pq_crypto 0.6.4 → 0.6.6

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/auto.h

@@ -24,38 +24,44 @@
 /*
  * Keccak-f1600
  *
- * - On Arm-based Apple CPUs, we pick a pure Neon implementation.
+ * - On Arm-based Apple CPUs, or if MLK_SYS_AARCH64_FAST_SHA3 is set,
+ *   we pick a pure Neon implementation.
  * - Otherwise, unless MLK_SYS_AARCH64_SLOW_BARREL_SHIFTER is set,
  *   we use lazy-rotation scalar assembly from @[HYBRID].
  * - Otherwise, if MLK_SYS_AARCH64_SLOW_BARREL_SHIFTER is set, we
  *   fall back to the standard C implementation.
  */
-#if defined(__ARM_FEATURE_SHA3) && defined(__APPLE__)
+#if defined(__ARM_FEATURE_SHA3) && \
+    (defined(__APPLE__) || defined(MLK_SYS_AARCH64_FAST_SHA3))
 #include "x1_v84a.h"
 #elif !defined(MLK_SYS_AARCH64_SLOW_BARREL_SHIFTER)
 #include "x1_scalar.h"
 #endif
 
+/* Batched, SIMD-based Keccak-f1600 implementations. */
+#if defined(MLK_SYS_AARCH64_NEON)
+
 /*
  * Keccak-f1600x2/x4
  *
  * The optimal implementation is highly CPU-specific; see @[HYBRID].
  *
  * For now, if v8.4-A is not implemented, we fall back to Keccak-f1600.
- * If v8.4-A is implemented and we are on an Apple CPU, we use a plain
- * Neon-based implementation.
- * If v8.4-A is implemented and we are not on an Apple CPU, we use a
- * scalar/Neon/Neon hybrid.
- * The reason for this distinction is that Apple CPUs appear to implement
- * the SHA3 instructions on all SIMD units, while Arm CPUs prior to Cortex-X4
- * don't, and ordinary Neon instructions are still needed.
+ * If v8.4-A is implemented and we are on an Apple CPU or
+ * MLK_SYS_AARCH64_FAST_SHA3 is set, we use a plain Neon-based
+ * implementation.
+ * Otherwise, if v8.4-A is implemented, we use a scalar/Neon/Neon hybrid.
+ * The reason for this distinction is that Apple CPUs (and CPUs flagged with
+ * MLK_SYS_AARCH64_FAST_SHA3) implement the SHA3 instructions on all SIMD
+ * units, while Arm CPUs prior to Cortex-X4 don't, and ordinary Neon
+ * instructions are still needed.
  */
 #if defined(__ARM_FEATURE_SHA3)
 /*
- * For Apple-M cores, we use a plain implementation leveraging SHA3
- * instructions only.
+ * For Apple-M cores (and CPUs flagged with MLK_SYS_AARCH64_FAST_SHA3), we
+ * use a plain implementation leveraging SHA3 instructions only.
  */
-#if defined(__APPLE__)
+#if defined(__APPLE__) || defined(MLK_SYS_AARCH64_FAST_SHA3)
 #include "x2_v84a.h"
 #else
 #include "x4_v8a_v84a_scalar.h"
@@ -67,4 +73,6 @@
 
 #endif /* !__ARM_FEATURE_SHA3 */
 
+#endif /* MLK_SYS_AARCH64_NEON */
+
 #endif /* !MLK_FIPS202_NATIVE_AARCH64_AUTO_H */

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/src/keccak_f1600_x1_scalar_aarch64_asm.S

@@ -13,6 +13,8 @@
   Description: AArch64 scalar implementation of Keccak-f[1600] permutation for single state
   Signature: void mlk_keccak_f1600_x1_scalar_aarch64_asm(uint64_t state[25], const uint64_t rc[24])
   ABI:
+    Architecture: aarch64
+    CallingConvention: AAPCS64
     x0:
       type: buffer
       size_bytes: 200
@@ -66,7 +68,7 @@ MLK_ASM_FN_SYMBOL(keccak_f1600_x1_scalar_aarch64_asm)
         .cfi_rel_offset x29, 0x70
         .cfi_rel_offset x30, 0x78
 
-Lkeccak_f1600_x1_scalar_initial:
+Lmlk_keccak_f1600_x1_scalar_initial:
         mov x26, x1
         str x1, [sp, #0x8]
         ldp x1, x6, [x0]
@@ -191,7 +193,7 @@ Lkeccak_f1600_x1_scalar_initial:
         eor x14, x26, x6, ror #46
         eor x6, x27, x29, ror #41
 
