projects / openssl.git / blobdiff
? search:
summary | shortlog | log | commit | commitdiff | tree
raw | inline | side by side
Lppc_AES_[en|de]crypt_compact: size optimization.
[openssl.git] / crypto / aes / asm / aes-ppc.pl
diff --git a/crypto/aes/asm/aes-ppc.pl b/crypto/aes/asm/aes-ppc.pl
index 685fccf04eb2929b3a50cc0614b68ce745963638..8a7a2218f646a5cfc385f4d82ac5eba2a71e67fb 100644 (file)
--- a/crypto/aes/asm/aes-ppc.pl
+++ b/crypto/aes/asm/aes-ppc.pl
@@ -12,9 +12,9 @@
 # ppc_AES_[en|de]crypt perform at 18 cycles per byte processed with
 # 128-bit key, which is ~40% better than 64-bit code generated by gcc
 # 4.0. But these are not the ones currently used! Their "compact"
-# counterparts are, for security reason. ppc_AES_crypt_compact runs at
-# 1/2 of ppc_AES_encrypt speed, while ppc_AES_decrypt_compact - at 1/3
-# of ppc_AES_decrypt.
+# counterparts are, for security reason. ppc_AES_encrypt_compact runs
+# at 1/2 of ppc_AES_encrypt speed, while ppc_AES_decrypt_compact -
+# at 1/3 of ppc_AES_decrypt.
 
 $output = shift;
 
@@ -528,17 +528,16 @@ Lppc_AES_encrypt_compact:
        addi    $Tbl1,$Tbl0,2048
        lis     $mask80,0x8080
        lis     $mask1b,0x1b1b
-       addi    $acc00,$acc00,-1
        addi    $key,$key,16
        ori     $mask80,$mask80,0x8080
        ori     $mask1b,$mask1b,0x1b1b
+       mtctr   $acc00
+.align 4
+Lenc_compact_loop:
        xor     $s0,$s0,$t0
        xor     $s1,$s1,$t1
        xor     $s2,$s2,$t2
        xor     $s3,$s3,$t3
-       mtctr   $acc00
-.align 4
-Lenc_compact_loop:
        rlwinm  $acc00,$s0,`32-24`,24,31
        rlwinm  $acc01,$s1,`32-24`,24,31
        rlwinm  $acc02,$s2,`32-24`,24,31
@@ -592,6 +591,9 @@ Lenc_compact_loop:
        or      $s2,$s2,$acc14
        or      $s3,$s3,$acc15
 
+       addi    $key,$key,16
+       bdz     Lenc_compact_done
+
        and     $acc00,$s0,$mask80      # r1=r0&0x80808080
        and     $acc01,$s1,$mask80
        and     $acc02,$s2,$mask80
@@ -650,66 +652,9 @@ Lenc_compact_loop:
        xor     $s2,$s2,$acc10
        xor     $s3,$s3,$acc11
 
-       xor     $s0,$s0,$t0
-       xor     $s1,$s1,$t1
-       xor     $s2,$s2,$t2
-       xor     $s3,$s3,$t3
-
-       addi    $key,$key,16
-       bdnz-   Lenc_compact_loop
-
-       rlwinm  $acc00,$s0,`32-24`,24,31
-       rlwinm  $acc01,$s1,`32-24`,24,31
-       rlwinm  $acc02,$s2,`32-24`,24,31
-       rlwinm  $acc03,$s3,`32-24`,24,31
-       lwz     $t0,0($key)
-       lwz     $t1,4($key)
-       lwz     $t2,8($key)
-       lwz     $t3,12($key)
-       rlwinm  $acc04,$s1,`32-16`,24,31
-       rlwinm  $acc05,$s2,`32-16`,24,31
-       rlwinm  $acc06,$s3,`32-16`,24,31
-       rlwinm  $acc07,$s0,`32-16`,24,31
-       lbzx    $acc00,$Tbl1,$acc00
-       lbzx    $acc01,$Tbl1,$acc01
-       lbzx    $acc02,$Tbl1,$acc02
-       lbzx    $acc03,$Tbl1,$acc03
-       rlwinm  $acc08,$s2,`32-8`,24,31
-       rlwinm  $acc09,$s3,`32-8`,24,31
-       rlwinm  $acc10,$s0,`32-8`,24,31
-       rlwinm  $acc11,$s1,`32-8`,24,31
-       lbzx    $acc04,$Tbl1,$acc04
-       lbzx    $acc05,$Tbl1,$acc05
-       lbzx    $acc06,$Tbl1,$acc06
-       lbzx    $acc07,$Tbl1,$acc07
-       rlwinm  $acc12,$s3,`0`,24,31
-       rlwinm  $acc13,$s0,`0`,24,31
-       rlwinm  $acc14,$s1,`0`,24,31
-       rlwinm  $acc15,$s2,`0`,24,31
-       lbzx    $acc08,$Tbl1,$acc08
-       lbzx    $acc09,$Tbl1,$acc09
-       lbzx    $acc10,$Tbl1,$acc10
-       lbzx    $acc11,$Tbl1,$acc11
-       rlwinm  $s0,$acc00,24,0,7
-       rlwinm  $s1,$acc01,24,0,7
-       rlwinm  $s2,$acc02,24,0,7
-       rlwinm  $s3,$acc03,24,0,7
-       lbzx    $acc12,$Tbl1,$acc12
-       lbzx    $acc13,$Tbl1,$acc13
-       lbzx    $acc14,$Tbl1,$acc14
-       lbzx    $acc15,$Tbl1,$acc15
-       rlwimi  $s0,$acc04,16,8,15
-       rlwimi  $s1,$acc05,16,8,15
-       rlwimi  $s2,$acc06,16,8,15
-       rlwimi  $s3,$acc07,16,8,15
-       rlwimi  $s0,$acc08,8,16,23
-       rlwimi  $s1,$acc09,8,16,23
-       rlwimi  $s2,$acc10,8,16,23
-       rlwimi  $s3,$acc11,8,16,23
-       or      $s0,$s0,$acc12
-       or      $s1,$s1,$acc13
-       or      $s2,$s2,$acc14
-       or      $s3,$s3,$acc15
+       b       Lenc_compact_loop
+.align 4
+Lenc_compact_done:
        xor     $s0,$s0,$t0
        xor     $s1,$s1,$t1
        xor     $s2,$s2,$t2
@@ -932,17 +877,16 @@ Lppc_AES_decrypt_compact:
        addi    $Tbl1,$Tbl0,2048
        lis     $mask80,0x8080
        lis     $mask1b,0x1b1b
-       addi    $acc00,$acc00,-1
        addi    $key,$key,16
        ori     $mask80,$mask80,0x8080
        ori     $mask1b,$mask1b,0x1b1b
+       mtctr   $acc00
+.align 4
+Ldec_compact_loop:
        xor     $s0,$s0,$t0
        xor     $s1,$s1,$t1
        xor     $s2,$s2,$t2
        xor     $s3,$s3,$t3
-       mtctr   $acc00
-.align 4
-Ldec_compact_loop:
        rlwinm  $acc00,$s0,`32-24`,24,31
        rlwinm  $acc01,$s1,`32-24`,24,31
        rlwinm  $acc02,$s2,`32-24`,24,31
@@ -996,6 +940,9 @@ Ldec_compact_loop:
        or      $s2,$s2,$acc14
        or      $s3,$s3,$acc15
 
