diff --git a/BUILD.bazel b/BUILD.bazel
index 16fbf7adac47..8a7601b8c111 100644
--- a/BUILD.bazel
+++ b/BUILD.bazel
@@ -766,6 +766,7 @@ xnnpack_microkernel_cc_library(
     srcs_prod = ["src/amalgam/gen/scalar.c"],
     srcs_test = ALL_SCALAR_MICROKERNEL_SRCS,
     deps = MICROKERNEL_DEPS + [
+        ":config_hdrs",
         "@FP16",
         "@FXdiv",
     ],
@@ -855,6 +856,7 @@ xnnpack_microkernel_cc_library(
     msvc_copts = xnnpack_msvc_std_copts(),
     deps = MICROKERNEL_DEPS + xnnpack_if_kleidiai_enabled([
         "@KleidiAI//kai/ukernels/matmul",
+        "@KleidiAI//kai/ukernels/matmul:lhs_quant_pack_qai8dxp_f32",
     ]),
 )
 
@@ -977,6 +979,7 @@ xnnpack_microkernel_cc_library(
     msvc_copts = xnnpack_msvc_std_copts(),
     deps = MICROKERNEL_DEPS + xnnpack_if_kleidiai_enabled([
         "@KleidiAI//kai/ukernels/matmul",
+        "@KleidiAI//kai/ukernels/matmul:clamp_f32_qai8dxp1x8_qsi4cxp8x8_1x8x32_neon_dotprod",
     ]),
 )
 
@@ -1000,6 +1003,7 @@ xnnpack_microkernel_cc_library(
     msvc_copts = xnnpack_msvc_std_copts(),
     deps = MICROKERNEL_DEPS + xnnpack_if_kleidiai_enabled([
         "@KleidiAI//kai/ukernels/matmul",
+        "@KleidiAI//kai/ukernels/matmul:clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm",
     ]),
 )
 
@@ -1786,10 +1790,15 @@ xnnpack_cc_library(
         ":config_hdrs",
         ":logging",
         ":math",
+        ":microparams",
+        ":params",
         ":unaligned",
+        ":xnnpack_h",
         "@FP16",
     ] + xnnpack_if_kleidiai_enabled([
         "@KleidiAI//kai/ukernels/matmul",
+        "@KleidiAI//kai/ukernels/matmul:rhs_pack_kxn_qsi4cxp_qsu4cxs1s0",
+        "@KleidiAI//kai/ukernels/matmul:rhs_pack_nxk_qsi4cxp_qsu4cxs1s0",
     ]),
 )
 
@@ -2007,6 +2016,7 @@ xnnpack_cc_library(
     ],
     deps = [
         ":allocator",
+        ":logging",
         ":microkernel_configs",
         ":microparams",
         ":params",
@@ -2040,10 +2050,12 @@ xnnpack_cc_library(
     gcc_copts = xnnpack_gcc_std_copts(),
     msvc_copts = xnnpack_msvc_std_copts(),
     deps = [
-        ":cache",
+        ":allocator",
         ":common",
+        ":logging",
         ":math",
         ":operator_h",
+        ":params",
         ":xnnpack_h",
     ],
 )
@@ -2076,6 +2088,7 @@ xnnpack_cc_library(
         ":microkernel_configs",
         ":microkernel_type",
         ":microkernel_utils",
+        ":microkernels_h",
         ":microparams_init",
         ":normalization",
         ":operator_type",
@@ -2083,6 +2096,7 @@ xnnpack_cc_library(
         ":packing",
         ":params",
         ":post_operation",
+        ":quantization",
         ":xnnpack_h",
         "@FP16",
         "@pthreadpool",
@@ -2226,6 +2240,9 @@ xnnpack_cc_library(
         ":allocator",
         ":build_identifier",
         ":common",
+        ":config_hdrs",
+        ":hardware_config",
+        ":init_once",
         ":logging",
         ":math",
         ":microkernels_h",
@@ -2234,6 +2251,7 @@ xnnpack_cc_library(
         ":operators",
         ":params",
         ":subgraph",
+        "@pthreadpool",
     ],
 )
 
