diff --git a/bench/f32-gemm-e2e.cc b/bench/f32-gemm-e2e.cc
index 2523fda..190d55c 100644
--- a/bench/f32-gemm-e2e.cc
+++ b/bench/f32-gemm-e2e.cc
@@ -283,6 +283,15 @@
       4 /* mr */, 8 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
       benchmark::utils::CheckNEON);
   }
+  static void f32_gemm_4x8__aarch32_neon_cortex_a55(benchmark::State& state, models::ExecutionPlanFactory model) {
+    GEMMEnd2EndBenchmark(state, model,
+      xnn_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a55,
+      xnn_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a55,
+      xnn_f32_gemm_ukernel_1x8__neon_lane_ld64,
+      xnn_f32_igemm_ukernel_1x8__neon_lane_ld64,
+      4 /* mr */, 8 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
+      benchmark::utils::CheckNEON);
+  }
   static void f32_gemm_4x8__aarch32_neon_cortex_a75(benchmark::State& state, models::ExecutionPlanFactory model) {
     GEMMEnd2EndBenchmark(state, model,
       xnn_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a75,
@@ -304,6 +313,7 @@
 
   BENCHMARK_END2END(f32_gemm_4x8__aarch32_neon_ld64);
   BENCHMARK_END2END(f32_gemm_4x8__aarch32_neon_cortex_a53);
+  BENCHMARK_END2END(f32_gemm_4x8__aarch32_neon_cortex_a55);
   BENCHMARK_END2END(f32_gemm_4x8__aarch32_neon_cortex_a75);
   BENCHMARK_END2END(f32_gemm_4x8__aarch32_neon_pld_cortex_a75);
 #endif  // XNN_ARCH_ARM && XNN_ENABLE_ASSEMBLY
diff --git a/bench/f32-gemm.cc b/bench/f32-gemm.cc
index 8788f23..c68578f 100644
--- a/bench/f32-gemm.cc
+++ b/bench/f32-gemm.cc
@@ -457,6 +457,9 @@
   static void f32_gemm_4x8__aarch32_neon_cortex_a53(benchmark::State& state, const char* net) {
     GEMMBenchmark(state, xnn_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a53, 4, 8, 1, 1, benchmark::utils::CheckNEON);
   }
+  static void f32_gemm_4x8__aarch32_neon_cortex_a55(benchmark::State& state, const char* net) {
+    GEMMBenchmark(state, xnn_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a55, 4, 8, 1, 1, benchmark::utils::CheckNEON);
+  }
   static void f32_gemm_4x8__aarch32_neon_cortex_a75(benchmark::State& state, const char* net) {
     GEMMBenchmark(state, xnn_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a75, 4, 8, 1, 1, benchmark::utils::CheckNEON);
   }
@@ -466,6 +469,7 @@
 
   BENCHMARK_GEMM(f32_gemm_4x8__aarch32_neon_ld64)
   BENCHMARK_GEMM(f32_gemm_4x8__aarch32_neon_cortex_a53)
+  BENCHMARK_GEMM(f32_gemm_4x8__aarch32_neon_cortex_a55)
   BENCHMARK_GEMM(f32_gemm_4x8__aarch32_neon_cortex_a75)
   BENCHMARK_GEMM(f32_gemm_4x8__aarch32_neon_pld_cortex_a75)
 #endif  // XNN_ARCH_ARM
diff --git a/bench/f32-igemm.cc b/bench/f32-igemm.cc
index 10bcbee..1aa6fcd 100644
--- a/bench/f32-igemm.cc
+++ b/bench/f32-igemm.cc
@@ -262,8 +262,24 @@
   static void f32_igemm_4x8__aarch32_neon_ld64(benchmark::State& state, const char* net) {
     IGEMMBenchmark(state, xnn_f32_igemm_ukernel_4x8__aarch32_neon_ld64, 4, 8, 1, 1);
   }
+  static void f32_igemm_4x8__aarch32_neon_cortex_a53(benchmark::State& state, const char* net) {
+    IGEMMBenchmark(state, xnn_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a53, 4, 8, 1, 1);
+  }
+  static void f32_igemm_4x8__aarch32_neon_cortex_a55(benchmark::State& state, const char* net) {
+    IGEMMBenchmark(state, xnn_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a55, 4, 8, 1, 1);
+  }
+  static void f32_igemm_4x8__aarch32_neon_pld_cortex_a75(benchmark::State& state, const char* net) {
+    IGEMMBenchmark(state, xnn_f32_igemm_ukernel_4x8__aarch32_neon_pld_cortex_a75, 4, 8, 1, 1);
+  }
+  static void f32_igemm_4x8__aarch32_neon_cortex_a75(benchmark::State& state, const char* net) {
+    IGEMMBenchmark(state, xnn_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a75, 4, 8, 1, 1);
+  }
 
   BENCHMARK_CONV(f32_igemm_4x8__aarch32_neon_ld64)
+  BENCHMARK_CONV(f32_igemm_4x8__aarch32_neon_cortex_a53)
+  BENCHMARK_CONV(f32_igemm_4x8__aarch32_neon_cortex_a55)
+  BENCHMARK_CONV(f32_igemm_4x8__aarch32_neon_pld_cortex_a75)
+  BENCHMARK_CONV(f32_igemm_4x8__aarch32_neon_cortex_a75)
 #endif  /* XNN_ARCH_ARM */
 
 #if XNN_ARCH_ARM64 && XNN_ENABLE_ASSEMBLY
@@ -379,22 +395,6 @@
   BENCHMARK_CONV(f32_igemm_6x8__neonfma_lane_ld128)
 #endif  /* XNN_ARCH_ARM64 */
 
-#if XNN_ARCH_ARM && XNN_ENABLE_ASSEMBLY
-  static void f32_igemm_4x8__aarch32_neon_pld_cortex_a75(benchmark::State& state, const char* net) {
-    IGEMMBenchmark(state, xnn_f32_igemm_ukernel_4x8__aarch32_neon_pld_cortex_a75, 4, 8, 1, 1);
-  }
-  static void f32_igemm_4x8__aarch32_neon_cortex_a75(benchmark::State& state, const char* net) {
-    IGEMMBenchmark(state, xnn_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a75, 4, 8, 1, 1);
-  }
-  static void f32_igemm_4x8__aarch32_neon_cortex_a53(benchmark::State& state, const char* net) {
-    IGEMMBenchmark(state, xnn_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a53, 4, 8, 1, 1);
-  }
-
-  BENCHMARK_CONV(f32_igemm_4x8__aarch32_neon_pld_cortex_a75)
-  BENCHMARK_CONV(f32_igemm_4x8__aarch32_neon_cortex_a75)
-  BENCHMARK_CONV(f32_igemm_4x8__aarch32_neon_cortex_a53)
-#endif  /* XNN_ARCH_ARM */
-
 #if XNN_ARCH_X86 || XNN_ARCH_X86_64
   static void f32_igemm_1x8__sse_load1(benchmark::State& state, const char* net) {
     IGEMMBenchmark(state, xnn_f32_igemm_ukernel_1x8__sse_load1, 1, 8, 1, 1);