_run_topi_GPU_2_of_3()
   },
-  'topi: GPU 3 of 4': {
-    shard_run_topi_GPU_3_of_4()
-  },
-  'topi: GPU 4 of 4': {
-    shard_run_topi_GPU_4_of_4()
+  'topi: GPU 3 of 3': {
+    shard_run_topi_GPU_3_of_3()
   },
   'frontend: GPU 1 of 6': {
     shard_run_frontend_GPU_1_of_6()
@@ -5420,29 +5086,41 @@ stage('Test') {
   'frontend: aarch64 2 of 2': {
     shard_run_frontend_aarch64_2_of_2()
   },
-  'test: Cortex-M 1 of 8': {
-    shard_run_test_Cortex_M_1_of_8()
+  'test: Cortex-M 1 of 12': {
+    shard_run_test_Cortex_M_1_of_12()
+  },
+  'test: Cortex-M 2 of 12': {
+    shard_run_test_Cortex_M_2_of_12()
   },
-  'test: Cortex-M 2 of 8': {
-    shard_run_test_Cortex_M_2_of_8()
+  'test: Cortex-M 3 of 12': {
+    shard_run_test_Cortex_M_3_of_12()
   },
-  'test: Cortex-M 3 of 8': {
-    shard_run_test_Cortex_M_3_of_8()
+  'test: Cortex-M 4 of 12': {
+    shard_run_test_Cortex_M_4_of_12()
   },
-  'test: Cortex-M 4 of 8': {
-    shard_run_test_Cortex_M_4_of_8()
+  'test: Cortex-M 5 of 12': {
+    shard_run_test_Cortex_M_5_of_12()
   },
-  'test: Cortex-M 5 of 8': {
-    shard_run_test_Cortex_M_5_of_8()
+  'test: Cortex-M 6 of 12': {
+    shard_run_test_Cortex_M_6_of_12()
   },
-  'test: Cortex-M 6 of 8': {
-    shard_run_test_Cortex_M_6_of_8()
+  'test: Cortex-M 7 of 12': {
+    shard_run_test_Cortex_M_7_of_12()
   },
-  'test: Cortex-M 7 of 8': {
-    shard_run_test_Cortex_M_7_of_8()
+  'test: Cortex-M 8 of 12': {
+    shard_run_test_Cortex_M_8_of_12()
   },
-  'test: Cortex-M 8 of 8': {
-    shard_run_test_Cortex_M_8_of_8()
+  'test: Cortex-M 9 of 12': {
+    shard_run_test_Cortex_M_9_of_12()
+  },
+  'test: Cortex-M 10 of 12': {
+    shard_run_test_Cortex_M_10_of_12()
+  },
+  'test: Cortex-M 11 of 12': {
+    shard_run_test_Cortex_M_11_of_12()
+  },
+  'test: Cortex-M 12 of 12': {
+    shard_run_test_Cortex_M_12_of_12()
   },
   'test: RISC-V 1 of 1': {
     shard_run_test_RISC_V_1_of_1()
@@ -5684,7 +5362,8 @@ stage('Test') {
   },
   )
 }
-}/*
+}
+/*
 stage('Build packages') {
   parallel 'conda CPU': {
     node('CPU') {
diff --git a/ci/jenkins/Test.groovy.j2 b/ci/jenkins/Test.groovy.j2
index 9e8c9ac28b01..4ed149da9be0 100644
--- a/ci/jenkins/Test.groovy.j2
+++ b/ci/jenkins/Test.groovy.j2
@@ -42,7 +42,7 @@
 {% call(shard_index, num_shards) m.sharded_test_step(
   name="integration: CPU",
   node="CPU-SMALL",
-  num_shards=10,
+  num_shards=4,
   ws="tvm/integration-python-cpu",
   platform="cpu",
   docker_image="ci_cpu",
@@ -58,7 +58,7 @@
 {% call(shard_index, num_shards) m.sharded_test_step(
   name="python: i386",
   node="CPU-SMALL",
-  num_shards=5,
+  num_shards=3,
   ws="tvm/integration-python-i386",
   platform="i386",
   docker_image="ci_i386",
@@ -85,7 +85,7 @@
   platform="hexagon",
   docker_image="ci_hexagon",
   test_method_names=test_method_names,
-  num_shards=7,
+  num_shards=8,
 ) %}
   {{ m.download_artifacts(tag='hexagon', filenames=tvm_lib, folders=hexagon_api) }}
   add_hexagon_permissions()
@@ -118,7 +118,7 @@
 {% call(shard_index, num_shards) m.sharded_test_step(
   name="topi: GPU",
   node="GPU",
-  num_shards=4,
+  num_shards=3,
   ws="tvm/topi-python-gpu",
   platform="gpu",
   docker_image="ci_gpu",
@@ -192,7 +192,7 @@
   ws="tvm/test-cortexm",
   platform="cortexm",
   docker_image="ci_cortexm",
-  num_shards=8,
+  num_shards=12,
   test_method_names=test_method_names,
 ) %}
   {{ m.download_artifacts(tag='cortexm', filenames=tvm_lib, folders=microtvm_template_projects) }}
@@ -316,4 +316,4 @@ stage('Test') {
   },
   )
 }
-}
\ No newline at end of file
+}
diff --git a/ci/jenkins/generate.py b/ci/jenkins/generate.py
index 3ccdedc6d924..07bf4b5a8dad 100644
--- a/ci/jenkins/generate.py
+++ b/ci/jenkins/generate.py
@@ -31,6 +31,12 @@
 JENKINSFILE = REPO_ROOT / "Jenkinsfile"
 