@@ -2257,17 +2275,12 @@ xnnpack_cc_library(
     visibility = xnnpack_visibility(),
     deps = [
         ":allocator",
-        ":build_identifier",
         ":common",
+        ":config_hdrs",
+        ":init_once",
         ":logging",
-        ":math",
-        ":microkernels_h",
-        ":microparams",
-        ":microparams_init",
-        ":operators",
         ":params",
-        ":prod_microkernels",
-        ":subgraph",
+        "@pthreadpool",
     ],
 )
 
@@ -2291,16 +2304,12 @@ xnnpack_cc_library(
     visibility = xnnpack_visibility(),
     deps = [
         ":allocator",
-        ":build_identifier",
         ":common",
+        ":config_hdrs",
+        ":init_once",
         ":logging",
-        ":math",
-        ":microkernels_h",
-        ":microparams",
-        ":microparams_init",
-        ":operators",
         ":params",
-        ":prod_microkernels",
+        "@pthreadpool",
     ],
 )
 
diff --git a/src/f16-avgpool/f16-avgpool-9p8x-minmax-f16c-c8.c b/src/f16-avgpool/f16-avgpool-9p8x-minmax-f16c-c8.c
index 6f0d7328f750..f1e8ac057dae 100644
--- a/src/f16-avgpool/f16-avgpool-9p8x-minmax-f16c-c8.c
+++ b/src/f16-avgpool/f16-avgpool-9p8x-minmax-f16c-c8.c
@@ -7,8 +7,13 @@
 
 #include <immintrin.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
 #include "xnnpack/intrinsics-polyfill.h"
+#include "xnnpack/microparams.h"
 
 
 void xnn_f16_avgpool_minmax_ukernel_9p8x__f16c_c8(
diff --git a/src/f16-avgpool/f16-avgpool-9x-minmax-f16c-c8.c b/src/f16-avgpool/f16-avgpool-9x-minmax-f16c-c8.c
index 579b593039cc..95b836fdc28f 100644
--- a/src/f16-avgpool/f16-avgpool-9x-minmax-f16c-c8.c
+++ b/src/f16-avgpool/f16-avgpool-9x-minmax-f16c-c8.c
@@ -7,8 +7,13 @@
 
 #include <immintrin.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
 #include "xnnpack/intrinsics-polyfill.h"
+#include "xnnpack/microparams.h"
 
 
 void xnn_f16_avgpool_minmax_ukernel_9x__f16c_c8(
diff --git a/src/f16-pavgpool/f16-pavgpool-9p8x-minmax-avx2-c8.c b/src/f16-pavgpool/f16-pavgpool-9p8x-minmax-avx2-c8.c
index 0f66c40d087a..adf2b9c7217f 100644
--- a/src/f16-pavgpool/f16-pavgpool-9p8x-minmax-avx2-c8.c
+++ b/src/f16-pavgpool/f16-pavgpool-9p8x-minmax-avx2-c8.c
@@ -7,7 +7,12 @@
 
 #include <immintrin.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
+#include "xnnpack/common.h"
 #include "xnnpack/intrinsics-polyfill.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/pavgpool.h"
 
 
diff --git a/src/f16-pavgpool/f16-pavgpool-9x-minmax-avx2-c8.c b/src/f16-pavgpool/f16-pavgpool-9x-minmax-avx2-c8.c
index 4b2b8eae2c73..bf9a896fcb9e 100644
--- a/src/f16-pavgpool/f16-pavgpool-9x-minmax-avx2-c8.c
+++ b/src/f16-pavgpool/f16-pavgpool-9x-minmax-avx2-c8.c
@@ -7,7 +7,12 @@
 
