Return xnn_status instead of hard coded integers in JIT generators

PiperOrigin-RevId: 465095604

26 files changed, 102 insertions, 76 deletions

diff --git a/src/f32-gemm/1x8-aarch64-neonfma-cortex-a75.cc b/src/f32-gemm/1x8-aarch64-neonfma-cortex-a75.cc
index 27329459f..d7523a6de 100644
--- a/src/f32-gemm/1x8-aarch64-neonfma-cortex-a75.cc
+++ b/src/f32-gemm/1x8-aarch64-neonfma-cortex-a75.cc
@@ -8,6 +8,7 @@
 #include <cstddef>
 #include <limits>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch64-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -270,9 +271,9 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75(
   g.generate(false, max_mr, nc_mod_nr, kc, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_f32_gemm_ukernel_1x8__aarch64_neonfma_prfm_cortex_a75(
@@ -285,7 +286,7 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_1x8__aarch64_neonfma_prfm_cortex_a75(
   g.generate(true, max_mr, nc_mod_nr, kc, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-gemm/4x8-aarch32-neon-cortex-a53.cc b/src/f32-gemm/4x8-aarch32-neon-cortex-a53.cc
index c55312f21..21cf97ba6 100644
--- a/src/f32-gemm/4x8-aarch32-neon-cortex-a53.cc
+++ b/src/f32-gemm/4x8-aarch32-neon-cortex-a53.cc
@@ -7,6 +7,7 @@
 #include <cassert>
 #include <limits>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -552,7 +553,7 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a53(xnn_code
   g.generate(max_mr, nc_mod_nr, kc, p->f32_minmax.min, p->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-gemm/4x8-aarch32-neon-cortex-a55.cc b/src/f32-gemm/4x8-aarch32-neon-cortex-a55.cc
index dc65448b2..e6c64bbc9 100644
--- a/src/f32-gemm/4x8-aarch32-neon-cortex-a55.cc
+++ b/src/f32-gemm/4x8-aarch32-neon-cortex-a55.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -443,7 +444,7 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a55(xnn_code
   g.generate(max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-gemm/4x8-aarch32-neon-cortex-a7.cc b/src/f32-gemm/4x8-aarch32-neon-cortex-a7.cc
index 9fb8bf161..4eceb89a0 100644
--- a/src/f32-gemm/4x8-aarch32-neon-cortex-a7.cc
+++ b/src/f32-gemm/4x8-aarch32-neon-cortex-a7.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -260,7 +261,7 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a7(xnn_code_
   g.generate(max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-gemm/4x8-aarch32-neon-cortex-a75.cc b/src/f32-gemm/4x8-aarch32-neon-cortex-a75.cc
index aa6d1286b..961e3ec2d 100644
--- a/src/f32-gemm/4x8-aarch32-neon-cortex-a75.cc
+++ b/src/f32-gemm/4x8-aarch32-neon-cortex-a75.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -393,9 +394,9 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a75(xnn_code
   g.generate(false, max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch32_neon_prfm_cortex_a75(xnn_code_buffer* code, size_t max_mr, size_t nc_mod_nr, size_t kc, const void* params) {
@@ -405,7 +406,7 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch32_neon_prfm_cortex_a75(xnn
   g.generate(true, max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-gemm/4x8-aarch32-neon-ld64.cc b/src/f32-gemm/4x8-aarch32-neon-ld64.cc
index 7bf5f1c26..33c38e42e 100644
--- a/src/f32-gemm/4x8-aarch32-neon-ld64.cc
+++ b/src/f32-gemm/4x8-aarch32-neon-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -240,7 +241,7 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch32_neon_ld64(xnn_code_buffe
   g.generate(max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-gemm/4x8-aarch64-neonfma-cortex-a75.cc b/src/f32-gemm/4x8-aarch64-neonfma-cortex-a75.cc
index e20a0c9e3..6c9e9b730 100644
--- a/src/f32-gemm/4x8-aarch64-neonfma-cortex-a75.cc
+++ b/src/f32-gemm/4x8-aarch64-neonfma-cortex-a75.cc
@@ -7,6 +7,7 @@
 #include <cstddef>
 #include <limits>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch64-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -522,9 +523,9 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75(xnn_c
   g.generate(false, max_mr, nc_mod_nr, kc, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch64_neonfma_prfm_cortex_a75(xnn_code_buffer* code, size_t max_mr, size_t nc_mod_nr, size_t kc, const void* params) {
@@ -535,7 +536,7 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_4x8__aarch64_neonfma_prfm_cortex_a75(
   g.generate(true, max_mr, nc_mod_nr, kc, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-gemm/6x8-aarch64-neonfma-ld128.cc b/src/f32-gemm/6x8-aarch64-neonfma-ld128.cc
index 858d55c56..ab15614fb 100644
--- a/src/f32-gemm/6x8-aarch64-neonfma-ld128.cc
+++ b/src/f32-gemm/6x8-aarch64-neonfma-ld128.cc
@@ -7,6 +7,7 @@
 #include <cstddef>
 #include <limits>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch64-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -393,7 +394,7 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128(
   g.generate(max_mr, nc_mod_nr, kc, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-gemm/upto6x8-aarch64-neonfma-cortex-a75.cc b/src/f32-gemm/upto6x8-aarch64-neonfma-cortex-a75.cc
index 88b2acd38..4ddd4af70 100644
--- a/src/f32-gemm/upto6x8-aarch64-neonfma-cortex-a75.cc
+++ b/src/f32-gemm/upto6x8-aarch64-neonfma-cortex-a75.cc
@@ -7,6 +7,7 @@
 #include <cstddef>
 #include <limits>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch64-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -1357,9 +1358,9 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_upto6x8__aarch64_neonfma_cortex_a75(x
   g.generate(false, max_mr, nc_mod_nr, kc, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_f32_gemm_ukernel_upto6x8__aarch64_neonfma_prfm_cortex_a75(xnn_code_buffer* code, size_t max_mr, size_t nc_mod_nr, size_t kc, const void* params) {
