Thumb2 ASM fixes

Make a separate AES for IAR that has AES_encrypt_block and
AES_decrypt_block inlined. Default code is relying on compiler to use
specific registers and not modify others.

Improve performance of small SP ASM code for RSA.

Author: SparkiDev

wolfcrypt/src/port/arm/thumb2-aes-asm.S

@@ -925,10 +925,10 @@ L_SHA256_transform_len_start:
 	STR	r9, [sp, #60]
 	ADD	r3, r3, #0x40
 	SUBS	r12, r12, #0x1
-#if defined(__GNUC__) || defined(__ICCARM__) || defined(__IAR_SYSTEMS_ICC__)
+#ifdef __GNUC__
 	BNE	L_SHA256_transform_len_start
 #else
-	BNE.N	L_SHA256_transform_len_start
+	BNE.W	L_SHA256_transform_len_start
 #endif
 	/* Round 0 */
 	LDR	r5, [r0, #16]
@@ -1470,10 +1470,10 @@ L_SHA256_transform_len_start:
 	SUBS	r2, r2, #0x40
 	SUB	r3, r3, #0xc0
 	ADD	r1, r1, #0x40
-#if defined(__GNUC__) || defined(__ICCARM__) || defined(__IAR_SYSTEMS_ICC__)
+#ifdef __GNUC__
 	BNE	L_SHA256_transform_len_begin
 #else
-	BNE.N	L_SHA256_transform_len_begin
+	BNE.W	L_SHA256_transform_len_begin
 #endif
 	ADD	sp, sp, #0xc0
 	POP	{r4, r5, r6, r7, r8, r9, r10, r11, pc}

wolfcrypt/src/port/arm/thumb2-aes-asm_c.c

@@ -904,10 +904,10 @@ void Transform_Sha256_Len(wc_Sha256* sha256, const byte* data, word32 len)
         "STR	r9, [sp, #60]\n\t"
         "ADD	r3, r3, #0x40\n\t"
         "SUBS	r12, r12, #0x1\n\t"
-#if defined(__GNUC__) || defined(__ICCARM__) || defined(__IAR_SYSTEMS_ICC__)
+#ifdef __GNUC__
         "BNE	L_SHA256_transform_len_start\n\t"
 #else
-        "BNE.N	L_SHA256_transform_len_start\n\t"
+        "BNE.W	L_SHA256_transform_len_start\n\t"
 #endif
         /* Round 0 */
         "LDR	r5, [%[sha256], #16]\n\t"
@@ -1449,19 +1449,20 @@ void Transform_Sha256_Len(wc_Sha256* sha256, const byte* data, word32 len)
         "SUBS	%[len], %[len], #0x40\n\t"
         "SUB	r3, r3, #0xc0\n\t"
         "ADD	%[data], %[data], #0x40\n\t"
-#if defined(__GNUC__) || defined(__ICCARM__) || defined(__IAR_SYSTEMS_ICC__)
+#ifdef __GNUC__
         "BNE	L_SHA256_transform_len_begin\n\t"
 #else
-        "BNE.N	L_SHA256_transform_len_begin\n\t"
+        "BNE.W	L_SHA256_transform_len_begin\n\t"
 #endif
         "ADD	sp, sp, #0xc0\n\t"
-        : [sha256] "+r" (sha256), [data] "+r" (data), [len] "+r" (len),
 #ifndef WOLFSSL_NO_VAR_ASSIGN_REG
-         [L_SHA256_transform_len_k] "+r" (L_SHA256_transform_len_k_c)
+        : [sha256] "+r" (sha256), [data] "+r" (data), [len] "+r" (len),
+          [L_SHA256_transform_len_k] "+r" (L_SHA256_transform_len_k_c)
+        :
 #else
-         [L_SHA256_transform_len_k] "r" (L_SHA256_transform_len_k)
+        : [sha256] "+r" (sha256), [data] "+r" (data), [len] "+r" (len)
+        : [L_SHA256_transform_len_k] "r" (L_SHA256_transform_len_k)
 #endif /* WOLFSSL_NO_VAR_ASSIGN_REG */
-        :
         : "memory", "r4", "r5", "r6", "r7", "r8", "r9", "r10", "r11", "r12"
     );
 }