 #include <immintrin.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
+#include "xnnpack/common.h"
 #include "xnnpack/intrinsics-polyfill.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/pavgpool.h"
 
 
diff --git a/src/f32-avgpool/f32-avgpool-9p8x-minmax-scalar-c1.c b/src/f32-avgpool/f32-avgpool-9p8x-minmax-scalar-c1.c
index 2ee82bd76224..b4d6763952ce 100644
--- a/src/f32-avgpool/f32-avgpool-9p8x-minmax-scalar-c1.c
+++ b/src/f32-avgpool/f32-avgpool-9p8x-minmax-scalar-c1.c
@@ -5,8 +5,13 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
 #include "xnnpack/math.h"
+#include "xnnpack/microparams.h"
 
 
 void xnn_f32_avgpool_minmax_ukernel_9p8x__scalar_c1(
diff --git a/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c b/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c
index d9295c4e4595..b0b4047a381f 100644
--- a/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c
+++ b/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c
@@ -5,9 +5,13 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
 #include <xmmintrin.h>
 
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 
 
 void xnn_f32_avgpool_minmax_ukernel_9p8x__sse_c4(
diff --git a/src/f32-avgpool/f32-avgpool-9x-minmax-scalar-c1.c b/src/f32-avgpool/f32-avgpool-9x-minmax-scalar-c1.c
index d82716ee27df..f0832345bf0a 100644
--- a/src/f32-avgpool/f32-avgpool-9x-minmax-scalar-c1.c
+++ b/src/f32-avgpool/f32-avgpool-9x-minmax-scalar-c1.c
@@ -5,8 +5,13 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
 #include "xnnpack/math.h"
+#include "xnnpack/microparams.h"
 
 
 void xnn_f32_avgpool_minmax_ukernel_9x__scalar_c1(
diff --git a/src/f32-avgpool/f32-avgpool-9x-minmax-sse-c4.c b/src/f32-avgpool/f32-avgpool-9x-minmax-sse-c4.c
index 206a8b3e1b61..7054d44c5b21 100644
--- a/src/f32-avgpool/f32-avgpool-9x-minmax-sse-c4.c
+++ b/src/f32-avgpool/f32-avgpool-9x-minmax-sse-c4.c
@@ -5,9 +5,14 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
 #include <xmmintrin.h>
 
+
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 
 
 void xnn_f32_avgpool_minmax_ukernel_9x__sse_c4(
diff --git a/src/f32-pavgpool/f32-pavgpool-9p8x-minmax-scalar-c1.c b/src/f32-pavgpool/f32-pavgpool-9p8x-minmax-scalar-c1.c
index d9ecc2d33195..231af6ca5a4a 100644
--- a/src/f32-pavgpool/f32-pavgpool-9p8x-minmax-scalar-c1.c
+++ b/src/f32-pavgpool/f32-pavgpool-9p8x-minmax-scalar-c1.c
@@ -5,6 +5,11 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/pavgpool.h"
 #include "xnnpack/math.h"
 
diff --git a/src/f32-pavgpool/f32-pavgpool-9p8x-minmax-sse-c4.c b/src/f32-pavgpool/f32-pavgpool-9p8x-minmax-sse-c4.c
index b25cdd663d88..7e5973a0c986 100644
--- a/src/f32-pavgpool/f32-pavgpool-9p8x-minmax-sse-c4.c
+++ b/src/f32-pavgpool/f32-pavgpool-9p8x-minmax-sse-c4.c
@@ -5,8 +5,13 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
 #include <xmmintrin.h>
 
+
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/pavgpool.h"
 
 
diff --git a/src/f32-pavgpool/f32-pavgpool-9x-minmax-scalar-c1.c b/src/f32-pavgpool/f32-pavgpool-9x-minmax-scalar-c1.c
index a60dc4f9f3fe..fa3992472473 100644
--- a/src/f32-pavgpool/f32-pavgpool-9x-minmax-scalar-c1.c
+++ b/src/f32-pavgpool/f32-pavgpool-9x-minmax-scalar-c1.c
@@ -5,6 +5,11 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/pavgpool.h"
 #include "xnnpack/math.h"
 
diff --git a/src/f32-pavgpool/f32-pavgpool-9x-minmax-sse-c4.c b/src/f32-pavgpool/f32-pavgpool-9x-minmax-sse-c4.c
index 8a613d097b27..2743e3c723e2 100644
--- a/src/f32-pavgpool/f32-pavgpool-9x-minmax-sse-c4.c
+++ b/src/f32-pavgpool/f32-pavgpool-9x-minmax-sse-c4.c
@@ -5,8 +5,12 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
 #include <xmmintrin.h>
 