-Lkeccak_f1600_x1_scalar_loop:
+Lmlk_keccak_f1600_x1_scalar_loop:
         eor x0, x15, x11, ror #52
         eor x0, x0, x13, ror #48
         eor x26, x8, x9, ror #57
@@ -304,7 +306,7 @@ Lkeccak_f1600_x1_scalar_loop:
         bic x3, x0, x30, ror #5
         eor x23, x3, x26, ror #52
         eor x3, x29, x30, ror #24
-        b.le Lkeccak_f1600_x1_scalar_loop
+        b.le Lmlk_keccak_f1600_x1_scalar_loop
         ror x6, x6, #0x2b
         ror x11, x11, #0x32
         ror x21, x21, #0x14

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/src/keccak_f1600_x1_v84a_aarch64_asm.S

@@ -19,6 +19,9 @@
   Description: AArch64 ARMv8.4-A implementation of Keccak-f[1600] permutation for single state
   Signature: void mlk_keccak_f1600_x1_v84a_aarch64_asm(uint64_t state[25], const uint64_t rc[24])
   ABI:
+    Architecture: aarch64
+    CallingConvention: AAPCS64
+    Features: [NEON, SHA3]
     x0:
       type: buffer
       size_bytes: 200
@@ -91,7 +94,7 @@ MLK_ASM_FN_SYMBOL(keccak_f1600_x1_v84a_aarch64_asm)
         ldr d24, [x0, #0xc0]
         mov x2, #0x18               // =24
 
-Lkeccak_f1600_x1_v84a_loop:
+Lmlk_keccak_f1600_x1_v84a_loop:
         eor3 v30.16b, v0.16b, v5.16b, v10.16b
         eor3 v29.16b, v1.16b, v6.16b, v11.16b
         eor3 v28.16b, v2.16b, v7.16b, v12.16b
@@ -160,7 +163,7 @@ Lkeccak_f1600_x1_v84a_loop:
         bcax v4.16b, v4.16b, v27.16b, v30.16b
         eor v0.16b, v0.16b, v31.16b
         sub x2, x2, #0x1
-        cbnz x2, Lkeccak_f1600_x1_v84a_loop
+        cbnz x2, Lmlk_keccak_f1600_x1_v84a_loop
         stp d0, d1, [x0]
         stp d2, d3, [x0, #0x10]
         stp d4, d5, [x0, #0x20]

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/src/keccak_f1600_x2_v84a_aarch64_asm.S

@@ -19,6 +19,9 @@
   Description: AArch64 ARMv8.4-A implementation of Keccak-f[1600] permutation for two sequential states
   Signature: void mlk_keccak_f1600_x2_v84a_aarch64_asm(uint64_t state[50], const uint64_t rc[24])
   ABI:
+    Architecture: aarch64
+    CallingConvention: AAPCS64
+    Features: [NEON, SHA3]
     x0:
       type: buffer
       size_bytes: 400
@@ -118,7 +121,7 @@ MLK_ASM_FN_SYMBOL(keccak_f1600_x2_v84a_aarch64_asm)
         trn1 v24.2d, v25.2d, v27.2d
         mov x2, #0x18               // =24
 
-Lkeccak_f1600_x2_v84a_loop:
+Lmlk_keccak_f1600_x2_v84a_loop:
         eor3 v30.16b, v0.16b, v5.16b, v10.16b
         eor3 v29.16b, v1.16b, v6.16b, v11.16b
         eor3 v28.16b, v2.16b, v7.16b, v12.16b
@@ -187,7 +190,7 @@ Lkeccak_f1600_x2_v84a_loop:
         bcax v4.16b, v4.16b, v27.16b, v30.16b
         eor v0.16b, v0.16b, v31.16b
         sub x2, x2, #0x1
-        cbnz x2, Lkeccak_f1600_x2_v84a_loop
+        cbnz x2, Lmlk_keccak_f1600_x2_v84a_loop
         sub x0, x0, #0xc0
         add x2, x0, #0xc8
         trn1 v25.2d, v0.2d, v1.2d

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/src/keccak_f1600_x4_v8a_scalar_hybrid_aarch64_asm.S