wolfcrypt/src/port/arm/thumb2-sha256-asm.S

@@ -2319,10 +2319,10 @@ L_SHA512_transform_len_start:
 	STRD	r4, r5, [sp, #120]
 	ADD	r3, r3, #0x80
 	SUBS	r12, r12, #0x1
-#if defined(__GNUC__) || defined(__ICCARM__) || defined(__IAR_SYSTEMS_ICC__)
+#ifdef __GNUC__
 	BNE	L_SHA512_transform_len_start
 #else
-	BNE.N	L_SHA512_transform_len_start
+	BNE.W	L_SHA512_transform_len_start
 #endif
 	/* Round 0 */
 	LDRD	r4, r5, [r0, #32]
@@ -3656,10 +3656,10 @@ L_SHA512_transform_len_start:
 	SUBS	r2, r2, #0x80
 	SUB	r3, r3, #0x200
 	ADD	r1, r1, #0x80
-#if defined(__GNUC__) || defined(__ICCARM__) || defined(__IAR_SYSTEMS_ICC__)
+#ifdef __GNUC__
 	BNE	L_SHA512_transform_len_begin
 #else
-	BNE.N	L_SHA512_transform_len_begin
+	BNE.W	L_SHA512_transform_len_begin
 #endif
 	EOR	r0, r0, r0
 	ADD	sp, sp, #0xc0

wolfcrypt/src/port/arm/thumb2-sha256-asm_c.c

@@ -2226,10 +2226,10 @@ void Transform_Sha512_Len(wc_Sha512* sha512, const byte* data, word32 len)
         "STRD	r4, r5, [sp, #120]\n\t"
         "ADD	r3, r3, #0x80\n\t"
         "SUBS	r12, r12, #0x1\n\t"
-#if defined(__GNUC__) || defined(__ICCARM__) || defined(__IAR_SYSTEMS_ICC__)
+#ifdef __GNUC__
         "BNE	L_SHA512_transform_len_start\n\t"
 #else
-        "BNE.N	L_SHA512_transform_len_start\n\t"
+        "BNE.W	L_SHA512_transform_len_start\n\t"
 #endif
         /* Round 0 */
         "LDRD	r4, r5, [%[sha512], #32]\n\t"
@@ -3563,20 +3563,21 @@ void Transform_Sha512_Len(wc_Sha512* sha512, const byte* data, word32 len)
         "SUBS	%[len], %[len], #0x80\n\t"
         "SUB	r3, r3, #0x200\n\t"
         "ADD	%[data], %[data], #0x80\n\t"
-#if defined(__GNUC__) || defined(__ICCARM__) || defined(__IAR_SYSTEMS_ICC__)
+#ifdef __GNUC__
         "BNE	L_SHA512_transform_len_begin\n\t"
 #else
-        "BNE.N	L_SHA512_transform_len_begin\n\t"
+        "BNE.W	L_SHA512_transform_len_begin\n\t"
 #endif
         "EOR	r0, r0, r0\n\t"
         "ADD	sp, sp, #0xc0\n\t"
-        : [sha512] "+r" (sha512), [data] "+r" (data), [len] "+r" (len),
 #ifndef WOLFSSL_NO_VAR_ASSIGN_REG
-         [L_SHA512_transform_len_k] "+r" (L_SHA512_transform_len_k_c)
+        : [sha512] "+r" (sha512), [data] "+r" (data), [len] "+r" (len),
+          [L_SHA512_transform_len_k] "+r" (L_SHA512_transform_len_k_c)
+        :
 #else
-         [L_SHA512_transform_len_k] "r" (L_SHA512_transform_len_k)
+        : [sha512] "+r" (sha512), [data] "+r" (data), [len] "+r" (len)
+        : [L_SHA512_transform_len_k] "r" (L_SHA512_transform_len_k)
 #endif /* WOLFSSL_NO_VAR_ASSIGN_REG */
-        :
         : "memory", "r4", "r5", "r6", "r7", "r8", "r9", "r10", "r11", "r12"
     );
 }