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/pavgpool.h"
 
 
diff --git a/src/qu8-avgpool/qu8-avgpool-9p8x-minmax-fp32-scalar-imagic-c1.c b/src/qu8-avgpool/qu8-avgpool-9p8x-minmax-fp32-scalar-imagic-c1.c
index 3f4a169a453e..a52f68040c6a 100644
--- a/src/qu8-avgpool/qu8-avgpool-9p8x-minmax-fp32-scalar-imagic-c1.c
+++ b/src/qu8-avgpool/qu8-avgpool-9p8x-minmax-fp32-scalar-imagic-c1.c
@@ -8,8 +8,13 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
 #include "xnnpack/math.h"
+#include "xnnpack/microparams.h"
 
 
 void xnn_qu8_avgpool_minmax_fp32_ukernel_9p8x__scalar_imagic_c1(
diff --git a/src/qu8-avgpool/qu8-avgpool-9p8x-minmax-fp32-sse2-c8.c b/src/qu8-avgpool/qu8-avgpool-9p8x-minmax-fp32-sse2-c8.c
index d4a0e41af921..d70b3aebdda5 100644
--- a/src/qu8-avgpool/qu8-avgpool-9p8x-minmax-fp32-sse2-c8.c
+++ b/src/qu8-avgpool/qu8-avgpool-9p8x-minmax-fp32-sse2-c8.c
@@ -9,8 +9,12 @@
 #include <assert.h>
 
 #include <emmintrin.h>
+#include <stddef.h>
+#include <stdint.h>
 
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/unaligned.h"
 
 
diff --git a/src/qu8-avgpool/qu8-avgpool-9x-minmax-fp32-scalar-imagic-c1.c b/src/qu8-avgpool/qu8-avgpool-9x-minmax-fp32-scalar-imagic-c1.c
index 83018ea27216..b61405fac6b4 100644
--- a/src/qu8-avgpool/qu8-avgpool-9x-minmax-fp32-scalar-imagic-c1.c
+++ b/src/qu8-avgpool/qu8-avgpool-9x-minmax-fp32-scalar-imagic-c1.c
@@ -8,8 +8,13 @@
 
 #include <assert.h>
 
+#include <stddef.h>
+#include <stdint.h>
+
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
 #include "xnnpack/math.h"
+#include "xnnpack/microparams.h"
 
 
 void xnn_qu8_avgpool_minmax_fp32_ukernel_9x__scalar_imagic_c1(
diff --git a/src/qu8-avgpool/qu8-avgpool-9x-minmax-fp32-sse2-c8.c b/src/qu8-avgpool/qu8-avgpool-9x-minmax-fp32-sse2-c8.c
index 86566f6235b7..8341f6517d88 100644
--- a/src/qu8-avgpool/qu8-avgpool-9x-minmax-fp32-sse2-c8.c
+++ b/src/qu8-avgpool/qu8-avgpool-9x-minmax-fp32-sse2-c8.c
@@ -9,8 +9,12 @@
 #include <assert.h>
 
 #include <emmintrin.h>
+#include <stddef.h>
+#include <stdint.h>
 
 #include "xnnpack/avgpool.h"
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/unaligned.h"
 
 
diff --git a/src/s8-vclamp/s8-vclamp-scalar-u4.c b/src/s8-vclamp/s8-vclamp-scalar-u4.c
index ff5802cf92d6..8f75582f4ef0 100644
--- a/src/s8-vclamp/s8-vclamp-scalar-u4.c
+++ b/src/s8-vclamp/s8-vclamp-scalar-u4.c
@@ -4,8 +4,12 @@
 // LICENSE file in the root directory of this source tree.
 