@@ -13,6 +13,9 @@
   Description: AArch64 hybrid scalar/vector implementation of Keccak-f[1600] permutation for four sequential states
   Signature: void mlk_keccak_f1600_x4_v8a_scalar_hybrid_aarch64_asm(uint64_t state[100], const uint64_t rc[24])
   ABI:
+    Architecture: aarch64
+    CallingConvention: AAPCS64
+    Features: [NEON]
     x0:
       type: buffer
       size_bytes: 800
@@ -140,7 +143,7 @@ MLK_ASM_FN_SYMBOL(keccak_f1600_x4_v8a_scalar_hybrid_aarch64_asm)
         ldr x25, [x0, #0xc0]
         sub x0, x0, #0x190
 
-Lkeccak_f1600_x4_v8a_scalar_hybrid_initial:
+Lmlk_keccak_f1600_x4_v8a_scalar_hybrid_initial:
         eor x30, x24, x25
         eor x27, x9, x10
         eor v30.16b, v0.16b, v5.16b
@@ -523,7 +526,7 @@ Lkeccak_f1600_x4_v8a_scalar_hybrid_initial:
         str x30, [sp, #0x10]
         eor v0.16b, v0.16b, v28.16b
 
-Lkeccak_f1600_x4_v8a_scalar_hybrid_loop:
+Lmlk_keccak_f1600_x4_v8a_scalar_hybrid_loop:
         eor x0, x15, x11, ror #52
         eor x0, x0, x13, ror #48
         eor v30.16b, v0.16b, v5.16b
@@ -911,8 +914,8 @@ Lkeccak_f1600_x4_v8a_scalar_hybrid_loop:
         str x30, [sp, #0x10]
         eor v0.16b, v0.16b, v28.16b
 
-Lkeccak_f1600_x4_v8a_scalar_hybrid_loop_end:
-        b.le Lkeccak_f1600_x4_v8a_scalar_hybrid_loop
+Lmlk_keccak_f1600_x4_v8a_scalar_hybrid_loop_end:
+        b.le Lmlk_keccak_f1600_x4_v8a_scalar_hybrid_loop
         ror x2, x2, #0x3d
         ror x3, x3, #0x27
         ror x4, x4, #0x36
@@ -938,7 +941,7 @@ Lkeccak_f1600_x4_v8a_scalar_hybrid_loop_end:
         ror x25, x25, #0x9
         ldr x30, [sp, #0x20]
         cmp x30, #0x1
-        b.eq Lkeccak_f1600_x4_v8a_scalar_hybrid_done
+        b.eq Lmlk_keccak_f1600_x4_v8a_scalar_hybrid_done
         mov x30, #0x1               // =1
         str x30, [sp, #0x20]
         ldr x0, [sp]
@@ -972,9 +975,9 @@ Lkeccak_f1600_x4_v8a_scalar_hybrid_loop_end:
         ldp x15, x20, [x0, #0xb0]
         ldr x25, [x0, #0xc0]
         sub x0, x0, #0x258
-        b Lkeccak_f1600_x4_v8a_scalar_hybrid_initial
+        b Lmlk_keccak_f1600_x4_v8a_scalar_hybrid_initial
 
-Lkeccak_f1600_x4_v8a_scalar_hybrid_done:
+Lmlk_keccak_f1600_x4_v8a_scalar_hybrid_done:
         ldr x0, [sp]
         add x0, x0, #0x258
         stp x1, x6, [x0]

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/src/keccak_f1600_x4_v8a_v84a_scalar_hybrid_aarch64_asm.S

@@ -13,6 +13,9 @@
   Description: AArch64 hybrid scalar/vector implementation of Keccak-f[1600] permutation for four sequential states with ARMv8.4-A optimizations
   Signature: void mlk_keccak_f1600_x4_v8a_v84a_scalar_hybrid_aarch64_asm(uint64_t state[100], const uint64_t rc[24])
   ABI:
+    Architecture: aarch64
+    CallingConvention: AAPCS64
+    Features: [NEON, SHA3]
     x0:
       type: buffer
       size_bytes: 800
@@ -142,7 +145,7 @@ MLK_ASM_FN_SYMBOL(keccak_f1600_x4_v8a_v84a_scalar_hybrid_aarch64_asm)
         ldr x25, [x0, #0xc0]
         sub x0, x0, #0x190
 
-Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_initial:
+Lmlk_keccak_f1600_x4_v8a_v84a_scalar_hybrid_initial:
         eor x30, x24, x25
         eor x27, x9, x10
         eor3 v30.16b, v0.16b, v5.16b, v10.16b
@@ -478,7 +481,7 @@ Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_initial:
         str x30, [sp, #0x10]
         eor v0.16b, v0.16b, v28.16b
 
-Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_loop:
+Lmlk_keccak_f1600_x4_v8a_v84a_scalar_hybrid_loop:
         eor x0, x15, x11, ror #52
         eor x0, x0, x13, ror #48
         eor3 v30.16b, v0.16b, v5.16b, v10.16b
@@ -819,8 +822,8 @@ Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_loop:
         str x30, [sp, #0x10]
         eor v0.16b, v0.16b, v28.16b
 
-Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_loop_end:
-        b.le Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_loop
+Lmlk_keccak_f1600_x4_v8a_v84a_scalar_hybrid_loop_end:
+        b.le Lmlk_keccak_f1600_x4_v8a_v84a_scalar_hybrid_loop
         ror x2, x2, #0x3d
         ror x3, x3, #0x27
         ror x4, x4, #0x36
@@ -846,7 +849,7 @@ Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_loop_end:
         ror x25, x25, #0x9
         ldr x30, [sp, #0x20]
         cmp x30, #0x1
-        b.eq Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_done
+        b.eq Lmlk_keccak_f1600_x4_v8a_v84a_scalar_hybrid_done
         mov x30, #0x1               // =1
         str x30, [sp, #0x20]
         ldr x0, [sp]
@@ -880,9 +883,9 @@ Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_loop_end:
         ldp x15, x20, [x0, #0xb0]
         ldr x25, [x0, #0xc0]
         sub x0, x0, #0x258
-        b Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_initial
+        b Lmlk_keccak_f1600_x4_v8a_v84a_scalar_hybrid_initial
 
-Lkeccak_f1600_x4_v8a_v84a_scalar_hybrid_done:
+Lmlk_keccak_f1600_x4_v8a_v84a_scalar_hybrid_done:
         ldr x0, [sp]
         add x0, x0, #0x258
         stp x1, x6, [x0]

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/x1_scalar.h

@@ -7,7 +7,7 @@
 #define MLK_FIPS202_NATIVE_AARCH64_X1_SCALAR_H
 
 /* Part of backend API */
-#define MLK_USE_FIPS202_X1_NATIVE
+#define MLK_USE_NATIVE_FIPS202_X1
 /* Guard for assembly file */
 #define MLK_FIPS202_AARCH64_NEED_X1_SCALAR
 

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/x1_v84a.h

@@ -11,7 +11,7 @@
 #endif
 
 /* Part of backend API */
-#define MLK_USE_FIPS202_X1_NATIVE
+#define MLK_USE_NATIVE_FIPS202_X1
 /* Guard for assembly file */
 #define MLK_FIPS202_AARCH64_NEED_X1_V84A
 
@@ -21,7 +21,8 @@
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccak_f1600_x1_native(uint64_t *state)
 {
-  if (!mlk_sys_check_capability(MLK_SYS_CAP_SHA3))
+  if (!mlk_sys_check_capability(MLK_SYS_CAP_AARCH64_NEON) ||
+      !mlk_sys_check_capability(MLK_SYS_CAP_AARCH64_SHA3))
   {
     return MLK_NATIVE_FUNC_FALLBACK;
   }

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/x2_v84a.h

@@ -11,7 +11,7 @@
 #endif
 
 /* Part of backend API */
-#define MLK_USE_FIPS202_X4_NATIVE
+#define MLK_USE_NATIVE_FIPS202_X4
 /* Guard for assembly file */
 #define MLK_FIPS202_AARCH64_NEED_X2_V84A
 
@@ -21,7 +21,8 @@
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccak_f1600_x4_native(uint64_t *state)
 {
-  if (!mlk_sys_check_capability(MLK_SYS_CAP_SHA3))
+  if (!mlk_sys_check_capability(MLK_SYS_CAP_AARCH64_NEON) ||
+      !mlk_sys_check_capability(MLK_SYS_CAP_AARCH64_SHA3))
   {
     return MLK_NATIVE_FUNC_FALLBACK;
   }

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/x4_v8a_scalar.h

@@ -7,7 +7,7 @@
 #define MLK_FIPS202_NATIVE_AARCH64_X4_V8A_SCALAR_H
 
 /* Part of backend API */
-#define MLK_USE_FIPS202_X4_NATIVE
+#define MLK_USE_NATIVE_FIPS202_X4
 /* Guard for assembly file */
 #define MLK_FIPS202_AARCH64_NEED_X4_V8A_SCALAR_HYBRID
 