 
+class Change:
+    IMAGES_ONLY = object()
+    NONE = object()
+    FULL = object()
+
+
 data = {
     "images": [
         {
@@ -83,7 +89,7 @@ def lines_without_generated_tag(content):
     ]
 
 
-def is_changed_images_only(lines: List[str]) -> bool:
+def change_type(lines: List[str]) -> Change:
     """
     Return True if 'line' only edits an image tag or if 'line' is not a changed
     line in a diff
@@ -101,7 +107,7 @@ def is_changed_images_only(lines: List[str]) -> bool:
 
     if len(diff_lines) == 0:
         # no changes made
-        return True
+        return Change.NONE
 
     for line in diff_lines:
         is_add = line.startswith("+")
@@ -113,7 +119,7 @@ def is_changed_images_only(lines: List[str]) -> bool:
         )
         if match is None:
             # matched a non-image line, quit early
-            return False
+            return Change.FULL
 
         if is_add:
             added_images.append(match.groups()[0])
@@ -121,7 +127,10 @@ def is_changed_images_only(lines: List[str]) -> bool:
             removed_images.append(match.groups()[0])
 
     # make sure that the added image lines match the removed image lines
-    return len(added_images) > 0 and added_images == removed_images
+    if len(added_images) > 0 and added_images == removed_images:
+        return Change.IMAGES_ONLY
+    else:
+        return Change.FULL
 
 
 if __name__ == "__main__":
@@ -156,9 +165,11 @@ def is_changed_images_only(lines: List[str]) -> bool:
             lines_without_generated_tag(content), lines_without_generated_tag(new_content)
         )
     ]
-    if not args.force and is_changed_images_only(diff):
+    change = change_type(diff)
+    if not args.force and change == Change.IMAGES_ONLY or change == Change.NONE:
+        if change != Change.NONE:
+            print("Detected only Docker-image name changes, skipping timestamp update")
         new_content = new_content.replace(data["generated_time"], original_timestamp)
-        print("Detected only Docker-image name changed, skipping timestamp update")
 
     diff = "".join(diff)
 
diff --git a/python/tvm/contrib/hexagon/pytest_plugin.py b/python/tvm/contrib/hexagon/pytest_plugin.py
index 65475d67f555..0b9f65540c34 100644
--- a/python/tvm/contrib/hexagon/pytest_plugin.py
+++ b/python/tvm/contrib/hexagon/pytest_plugin.py
@@ -210,7 +210,8 @@ def pytest_configure(config):
 def pytest_configure_node(node):
     # the master for each node fills slaveinput dictionary
     # which pytest-xdist will transfer to the subprocess
-    node.workerinput["device_adr"] = node.config.iplist.pop()
+    if node.config.iplist is not None:
+        node.workerinput["device_adr"] = node.config.iplist.pop()
 
 
 @pytest.fixture
diff --git a/tests/scripts/setup-pytest-env.sh b/tests/scripts/setup-pytest-env.sh
index 895979293122..305f626d666c 100755
--- a/tests/scripts/setup-pytest-env.sh
+++ b/tests/scripts/setup-pytest-env.sh
@@ -47,6 +47,7 @@ function cleanup() {
 trap cleanup 0
 
 function run_pytest() {
+    set -e
     local ffi_type="$1"
     shift
     local test_suite_name="$1"
@@ -74,17 +75,14 @@ function run_pytest() {
 
     suite_name="${test_suite_name}-${current_shard}-${ffi_type}"
 
-    # Some test environments don't play well with parallelism
-    DEFAULT_PARALLELISM=2
-    if [[ "${TEST_STEP_NAME:-default}" == "frontend: GPU"* ]] || [[ "${TEST_STEP_NAME:-default}" == "test: Hexagon"* ]]; then
-        DEFAULT_PARALLELISM=1
-    fi
+    DEFAULT_PARALLELISM=1
 
-    if [ ! "${extra_args[@]}" == *" -n"* ] && [! "${extra_args[@]}" == *" -dist"* ]; then
+    if [[ ! "${extra_args[*]}" == *" -n"* ]] && [[ ! "${extra_args[*]}" == *" -dist"* ]]; then
         extra_args+=("-n=$DEFAULT_PARALLELISM")
     fi
 
     exit_code=0
+    set +e
     TVM_FFI=${ffi_type} python3 -m pytest \
            -o "junit_suite_name=${suite_name}" \
            "--junit-xml=${TVM_PYTEST_RESULT_DIR}/${suite_name}.xml" \