+       addi    $key,$key,16
+       bdz     Ldec_compact_done
+
        and     $acc00,$s0,$mask80      # r1=r0&0x80808080
        and     $acc01,$s1,$mask80
        and     $acc02,$s2,$mask80
@@ -1140,66 +1087,9 @@ Ldec_compact_loop:
        xor     $s2,$s2,$acc10  
        xor     $s3,$s3,$acc11  
 
-       xor     $s0,$s0,$t0
-       xor     $s1,$s1,$t1
-       xor     $s2,$s2,$t2
-       xor     $s3,$s3,$t3
-
-       addi    $key,$key,16
-       bdnz-   Ldec_compact_loop
-
-       rlwinm  $acc00,$s0,`32-24`,24,31
-       rlwinm  $acc01,$s1,`32-24`,24,31
-       rlwinm  $acc02,$s2,`32-24`,24,31
-       rlwinm  $acc03,$s3,`32-24`,24,31
-       lwz     $t0,0($key)
-       lwz     $t1,4($key)
-       lwz     $t2,8($key)
-       lwz     $t3,12($key)
-       rlwinm  $acc04,$s3,`32-16`,24,31
-       rlwinm  $acc05,$s0,`32-16`,24,31
-       rlwinm  $acc06,$s1,`32-16`,24,31
-       rlwinm  $acc07,$s2,`32-16`,24,31
-       lbzx    $acc00,$Tbl1,$acc00
-       lbzx    $acc01,$Tbl1,$acc01
-       lbzx    $acc02,$Tbl1,$acc02
-       lbzx    $acc03,$Tbl1,$acc03
-       rlwinm  $acc08,$s2,`32-8`,24,31
-       rlwinm  $acc09,$s3,`32-8`,24,31
-       rlwinm  $acc10,$s0,`32-8`,24,31
-       rlwinm  $acc11,$s1,`32-8`,24,31
-       lbzx    $acc04,$Tbl1,$acc04
-       lbzx    $acc05,$Tbl1,$acc05
-       lbzx    $acc06,$Tbl1,$acc06
-       lbzx    $acc07,$Tbl1,$acc07
-       rlwinm  $acc12,$s1,`0`,24,31
-       rlwinm  $acc13,$s2,`0`,24,31
-       rlwinm  $acc14,$s3,`0`,24,31
-       rlwinm  $acc15,$s0,`0`,24,31
-       lbzx    $acc08,$Tbl1,$acc08
-       lbzx    $acc09,$Tbl1,$acc09
-       lbzx    $acc10,$Tbl1,$acc10
-       lbzx    $acc11,$Tbl1,$acc11
-       rlwinm  $s0,$acc00,24,0,7
-       rlwinm  $s1,$acc01,24,0,7
-       rlwinm  $s2,$acc02,24,0,7
-       rlwinm  $s3,$acc03,24,0,7
-       lbzx    $acc12,$Tbl1,$acc12
-       lbzx    $acc13,$Tbl1,$acc13
-       lbzx    $acc14,$Tbl1,$acc14
-       lbzx    $acc15,$Tbl1,$acc15
-       rlwimi  $s0,$acc04,16,8,15
-       rlwimi  $s1,$acc05,16,8,15
-       rlwimi  $s2,$acc06,16,8,15
-       rlwimi  $s3,$acc07,16,8,15
-       rlwimi  $s0,$acc08,8,16,23
-       rlwimi  $s1,$acc09,8,16,23
-       rlwimi  $s2,$acc10,8,16,23
-       rlwimi  $s3,$acc11,8,16,23
-       or      $s0,$s0,$acc12
-       or      $s1,$s1,$acc13
-       or      $s2,$s2,$acc14
-       or      $s3,$s3,$acc15
+       b       Ldec_compact_loop
+.align 4
+Ldec_compact_done:
        xor     $s0,$s0,$t0
        xor     $s1,$s1,$t1
        xor     $s2,$s2,$t2
Unnamed repository; edit this file 'description' to name the repository.