@@ -17,6 +17,11 @@
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccak_f1600_x4_native(uint64_t *state)
 {
+  if (!mlk_sys_check_capability(MLK_SYS_CAP_AARCH64_NEON))
+  {
+    return MLK_NATIVE_FUNC_FALLBACK;
+  }
+
   mlk_keccak_f1600_x4_v8a_scalar_hybrid_aarch64_asm(
       state, mlk_keccakf1600_round_constants);
   return MLK_NATIVE_FUNC_SUCCESS;

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/aarch64/x4_v8a_v84a_scalar.h

@@ -11,7 +11,7 @@
 #endif
 
 /* Part of backend API */
-#define MLK_USE_FIPS202_X4_NATIVE
+#define MLK_USE_NATIVE_FIPS202_X4
 /* Guard for assembly file */
 #define MLK_FIPS202_AARCH64_NEED_X4_V8A_V84A_SCALAR_HYBRID
 
@@ -21,7 +21,8 @@
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccak_f1600_x4_native(uint64_t *state)
 {
-  if (!mlk_sys_check_capability(MLK_SYS_CAP_SHA3))
+  if (!mlk_sys_check_capability(MLK_SYS_CAP_AARCH64_NEON) ||
+      !mlk_sys_check_capability(MLK_SYS_CAP_AARCH64_SHA3))
   {
     return MLK_NATIVE_FUNC_FALLBACK;
   }

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/api.h

@@ -29,13 +29,13 @@
  * A _backend_ is a specific implementation of parts of this interface.
  *
  * You can replace 1-fold or 4-fold batched Keccak-F1600.
- * To enable, set MLK_USE_FIPS202_X1_NATIVE or MLK_USE_FIPS202_X4_NATIVE
+ * To enable, set MLK_USE_NATIVE_FIPS202_X1 or MLK_USE_NATIVE_FIPS202_X4
  * in your backend, and define the inline wrappers mlk_keccak_f1600_x1_native()
  * and/or mlk_keccak_f1600_x4_native(), respectively, to forward to your
  * implementation.
  */
 
-#if defined(MLK_USE_FIPS202_X1_NATIVE)
+#if defined(MLK_USE_NATIVE_FIPS202_X1)
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccak_f1600_x1_native(uint64_t *state)
 __contract__(
@@ -43,8 +43,8 @@ __contract__(
   assigns(memory_slice(state, sizeof(uint64_t) * 25 * 1))
   ensures(return_value == MLK_NATIVE_FUNC_FALLBACK || return_value == MLK_NATIVE_FUNC_SUCCESS)
   ensures((return_value == MLK_NATIVE_FUNC_FALLBACK) ==> array_unchanged_u64(state, 25 * 1)));
-#endif /* MLK_USE_FIPS202_X1_NATIVE */
-#if defined(MLK_USE_FIPS202_X4_NATIVE)
+#endif /* MLK_USE_NATIVE_FIPS202_X1 */
+#if defined(MLK_USE_NATIVE_FIPS202_X4)
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccak_f1600_x4_native(uint64_t *state)
 __contract__(
@@ -52,7 +52,7 @@ __contract__(
   assigns(memory_slice(state, sizeof(uint64_t) * 25 * 4))
   ensures(return_value == MLK_NATIVE_FUNC_FALLBACK || return_value == MLK_NATIVE_FUNC_SUCCESS)
   ensures((return_value == MLK_NATIVE_FUNC_FALLBACK) ==> array_unchanged_u64(state, 25 * 4)));
-#endif /* MLK_USE_FIPS202_X4_NATIVE */
+#endif /* MLK_USE_NATIVE_FIPS202_X4 */
 
 /*
  * Native x4 XOR bytes and extract bytes interface.
@@ -66,12 +66,12 @@ __contract__(
  * NOTE: We assume that the custom representation of the zero state is the
  * all-zero state.
  *
- * MLK_USE_FIPS202_X4_XOR_BYTES_NATIVE: Backend provides native XOR bytes
- * MLK_USE_FIPS202_X4_EXTRACT_BYTES_NATIVE: Backend provides native extract
+ * MLK_USE_NATIVE_FIPS202_X4_XOR_BYTES: Backend provides native XOR bytes
+ * MLK_USE_NATIVE_FIPS202_X4_EXTRACT_BYTES: Backend provides native extract
  * bytes
  */
 
