[CRYPTO] aes-i586: Nano-optimisation on key length check

Reduce the number of comparisons by one through the use of jb/je.
This patch also corrects the comments regarding the different key
lengths.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

diff --git a/arch/i386/crypto/aes-i586-asm.S b/arch/i386/crypto/aes-i586-asm.S
index 7b73c67cb4e81802cde151ce3733e2f34db7f956..911b15377f2efeded7a46a13c692770b232a79af 100644 (file)
--- a/arch/i386/crypto/aes-i586-asm.S
+++ b/arch/i386/crypto/aes-i586-asm.S
@@ -255,18 +255,17 @@ aes_enc_blk:
        xor     8(%ebp),%r4
        xor     12(%ebp),%r5
 
-       sub     $8,%esp           // space for register saves on stack
-       add     $16,%ebp          // increment to next round key
-       sub     $10,%r3          
-       je      4f              // 10 rounds for 128-bit key
-       add     $32,%ebp
-       sub     $2,%r3
-       je      3f              // 12 rounds for 128-bit key
-       add     $32,%ebp
-
-2:     fwd_rnd1( -64(%ebp) ,ft_tab)    // 14 rounds for 128-bit key
+       sub     $8,%esp         // space for register saves on stack
+       add     $16,%ebp        // increment to next round key
+       cmp     $12,%r3
+       jb      4f              // 10 rounds for 128-bit key
+       lea     32(%ebp),%ebp
+       je      3f              // 12 rounds for 192-bit key
+       lea     32(%ebp),%ebp
+
+2:     fwd_rnd1( -64(%ebp) ,ft_tab)    // 14 rounds for 256-bit key
        fwd_rnd2( -48(%ebp) ,ft_tab)
-3:     fwd_rnd1( -32(%ebp) ,ft_tab)    // 12 rounds for 128-bit key
+3:     fwd_rnd1( -32(%ebp) ,ft_tab)    // 12 rounds for 192-bit key
        fwd_rnd2( -16(%ebp) ,ft_tab)
 4:     fwd_rnd1(    (%ebp) ,ft_tab)    // 10 rounds for 128-bit key
        fwd_rnd2( +16(%ebp) ,ft_tab)
@@ -334,18 +333,17 @@ aes_dec_blk:
        xor     8(%ebp),%r4
        xor     12(%ebp),%r5
 
-       sub     $8,%esp         // space for register saves on stack
-       sub     $16,%ebp        // increment to next round key
-       sub     $10,%r3          
-       je      4f              // 10 rounds for 128-bit key
-       sub     $32,%ebp
-       sub     $2,%r3
-       je      3f              // 12 rounds for 128-bit key
-       sub     $32,%ebp
+       sub     $8,%esp         // space for register saves on stack
+       sub     $16,%ebp        // increment to next round key
+       cmp     $12,%r3
+       jb      4f              // 10 rounds for 128-bit key
+       lea     -32(%ebp),%ebp
+       je      3f              // 12 rounds for 192-bit key
+       lea     -32(%ebp),%ebp
 
-2:     inv_rnd1( +64(%ebp), it_tab)    // 14 rounds for 128-bit key
+2:     inv_rnd1( +64(%ebp), it_tab)    // 14 rounds for 256-bit key
        inv_rnd2( +48(%ebp), it_tab)
-3:     inv_rnd1( +32(%ebp), it_tab)    // 12 rounds for 128-bit key
+3:     inv_rnd1( +32(%ebp), it_tab)    // 12 rounds for 192-bit key
        inv_rnd2( +16(%ebp), it_tab)
 4:     inv_rnd1(    (%ebp), it_tab)    // 10 rounds for 128-bit key
        inv_rnd2( -16(%ebp), it_tab)