symm/rijndael-x86-aseni.S: Unify encryption and decryption with a macro.

diff --git a/symm/rijndael-x86-aesni.S b/symm/rijndael-x86-aesni.S
index 13c1ad56a3c9fa84d2aeb212be1585235b91684e..c0cd437adbe9af5e1dc6d398f61b3f7d1b3f7347 100644 (file)
--- a/symm/rijndael-x86-aesni.S
+++ b/symm/rijndael-x86-aesni.S
@@ -309,7 +309,8 @@ ENDFUNC
 ///--------------------------------------------------------------------------
 /// Encrypting and decrypting blocks.
 
 ///--------------------------------------------------------------------------
 /// Encrypting and decrypting blocks.
 

-FUNC(rijndael_eblk_x86_aesni)
+       .macro  encdec op, aes, koff
+FUNC(rijndael_\op\()_x86_aesni)

 
        // On entry, we have:
        // [esp +  4] points to the context block
 
        // On entry, we have:
        // [esp +  4] points to the context block


@@ -326,7 +327,7 @@ FUNC(rijndael_eblk_x86_aesni)
        movdqu  xmm0, [eax]
        pshufb  xmm0, xmm5
        mov     eax, [esp + 4]
        movdqu  xmm0, [eax]
        pshufb  xmm0, xmm5
        mov     eax, [esp + 4]

-       lea     edx, [eax + w]
+       lea     edx, [eax + \koff]

        mov     eax, [eax + nr]
 
        // Initial whitening.
        mov     eax, [eax + nr]
 
        // Initial whitening.


@@ -336,77 +337,77 @@ FUNC(rijndael_eblk_x86_aesni)
 
        // Dispatch to the correct code.
        cmp     eax, 10
 
        // Dispatch to the correct code.
        cmp     eax, 10

-       je      er10
+       je      10f

        jb      bogus
        cmp     eax, 14
        jb      bogus
        cmp     eax, 14

-       je      er14
+       je      14f

        ja      bogus
        cmp     eax, 12
        ja      bogus
        cmp     eax, 12

-       je      er12
-       jb      er11
-       jmp     er13
+       je      12f
+       jb      11f
+       jmp     13f

 
        .align  2
 
        // 14 rounds...
 
        .align  2
 
        // 14 rounds...

-er14:  movdqu  xmm1, [edx]
+14:    movdqu  xmm1, [edx]

        add     edx, 16
        add     edx, 16

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 13 rounds...
 
        // 13 rounds...

-er13:  movdqu  xmm1, [edx]
+13:    movdqu  xmm1, [edx]

        add     edx, 16
        add     edx, 16

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 12 rounds...
 
        // 12 rounds...

-er12:  movdqu  xmm1, [edx]
+12:    movdqu  xmm1, [edx]

        add     edx, 16
        add     edx, 16

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 11 rounds...
 
        // 11 rounds...

-er11:  movdqu  xmm1, [edx]
+11:    movdqu  xmm1, [edx]

        add     edx, 16
        add     edx, 16

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 10 rounds...
 
        // 10 rounds...

-er10:  movdqu  xmm1, [edx]
-       aesenc  xmm0, xmm1
+10:    movdqu  xmm1, [edx]
+       \aes    xmm0, xmm1

 
        // 9 rounds...
        movdqu  xmm1, [edx + 16]
 
        // 9 rounds...
        movdqu  xmm1, [edx + 16]

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 8 rounds...
        movdqu  xmm1, [edx + 32]
 
        // 8 rounds...
        movdqu  xmm1, [edx + 32]

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 7 rounds...
        movdqu  xmm1, [edx + 48]
 
        // 7 rounds...
        movdqu  xmm1, [edx + 48]

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 6 rounds...
        movdqu  xmm1, [edx + 64]
 
        // 6 rounds...
        movdqu  xmm1, [edx + 64]

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 5 rounds...
        movdqu  xmm1, [edx + 80]
 
        // 5 rounds...
        movdqu  xmm1, [edx + 80]

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 4 rounds...
        movdqu  xmm1, [edx + 96]
 
        // 4 rounds...
        movdqu  xmm1, [edx + 96]

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 3 rounds...
        movdqu  xmm1, [edx + 112]
 
        // 3 rounds...
        movdqu  xmm1, [edx + 112]

