Index: clang/lib/CodeGen/BackendUtil.cpp
===================================================================
--- clang/lib/CodeGen/BackendUtil.cpp
+++ clang/lib/CodeGen/BackendUtil.cpp
@@ -74,6 +74,7 @@
 #include "llvm/Transforms/ObjCARC.h"
 #include "llvm/Transforms/Scalar.h"
 #include "llvm/Transforms/Scalar/GVN.h"
+#include "llvm/Transforms/Scalar/LowerMatrixIntrinsics.h"
 #include "llvm/Transforms/Utils.h"
 #include "llvm/Transforms/Utils/CanonicalizeAliases.h"
 #include "llvm/Transforms/Utils/EntryExitInstrumenter.h"
@@ -88,6 +89,8 @@
   llvm::PassPluginLibraryInfo get##Ext##PluginInfo();
 #include "llvm/Support/Extension.def"
 
+extern cl::opt<bool> EnableMatrix;
+
 namespace {
 
 // Default filename used for profile generation.
@@ -1372,6 +1375,13 @@
     }
 
     if (CodeGenOpts.OptimizationLevel == 0) {
+      // FIXME: the backends do not handle matrix intrinsics currently. Make
+      // sure they are also lowered in O0. A lightweight version of the pass
+      // should run in the backend pipeline on demand.
+      if (EnableMatrix)
+        MPM.addPass(
+            createModuleToFunctionPassAdaptor(LowerMatrixIntrinsicsPass()));
+
       addCoroutinePassesAtO0(MPM, LangOpts, CodeGenOpts);
       addSanitizersAtO0(MPM, TargetTriple, LangOpts, CodeGenOpts);
     }
Index: clang/test/CodeGen/matrix-lowering-opt-levels.c
===================================================================
--- /dev/null
+++ clang/test/CodeGen/matrix-lowering-opt-levels.c
@@ -0,0 +1,20 @@
+// RUN: %clang -O0 -fenable-matrix -S -emit-llvm %s -o - | FileCheck  %s
+// RUN: %clang -O1 -fenable-matrix -S -emit-llvm %s -o - | FileCheck  %s
+// RUN: %clang -O2 -fenable-matrix -S -emit-llvm %s -o - | FileCheck  %s
+// RUN: %clang -O3 -fenable-matrix -S -emit-llvm %s -o - | FileCheck  %s
+// RUN: %clang -Ofast -fenable-matrix -S -emit-llvm %s -o - | FileCheck  %s
+// RUN: %clang -Os -fenable-matrix -S -emit-llvm %s -o - | FileCheck  %s
+// RUN: %clang -Oz -fenable-matrix -S -emit-llvm %s -o - | FileCheck  %s
+
+// Smoke test that the matrix intrinsics are lowered at any optimisation level.
+
+typedef float m4x4_t __attribute__((matrix_type(4, 4)));
+
+m4x4_t f(m4x4_t a, m4x4_t b, m4x4_t c) {
+  //
+  // CHECK-LABEL: f(
+  // CHECK-NOT:     @llvm.matrix
+  // CHECK:       }
+  //
+  return a + b * c;
+}
Index: llvm/lib/Passes/PassBuilder.cpp
===================================================================
--- llvm/lib/Passes/PassBuilder.cpp
+++ llvm/lib/Passes/PassBuilder.cpp
@@ -275,6 +275,8 @@
 extern cl::opt<AttributorRunOption> AttributorRun;
 extern cl::opt<bool> EnableKnowledgeRetention;
 
+extern cl::opt<bool> EnableMatrix;
+
 const PassBuilder::OptimizationLevel PassBuilder::OptimizationLevel::O0 = {
     /*SpeedLevel*/ 0,
     /*SizeLevel*/ 0};
@@ -929,6 +931,10 @@
   // Create an early function pass manager to cleanup the output of the
   // frontend.
   FunctionPassManager EarlyFPM(DebugLogging);
+
+  if (EnableMatrix)
+    EarlyFPM.addPass(LowerMatrixIntrinsicsPass());
+
   EarlyFPM.addPass(SimplifyCFGPass());
   EarlyFPM.addPass(SROA());
   EarlyFPM.addPass(EarlyCSEPass());
Index: llvm/lib/Transforms/IPO/PassManagerBuilder.cpp
===================================================================
--- llvm/lib/Transforms/IPO/PassManagerBuilder.cpp
+++ llvm/lib/Transforms/IPO/PassManagerBuilder.cpp
@@ -149,9 +149,9 @@
     "enable-order-file-instrumentation", cl::init(false), cl::Hidden,
     cl::desc("Enable order file instrumentation (default = off)"));
 
-static cl::opt<bool>
-    EnableMatrix("enable-matrix", cl::init(false), cl::Hidden,
-                 cl::desc("Enable lowering of the matrix intrinsics"));
+cl::opt<bool> EnableMatrix(
+    "enable-matrix", cl::init(false), cl::Hidden,
+    cl::desc("Enable lowering of the matrix intrinsics"));
 
 cl::opt<AttributorRunOption> AttributorRun(
     "attributor-enable", cl::Hidden, cl::init(AttributorRunOption::NONE),