symm/rijndael-x86ish-aesni.S: Load destination pointer earlier on 32-bit.

We don't need EDX for anything for most of the code, so repurpose it
earlier ready for the final store.

diff --git a/symm/rijndael-x86ish-aesni.S b/symm/rijndael-x86ish-aesni.S
index c22d23a..5293d12 100644 (file)
--- a/symm/rijndael-x86ish-aesni.S
+++ b/symm/rijndael-x86ish-aesni.S
@@ -432,6 +432,9 @@ ENDFUNC
        movdqu  xmm1, [K]
        add     K, 16
        pxor    xmm0, xmm1
+#if CPUFAM_X86
+       mov     DST, [esp + 12]
+#endif
 
        // Dispatch to the correct code.
        cmp     NR, 10
@@ -509,9 +512,6 @@ ENDFUNC
 
        // Unpermute the ciphertext block and store it.
        pshufb  xmm0, xmm5
-#if CPUFAM_X86
-       mov     DST, [esp + 12]
-#endif
        movdqu  [DST], xmm0
 
        // And we're done.