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // 2 rounds...
        movdqu  xmm1, [edx + 128]
 
        // 2 rounds...
        movdqu  xmm1, [edx + 128]

-       aesenc  xmm0, xmm1
+       \aes    xmm0, xmm1

 
        // Final round...
        movdqu  xmm1, [edx + 144]
 
        // Final round...
        movdqu  xmm1, [edx + 144]

-       aesenclast xmm0, xmm1
+       \aes\()last xmm0, xmm1

 
        // Unpermute the ciphertext block and store it.
        pshufb  xmm0, xmm5
 
        // Unpermute the ciphertext block and store it.
        pshufb  xmm0, xmm5


@@ -417,115 +418,10 @@ er10:    movdqu  xmm1, [edx]
        ret
 
 ENDFUNC
        ret
 
 ENDFUNC

+       .endm

 
 

-FUNC(rijndael_dblk_x86_aesni)
-
-       // On entry, we have:
-       // [esp +  4] points to the context block
-       // [esp +  8] points to the input data block
-       // [esp + 12] points to the output buffer
-
-       // Find the magic endianness-swapping table.
-       ldgot   ecx
-       movdqa  xmm5, [INTADDR(endswap_tab, ecx)]
-
-       // Load the input block and end-swap it.  Also, start loading the
-       // keys.
-       mov     eax, [esp + 8]
-       movdqu  xmm0, [eax]
-       pshufb  xmm0, xmm5
-       mov     eax, [esp + 4]
-       lea     edx, [eax + wi]
-       mov     eax, [eax + nr]
-
-       // Initial whitening.
-       movdqu  xmm1, [edx]
-       add     edx, 16
-       pxor    xmm0, xmm1
-
-       // Dispatch to the correct code.
-       cmp     eax, 10
-       je      dr10
-       jb      bogus
-       cmp     eax, 14
-       je      dr14
-       ja      bogus
-       cmp     eax, 12
-       je      dr12
-       jb      dr11
-       jmp     dr13
-
-       .align  2
-
-       // 14 rounds...
-dr14:  movdqu  xmm1, [edx]
-       add     edx, 16
-       aesdec  xmm0, xmm1
-
-       // 13 rounds...
-dr13:  movdqu  xmm1, [edx]
-       add     edx, 16
-       aesdec  xmm0, xmm1
-
-       // 12 rounds...
-dr12:  movdqu  xmm1, [edx]
-       add     edx, 16
-       aesdec  xmm0, xmm1
-
-       // 11 rounds...
-dr11:  movdqu  xmm1, [edx]
-       add     edx, 16
-       aesdec  xmm0, xmm1
-
-       // 10 rounds...
-dr10:  movdqu  xmm1, [edx]
-       aesdec  xmm0, xmm1
-
-       // 9 rounds...
-       movdqu  xmm1, [edx + 16]
-       aesdec  xmm0, xmm1
-
-       // 8 rounds...
-       movdqu  xmm1, [edx + 32]
-       aesdec  xmm0, xmm1
-
-       // 7 rounds...
-       movdqu  xmm1, [edx + 48]
-       aesdec  xmm0, xmm1
-
-       // 6 rounds...
-       movdqu  xmm1, [edx + 64]
-       aesdec  xmm0, xmm1
-
-       // 5 rounds...
-       movdqu  xmm1, [edx + 80]
-       aesdec  xmm0, xmm1
-
-       // 4 rounds...
-       movdqu  xmm1, [edx + 96]
-       aesdec  xmm0, xmm1
-
-       // 3 rounds...
-       movdqu  xmm1, [edx + 112]
-       aesdec  xmm0, xmm1
-
-       // 2 rounds...
-       movdqu  xmm1, [edx + 128]
-       aesdec  xmm0, xmm1
-
-       // Final round...
-       movdqu  xmm1, [edx + 144]
-       aesdeclast xmm0, xmm1
-
-       // Unpermute the ciphertext block and store it.
-       pshufb  xmm0, xmm5
-       mov     eax, [esp + 12]
-       movdqu  [eax], xmm0
-
-       // And we're done.
-       ret
-
-ENDFUNC
+       encdec  eblk, aesenc, w
+       encdec  dblk, aesdec, wi

 
 ///--------------------------------------------------------------------------
 /// Random utilities.
 
 ///--------------------------------------------------------------------------
 /// Random utilities.