-#if defined(MLK_USE_FIPS202_X4_XOR_BYTES_NATIVE)
+#if defined(MLK_USE_NATIVE_FIPS202_X4_XOR_BYTES)
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccakf1600_xor_bytes_x4_native(
     uint64_t *state, const unsigned char *data0, const unsigned char *data1,
@@ -91,9 +91,9 @@ __contract__(
   assigns(memory_slice(state, sizeof(uint64_t) * 25 * 4))
   ensures(return_value == MLK_NATIVE_FUNC_FALLBACK || return_value == MLK_NATIVE_FUNC_SUCCESS)
   ensures((return_value == MLK_NATIVE_FUNC_FALLBACK) ==> array_unchanged_u64(state, 25 * 4)));
-#endif /* MLK_USE_FIPS202_X4_XOR_BYTES_NATIVE */
+#endif /* MLK_USE_NATIVE_FIPS202_X4_XOR_BYTES */
 
-#if defined(MLK_USE_FIPS202_X4_EXTRACT_BYTES_NATIVE)
+#if defined(MLK_USE_NATIVE_FIPS202_X4_EXTRACT_BYTES)
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccakf1600_extract_bytes_x4_native(
     uint64_t *state, unsigned char *data0, unsigned char *data1,
@@ -112,6 +112,6 @@ __contract__(
   assigns(memory_slice(data2, length))
   assigns(memory_slice(data3, length))
   ensures(return_value == MLK_NATIVE_FUNC_FALLBACK || return_value == MLK_NATIVE_FUNC_SUCCESS));
-#endif /* MLK_USE_FIPS202_X4_EXTRACT_BYTES_NATIVE */
+#endif /* MLK_USE_NATIVE_FIPS202_X4_EXTRACT_BYTES */
 
 #endif /* !MLK_FIPS202_NATIVE_API_H */

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/armv81m/mve.h

@@ -9,14 +9,15 @@
 #define MLK_FIPS202_NATIVE_ARMV81M
 
 /* Part of backend API */
-#define MLK_USE_FIPS202_X4_NATIVE
-#define MLK_USE_FIPS202_X4_XOR_BYTES_NATIVE
-#define MLK_USE_FIPS202_X4_EXTRACT_BYTES_NATIVE
+#define MLK_USE_NATIVE_FIPS202_X4
+#define MLK_USE_NATIVE_FIPS202_X4_XOR_BYTES
+#define MLK_USE_NATIVE_FIPS202_X4_EXTRACT_BYTES
 /* Guard for assembly file */
 #define MLK_FIPS202_ARMV81M_NEED_X4
 
 #if !defined(__ASSEMBLER__)
 #include "../api.h"
+#include "src/fips202_native_armv81m.h"
 
 /*
  * Native x4 permutation
@@ -35,42 +36,27 @@ static MLK_INLINE int mlk_keccak_f1600_x4_native(uint64_t *state)
 /*
  * Native x4 XOR bytes (with on-the-fly bit interleaving)
  */
-#define mlk_keccak_f1600_x4_state_xor_bytes \
-  MLK_NAMESPACE(keccak_f1600_x4_state_xor_bytes_asm)
-void mlk_keccak_f1600_x4_state_xor_bytes(void *state, const uint8_t *data0,
-                                         const uint8_t *data1,
-                                         const uint8_t *data2,
-                                         const uint8_t *data3, unsigned offset,
-                                         unsigned length);
-
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccakf1600_xor_bytes_x4_native(
     uint64_t *state, const uint8_t *data0, const uint8_t *data1,
     const uint8_t *data2, const uint8_t *data3, unsigned offset,
     unsigned length)
 {
-  mlk_keccak_f1600_x4_state_xor_bytes(state, data0, data1, data2, data3, offset,
-                                      length);
+  mlk_keccak_f1600_x4_state_xor_bytes_asm(state, data0, data1, data2, data3,
+                                          offset, length);
   return MLK_NATIVE_FUNC_SUCCESS;
 }
 