@@ -1370,7 +1371,7 @@ xnn_status_t xnn_generate_f32_gemm_ukernel_upto6x8__aarch64_neonfma_prfm_cortex_
   g.generate(true, max_mr, nc_mod_nr, kc, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-igemm/1x8-aarch64-neonfma-cortex-a75.cc b/src/f32-igemm/1x8-aarch64-neonfma-cortex-a75.cc
index 429909f2a..d1dd37d15 100644
--- a/src/f32-igemm/1x8-aarch64-neonfma-cortex-a75.cc
+++ b/src/f32-igemm/1x8-aarch64-neonfma-cortex-a75.cc
@@ -7,6 +7,7 @@
 #include <cstddef>
 #include <limits>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch64-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -302,9 +303,9 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_1x8__aarch64_neonfma_cortex_a75(
   g.generate(false, max_mr, nc_mod_nr, kc, ks, jit_params->f32_minmax.min, jit_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_f32_igemm_ukernel_1x8__aarch64_neonfma_prfm_cortex_a75(
@@ -317,7 +318,7 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_1x8__aarch64_neonfma_prfm_cortex_a75
   g.generate(true, max_mr, nc_mod_nr, kc, ks, jit_params->f32_minmax.min, jit_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-igemm/4x8-aarch32-neon-cortex-a53.cc b/src/f32-igemm/4x8-aarch32-neon-cortex-a53.cc
index ea07da368..bb3ac22f7 100644
--- a/src/f32-igemm/4x8-aarch32-neon-cortex-a53.cc
+++ b/src/f32-igemm/4x8-aarch32-neon-cortex-a53.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -536,7 +537,7 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a53(xnn_cod
   g.generate(max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-igemm/4x8-aarch32-neon-cortex-a55.cc b/src/f32-igemm/4x8-aarch32-neon-cortex-a55.cc
index f5e3be029..914cf6402 100644
--- a/src/f32-igemm/4x8-aarch32-neon-cortex-a55.cc
+++ b/src/f32-igemm/4x8-aarch32-neon-cortex-a55.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -469,7 +470,7 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a55(xnn_cod
   g.generate(max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-igemm/4x8-aarch32-neon-cortex-a7.cc b/src/f32-igemm/4x8-aarch32-neon-cortex-a7.cc
index 60d48548e..bfb8905c1 100644
--- a/src/f32-igemm/4x8-aarch32-neon-cortex-a7.cc
+++ b/src/f32-igemm/4x8-aarch32-neon-cortex-a7.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -288,7 +289,7 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a7(xnn_code
   g.generate(max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-igemm/4x8-aarch32-neon-cortex-a75.cc b/src/f32-igemm/4x8-aarch32-neon-cortex-a75.cc
index 781c38228..d63cc8313 100644
--- a/src/f32-igemm/4x8-aarch32-neon-cortex-a75.cc
+++ b/src/f32-igemm/4x8-aarch32-neon-cortex-a75.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -420,9 +421,9 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch32_neon_cortex_a75(xnn_cod
   g.generate(false, max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch32_neon_prfm_cortex_a75(xnn_code_buffer* code, size_t max_mr, size_t nc_mod_nr, size_t kc, size_t ks, const void* params) {
@@ -432,7 +433,7 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch32_neon_prfm_cortex_a75(xn
   g.generate(true, max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-igemm/4x8-aarch32-neon-ld64.cc b/src/f32-igemm/4x8-aarch32-neon-ld64.cc
index e26dc6333..b7e906d79 100644
--- a/src/f32-igemm/4x8-aarch32-neon-ld64.cc
+++ b/src/f32-igemm/4x8-aarch32-neon-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -267,7 +268,7 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch32_neon_ld64(xnn_code_buff
   g.generate(max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-igemm/4x8-aarch64-neonfma-cortex-a75.cc b/src/f32-igemm/4x8-aarch64-neonfma-cortex-a75.cc
index 09dfead01..63b21061c 100644
--- a/src/f32-igemm/4x8-aarch64-neonfma-cortex-a75.cc
+++ b/src/f32-igemm/4x8-aarch64-neonfma-cortex-a75.cc
@@ -7,6 +7,7 @@
 #include <cstddef>
 #include <limits>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch64-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -551,9 +552,9 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch64_neonfma_cortex_a75(xnn_
   g.generate(false, max_mr, nc_mod_nr, kc, ks, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch64_neonfma_prfm_cortex_a75(xnn_code_buffer* code, size_t max_mr, size_t nc_mod_nr, size_t kc, size_t ks, const void* params) {
@@ -564,7 +565,7 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_4x8__aarch64_neonfma_prfm_cortex_a75
   g.generate(true, max_mr, nc_mod_nr, kc, ks, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-igemm/6x8-aarch64-neonfma-ld128.cc b/src/f32-igemm/6x8-aarch64-neonfma-ld128.cc
index 93884704a..c133ce239 100644
--- a/src/f32-igemm/6x8-aarch64-neonfma-ld128.cc
+++ b/src/f32-igemm/6x8-aarch64-neonfma-ld128.cc
@@ -7,6 +7,7 @@
 #include <cstddef>
 #include <limits>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch64-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -423,7 +424,7 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_6x8__aarch64_neonfma_ld128(
   g.generate(max_mr, nc_mod_nr, kc, ks, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/f32-igemm/upto6x8-aarch64-neonfma-cortex-a75.cc b/src/f32-igemm/upto6x8-aarch64-neonfma-cortex-a75.cc
index b62ef10a0..6b9cef53b 100644
--- a/src/f32-igemm/upto6x8-aarch64-neonfma-cortex-a75.cc
+++ b/src/f32-igemm/upto6x8-aarch64-neonfma-cortex-a75.cc
@@ -7,6 +7,7 @@
 #include <cstddef>
 #include <limits>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch64-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -1392,9 +1393,9 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_upto6x8__aarch64_neonfma_cortex_a75(
   g.generate(false, max_mr, nc_mod_nr, kc, ks, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 
 }
 
