Reverted the latest two changes; update

bost src and dst pointers in the loop, and use plain addb when
updating j. These two previous changes slowed the code down on AMD
Duron.

Rev: src/nettle/x86/arcfour-crypt.asm:1.10

diff --git a/x86/arcfour-crypt.asm b/x86/arcfour-crypt.asm
index 43e4f7d304efc0e09a891ac19aeb111a9e4685c0..74dd77634b9f7cf3317b931628e63b818b8bfa62 100644 (file)
--- a/x86/arcfour-crypt.asm
+++ b/x86/arcfour-crypt.asm
@@ -47,7 +47,7 @@ C Register usage:
        C %ebx = j
        C %cl  = si
        C %ch  = sj
-       
+
        movl    24(%esp), %edx          C  length
        testl   %edx,%edx
        jz      .Lend
@@ -59,15 +59,15 @@ C Register usage:
        
        movzbl  256(%ebp), %eax         C  i
        movzbl  257(%ebp), %ebx         C  j
-       subl    %esi, %edi
 .Lloop:
 C      incb    %al
        incl    %eax
        andl    $0xff, %eax
        movzbl  (%ebp, %eax), %ecx      C  si. Clears high bytes
-C      addb    %cl, %bl
-       addl    %ecx, %ebx
-       andl    $0xff, %ebx
+       addb    %cl, %bl
+C The addl andl is preferable on PPro and PII, but slows thing down on AMD Duron.
+C      addl    %ecx, %ebx
+C      andl    $0xff, %ebx
        movb    (%ebp, %ebx), %ch       C  sj
        movb    %ch, (%ebp, %eax)       C  S[i] = sj
        movb    %cl, (%ebp, %ebx)       C  S[j] = si
@@ -76,8 +76,9 @@ C     addb    %cl, %bl
                                        C  for indexing.
        movb    (%ebp, %ecx), %cl
        xorb    (%esi), %cl
-       movb    %cl, (%esi,%edi)
        incl    %esi
+       movb    %cl, (%edi)
+       incl    %edi
        cmpl    %esi, %edx
        jne     .Lloop