 /*
  * Native x4 extract bytes (with on-the-fly bit de-interleaving)
  */
-#define mlk_keccak_f1600_x4_state_extract_bytes \
-  MLK_NAMESPACE(keccak_f1600_x4_state_extract_bytes_asm)
-void mlk_keccak_f1600_x4_state_extract_bytes(void *state, uint8_t *data0,
-                                             uint8_t *data1, uint8_t *data2,
-                                             uint8_t *data3, unsigned offset,
-                                             unsigned length);
-
 MLK_MUST_CHECK_RETURN_VALUE
 static MLK_INLINE int mlk_keccakf1600_extract_bytes_x4_native(
     uint64_t *state, uint8_t *data0, uint8_t *data1, uint8_t *data2,
     uint8_t *data3, unsigned offset, unsigned length)
 {
-  mlk_keccak_f1600_x4_state_extract_bytes(state, data0, data1, data2, data3,
-                                          offset, length);
+  mlk_keccak_f1600_x4_state_extract_bytes_asm(state, data0, data1, data2, data3,
+                                              offset, length);
   return MLK_NATIVE_FUNC_SUCCESS;
 }
 

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/armv81m/src/keccak_f1600_x4_mve.S

@@ -9,6 +9,9 @@
   Description: Armv8.1-M MVE implementation of batched (x4) Keccak-f[1600] permutation using bit-interleaved state
   Signature: void mlk_keccak_f1600_x4_mve_asm(void *state, void *tmpstate, const uint32_t *rc)
   ABI:
+    Architecture: armv81m
+    CallingConvention: AAPCS32
+    Features: [MVE]
     r0:
       type: buffer
       size_bytes: 800
@@ -111,9 +114,9 @@ MLK_ASM_FN_SYMBOL(keccak_f1600_x4_mve_asm)
         vldrw.u32 q0, [r3]
         vldrw.u32 q1, [r2]
         vldrw.u32 q2, [r2, #32]
-        wls lr, lr, Lkeccak_f1600_x4_mve_asm_roundend @ imm = #0x8c0
+        wls lr, lr, Lmlk_keccak_f1600_x4_mve_asm_roundend @ imm = #0x8c0
 
-Lkeccak_f1600_x4_mve_asm_roundstart:
+Lmlk_keccak_f1600_x4_mve_asm_roundstart:
         vldrw.u32 q6, [r2, #112]
         veor q7, q6, q2
         vldrw.u32 q2, [r2, #80]
@@ -674,10 +677,10 @@ Lkeccak_f1600_x4_mve_asm_roundstart:
         veor q0, q4, q6
         vstrw.32 q0, [r5]
 
-Lkeccak_f1600_x4_mve_asm_roundend_pre:
-        le lr, Lkeccak_f1600_x4_mve_asm_roundstart @ imm = #-0x8c0
+Lmlk_keccak_f1600_x4_mve_asm_roundend_pre:
+        le lr, Lmlk_keccak_f1600_x4_mve_asm_roundstart @ imm = #-0x8c0
 
-Lkeccak_f1600_x4_mve_asm_roundend:
+Lmlk_keccak_f1600_x4_mve_asm_roundend:
         add sp, #0x80
         .cfi_adjust_cfa_offset -0x80
         vpop {d8, d9, d10, d11, d12, d13, d14, d15}

data/ext/pqcrypto/vendor/mlkem-native/mlkem/src/fips202/native/armv81m/src/{state_extract_bytes_x4_mve.S → keccak_f1600_x4_state_extract_bytes_mve.S}

@@ -9,7 +9,7 @@
 // Overview
 // ---------------------------------------------------------------------------
 // MVE/Helium implementation of KeccakF1600x4_StateExtractBytes
-// (inverse of state_xor_bytes_x4_mve.S).
+// (inverse of keccak_f1600_x4_state_xor_bytes_mve.S).
 //
 // void KeccakF1600x4_StateExtractBytes(state, d0, d1, d2, d3, offset, length)
 //
@@ -62,7 +62,7 @@
 
 /*
  * WARNING: This file is auto-derived from the mlkem-native source file
- *   dev/fips202/armv81m/src/state_extract_bytes_x4_mve.S using scripts/simpasm. Do not modify it directly.
+ *   dev/fips202/armv81m/src/keccak_f1600_x4_state_extract_bytes_mve.S using scripts/simpasm. Do not modify it directly.
  */
 