 #include <assert.h>
+#include <stddef.h>
+#include <stdint.h>
 
+#include "xnnpack/common.h"
 #include "xnnpack/math.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/vunary.h"
 
 
diff --git a/src/s8-vclamp/s8-vclamp-sse2-u64.c b/src/s8-vclamp/s8-vclamp-sse2-u64.c
index e0b9525fec58..3b7b1e8ddfb9 100644
--- a/src/s8-vclamp/s8-vclamp-sse2-u64.c
+++ b/src/s8-vclamp/s8-vclamp-sse2-u64.c
@@ -6,7 +6,11 @@
 #include <assert.h>
 
 #include <emmintrin.h>
+#include <stddef.h>
+#include <stdint.h>
 
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/unaligned.h"
 #include "xnnpack/vunary.h"
 
diff --git a/src/s8-vclamp/s8-vclamp-sse41-u64.c b/src/s8-vclamp/s8-vclamp-sse41-u64.c
index 23bc5fa08785..f7d842b11733 100644
--- a/src/s8-vclamp/s8-vclamp-sse41-u64.c
+++ b/src/s8-vclamp/s8-vclamp-sse41-u64.c
@@ -6,7 +6,11 @@
 #include <assert.h>
 
 #include <smmintrin.h>
+#include <stddef.h>
+#include <stdint.h>
 
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/unaligned.h"
 #include "xnnpack/vunary.h"
 
diff --git a/src/subgraph/multiply2.c b/src/subgraph/multiply2.c
index 6c07c31ae7c5..46d2c9b52387 100644
--- a/src/subgraph/multiply2.c
+++ b/src/subgraph/multiply2.c
@@ -9,6 +9,7 @@
 #include <stdint.h>
 #include <string.h>
 
+
 #include "xnnpack.h"
 #include "xnnpack/common.h"
 #include "xnnpack/log.h"
diff --git a/src/u8-rmax/u8-rmax-scalar-u2.c b/src/u8-rmax/u8-rmax-scalar-u2.c
index 4d887d82965b..a928ed52e40d 100644
--- a/src/u8-rmax/u8-rmax-scalar-u2.c
+++ b/src/u8-rmax/u8-rmax-scalar-u2.c
@@ -4,6 +4,8 @@
 // LICENSE file in the root directory of this source tree.
 
 #include <assert.h>
+#include <stddef.h>
+#include <stdint.h>
 
 #include "xnnpack/reduce.h"
 
diff --git a/src/u8-rmax/u8-rmax-sse2-u16.c b/src/u8-rmax/u8-rmax-sse2-u16.c
index 50054f64e369..7c8e87e133be 100644
--- a/src/u8-rmax/u8-rmax-sse2-u16.c
+++ b/src/u8-rmax/u8-rmax-sse2-u16.c
@@ -9,7 +9,10 @@
 #include <assert.h>
 
 #include <emmintrin.h>
+#include <stddef.h>
+#include <stdint.h>
 
+#include "xnnpack/common.h"
 #include "xnnpack/reduce.h"
 
 
diff --git a/src/u8-vclamp/u8-vclamp-scalar-u4.c b/src/u8-vclamp/u8-vclamp-scalar-u4.c
index d70ce16b73b1..32472df6f292 100644
--- a/src/u8-vclamp/u8-vclamp-scalar-u4.c
+++ b/src/u8-vclamp/u8-vclamp-scalar-u4.c
@@ -4,8 +4,12 @@
 // LICENSE file in the root directory of this source tree.
 
 #include <assert.h>
+#include <stddef.h>
+#include <stdint.h>
 
+#include "xnnpack/common.h"
 #include "xnnpack/math.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/vunary.h"
 
 
diff --git a/src/u8-vclamp/u8-vclamp-sse2-u64.c b/src/u8-vclamp/u8-vclamp-sse2-u64.c
index c5ae43eb1a8a..dca2796a4ef6 100644
--- a/src/u8-vclamp/u8-vclamp-sse2-u64.c
+++ b/src/u8-vclamp/u8-vclamp-sse2-u64.c
@@ -6,7 +6,11 @@
 #include <assert.h>
 
 #include <emmintrin.h>
+#include <stddef.h>
+#include <stdint.h>
 
+#include "xnnpack/common.h"
+#include "xnnpack/microparams.h"
 #include "xnnpack/unaligned.h"
 #include "xnnpack/vunary.h"