@@ -1406,7 +1407,7 @@ xnn_status_t xnn_generate_f32_igemm_ukernel_upto6x8__aarch64_neonfma_prfm_cortex
   g.generate(true, max_mr, nc_mod_nr, kc, ks, gemm_params->f32_minmax.min, gemm_params->f32_minmax.max);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/qc8-gemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc b/src/qc8-gemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc
index ef44e1388..2a01a4d0f 100644
--- a/src/qc8-gemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc
+++ b/src/qc8-gemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -478,9 +479,9 @@ xnn_status_t xnn_generate_qc8_gemm_fp32_ukernel_4x8__aarch32_neonv8_mlal_lane_ld
   g.generate(false, max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_qc8_gemm_fp32_ukernel_4x8__aarch32_neonv8_mlal_lane_prfm_ld64(xnn_code_buffer* code, size_t max_mr, size_t nc_mod_nr, size_t kc, const void* params) {
@@ -489,7 +490,7 @@ xnn_status_t xnn_generate_qc8_gemm_fp32_ukernel_4x8__aarch32_neonv8_mlal_lane_pr
   g.generate(true, max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/qc8-gemm/4x8c4-fp32-aarch32-neondot-ld64.cc b/src/qc8-gemm/4x8c4-fp32-aarch32-neondot-ld64.cc
index f619972a4..ee3135403 100644
--- a/src/qc8-gemm/4x8c4-fp32-aarch32-neondot-ld64.cc
+++ b/src/qc8-gemm/4x8c4-fp32-aarch32-neondot-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -293,7 +294,7 @@ xnn_status_t xnn_generate_qc8_gemm_fp32_ukernel_4x8c4__aarch32_neondot_ld64(xnn_
   g.generate(max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/qc8-igemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc b/src/qc8-igemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc
index 1301d22f8..7fac2a5d7 100644
--- a/src/qc8-igemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc
+++ b/src/qc8-igemm/4x8-fp32-aarch32-neonv8-mlal-lane-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -501,9 +502,9 @@ xnn_status_t xnn_generate_qc8_igemm_fp32_ukernel_4x8__aarch32_neonv8_mlal_lane_l
   g.generate(false, max_mr, nc_mod_nr, kc, ks, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_qc8_igemm_fp32_ukernel_4x8__aarch32_neonv8_mlal_lane_prfm_ld64(xnn_code_buffer* code, size_t max_mr, size_t nc_mod_nr, size_t kc, size_t ks, const void* params) {
@@ -512,7 +513,7 @@ xnn_status_t xnn_generate_qc8_igemm_fp32_ukernel_4x8__aarch32_neonv8_mlal_lane_p
   g.generate(true, max_mr, nc_mod_nr, kc, ks, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/qc8-igemm/4x8c4-fp32-aarch32-neondot-ld64.cc b/src/qc8-igemm/4x8c4-fp32-aarch32-neondot-ld64.cc
index b51113d81..f58b0eefd 100644
--- a/src/qc8-igemm/4x8c4-fp32-aarch32-neondot-ld64.cc
+++ b/src/qc8-igemm/4x8c4-fp32-aarch32-neondot-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -319,7 +320,7 @@ xnn_status_t xnn_generate_qc8_igemm_fp32_ukernel_4x8c4__aarch32_neondot_ld64(xnn
   g.generate(max_mr, nc_mod_nr, kc, ks, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/qs8-gemm/4x8-rndnu-aarch32-neon-mlal-lane-ld64.cc b/src/qs8-gemm/4x8-rndnu-aarch32-neon-mlal-lane-ld64.cc
index e6b6e8e59..f48b92216 100644
--- a/src/qs8-gemm/4x8-rndnu-aarch32-neon-mlal-lane-ld64.cc
+++ b/src/qs8-gemm/4x8-rndnu-aarch32-neon-mlal-lane-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -483,9 +484,9 @@ xnn_status_t xnn_generate_qs8_gemm_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_ld6
   g.generate(false, max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_qs8_gemm_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_prfm_ld64(xnn_code_buffer* code, size_t max_mr, size_t nc_mod_nr, size_t kc, const void* params) {
@@ -494,7 +495,7 @@ xnn_status_t xnn_generate_qs8_gemm_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_prf
   g.generate(true, max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/qs8-gemm/4x8c4-rndnu-aarch32-neondot-ld64.cc b/src/qs8-gemm/4x8c4-rndnu-aarch32-neondot-ld64.cc
index e12b0e546..fb215943b 100644
--- a/src/qs8-gemm/4x8c4-rndnu-aarch32-neondot-ld64.cc
+++ b/src/qs8-gemm/4x8c4-rndnu-aarch32-neondot-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/gemm.h>
@@ -298,7 +299,7 @@ xnn_status_t xnn_generate_qs8_gemm_rndnu_ukernel_4x8c4__aarch32_neondot_ld64(xnn
   g.generate(max_mr, nc_mod_nr, kc, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/qs8-igemm/4x8-rndnu-aarch32-neon-mlal-lane-ld64.cc b/src/qs8-igemm/4x8-rndnu-aarch32-neon-mlal-lane-ld64.cc
index 43e62bab1..bbadb290d 100644
--- a/src/qs8-igemm/4x8-rndnu-aarch32-neon-mlal-lane-ld64.cc
+++ b/src/qs8-igemm/4x8-rndnu-aarch32-neon-mlal-lane-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -506,9 +507,9 @@ xnn_status_t xnn_generate_qs8_igemm_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_ld
   g.generate(false, max_mr, nc_mod_nr, kc, ks, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
 