 .thumb
@@ -99,13 +99,13 @@ MLK_ASM_FN_SYMBOL(keccak_f1600_x4_state_extract_bytes_asm)
         ldr.w r10, [sp, #0x6c]
         ldr r6, [sp, #0x70]
         cmp r6, #0x0
-        beq.w Lkeccak_f1600_x4_state_extract_bytes_asm_exit @ imm = #0x2ea
+        beq.w Lmlk_keccak_f1600_x4_state_extract_bytes_asm_exit @ imm = #0x2ea
         and r5, r10, #0x7
         bic r9, r10, #0x7
         add.w r8, r0, r9, lsl #1
         add.w r7, r8, #0x190
         cmp r5, #0x0
-        beq.w Lkeccak_f1600_x4_state_extract_bytes_asm_pre_main @ imm = #0x112
+        beq.w Lmlk_keccak_f1600_x4_state_extract_bytes_asm_pre_main @ imm = #0x112
         vldrw.u32 q0, [r8], #16
         vldrw.u32 q1, [r7], #16
         vrev32.16 q2, q0
@@ -175,22 +175,22 @@ MLK_ASM_FN_SYMBOL(keccak_f1600_x4_state_extract_bytes_asm)
         vstrbt.8 q3, [r4], #4
         subs.w r6, r6, lr
         cmp r6, #0x0
-        beq.w Lkeccak_f1600_x4_state_extract_bytes_asm_exit @ imm = #0x1cc
+        beq.w Lmlk_keccak_f1600_x4_state_extract_bytes_asm_exit @ imm = #0x1cc
         vmov q7[2], q7[0], r1, r3
         vmov q7[3], q7[1], r2, r4
-        b Lkeccak_f1600_x4_state_extract_bytes_asm_main_body @ imm = #0xe
+        b Lmlk_keccak_f1600_x4_state_extract_bytes_asm_main_body @ imm = #0xe
 
-Lkeccak_f1600_x4_state_extract_bytes_asm_pre_main:
+Lmlk_keccak_f1600_x4_state_extract_bytes_asm_pre_main:
         vmov q7[2], q7[0], r1, r3
         vmov q7[3], q7[1], r2, r4
         mov.w r12, #0x4
         vsub.i32 q7, q7, r12
 
-Lkeccak_f1600_x4_state_extract_bytes_asm_main_body:
+Lmlk_keccak_f1600_x4_state_extract_bytes_asm_main_body:
         lsr.w lr, r6, #0x3
-        wls lr, lr, Lkeccak_f1600_x4_state_extract_bytes_asm_main_loop_end @ imm = #0xb4
+        wls lr, lr, Lmlk_keccak_f1600_x4_state_extract_bytes_asm_main_loop_end @ imm = #0xb4
 
-Lkeccak_f1600_x4_state_extract_bytes_asm_main_loop_start:
+Lmlk_keccak_f1600_x4_state_extract_bytes_asm_main_loop_start:
         vldrw.u32 q0, [r8], #16
         vldrw.u32 q1, [r7], #16
         vrev32.16 q2, q0
@@ -235,11 +235,11 @@ Lkeccak_f1600_x4_state_extract_bytes_asm_main_loop_start:
         vorr q1, q1, q3
         vstrw.32 q0, [q7, #4]!
         vstrw.32 q1, [q7, #4]!
-        le lr, Lkeccak_f1600_x4_state_extract_bytes_asm_main_loop_start @ imm = #-0xb4
+        le lr, Lmlk_keccak_f1600_x4_state_extract_bytes_asm_main_loop_start @ imm = #-0xb4
 
-Lkeccak_f1600_x4_state_extract_bytes_asm_main_loop_end:
+Lmlk_keccak_f1600_x4_state_extract_bytes_asm_main_loop_end:
         ands r6, r6, #0x7
-        beq Lkeccak_f1600_x4_state_extract_bytes_asm_exit @ imm = #0xee
+        beq Lmlk_keccak_f1600_x4_state_extract_bytes_asm_exit @ imm = #0xee
         mov.w r12, #0x4
         vadd.i32 q7, q7, r12
         vmov r1, r3, q7[2], q7[0]
@@ -301,7 +301,7 @@ Lkeccak_f1600_x4_state_extract_bytes_asm_main_loop_end:
         vstrbt.8 q2, [r3], #4
         vstrbt.8 q3, [r4], #4
 
-Lkeccak_f1600_x4_state_extract_bytes_asm_exit:
+Lmlk_keccak_f1600_x4_state_extract_bytes_asm_exit:
         vpop {d8, d9, d10, d11, d12, d13, d14, d15}
         .cfi_restore d8
         .cfi_restore d9