 xnn_status_t xnn_generate_qs8_igemm_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_prfm_ld64(xnn_code_buffer* code, size_t max_mr, size_t nc_mod_nr, size_t kc, size_t ks, const void* params) {
@@ -517,7 +518,7 @@ xnn_status_t xnn_generate_qs8_igemm_rndnu_ukernel_4x8__aarch32_neon_mlal_lane_pr
   g.generate(true, max_mr, nc_mod_nr, kc, ks, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }
diff --git a/src/qs8-igemm/4x8c4-rndnu-aarch32-neondot-ld64.cc b/src/qs8-igemm/4x8c4-rndnu-aarch32-neondot-ld64.cc
index 7a407c75a..f193e15d0 100644
--- a/src/qs8-igemm/4x8c4-rndnu-aarch32-neondot-ld64.cc
+++ b/src/qs8-igemm/4x8c4-rndnu-aarch32-neondot-ld64.cc
@@ -6,6 +6,7 @@
 
 #include <cassert>
 
+#include <xnnpack.h>
 #include <xnnpack/aarch32-assembler.h>
 #include <xnnpack/allocator.h>
 #include <xnnpack/igemm.h>
@@ -324,7 +325,7 @@ xnn_status_t xnn_generate_qs8_igemm_rndnu_ukernel_4x8c4__aarch32_neondot_ld64(xn
   g.generate(max_mr, nc_mod_nr, kc, ks, nullptr);
   g.finalize();
   if (g.error() != xnnpack::Error::kNoError) {
-    return 3;
+    return xnn_status_invalid_state;
   }
-  return 0;
+  return xnn_status_success;
 }