I noticed that by factoring out common rounds from the
branches of the if-statements in the encryption and
decryption functions, the executable file size goes down
significantly, for crypto/cast5.ko from 26688 bytes
to 24336 bytes (amd64).
On my test system, I saw a slight speedup. This is the
first time I'm doing such a benchmark - I found a similar
one on the crypto mailing list, and I hope I did it right?
Before:
# cryptsetup create dm-test /dev/hda2 -c cast5-cbc-plain -s 128
Passsatz eingeben:
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,43484 s, 21,5 MB/s
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,4089 s, 21,8 MB/s
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,41091 s, 21,7 MB/s
After:
# cryptsetup create dm-test /dev/hda2 -c cast5-cbc-plain -s 128
Passsatz eingeben:
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,38128 s, 22,0 MB/s
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,29486 s, 22,8 MB/s
# dd if=/dev/zero of=/dev/mapper/dm-test bs=1M count=50
52428800 Bytes (52 MB) kopiert, 2,37162 s, 22,1 MB/s
Signed-off-by: Nicolas Kaiser <[email protected]>
---
crypto/cast5.c | 74 ++++++++++++++++++-------------------------------------
1 files changed, 24 insertions(+), 50 deletions(-)
diff --git a/crypto/cast5.c b/crypto/cast5.c
index a1d2294..4a230dd 100644
--- a/crypto/cast5.c
+++ b/crypto/cast5.c
@@ -604,36 +604,23 @@ static void cast5_encrypt(struct crypto_tfm *tfm, u8 *outbuf, const u8 *inbuf)
* Rounds 3, 6, 9, 12, and 15 use f function Type 3.
*/
+ t = l; l = r; r = t ^ F1(r, Km[0], Kr[0]);
+ t = l; l = r; r = t ^ F2(r, Km[1], Kr[1]);
+ t = l; l = r; r = t ^ F3(r, Km[2], Kr[2]);
+ t = l; l = r; r = t ^ F1(r, Km[3], Kr[3]);
+ t = l; l = r; r = t ^ F2(r, Km[4], Kr[4]);
+ t = l; l = r; r = t ^ F3(r, Km[5], Kr[5]);
+ t = l; l = r; r = t ^ F1(r, Km[6], Kr[6]);
+ t = l; l = r; r = t ^ F2(r, Km[7], Kr[7]);
+ t = l; l = r; r = t ^ F3(r, Km[8], Kr[8]);
+ t = l; l = r; r = t ^ F1(r, Km[9], Kr[9]);
+ t = l; l = r; r = t ^ F2(r, Km[10], Kr[10]);
+ t = l; l = r; r = t ^ F3(r, Km[11], Kr[11]);
if (!(c->rr)) {
- t = l; l = r; r = t ^ F1(r, Km[0], Kr[0]);
- t = l; l = r; r = t ^ F2(r, Km[1], Kr[1]);
- t = l; l = r; r = t ^ F3(r, Km[2], Kr[2]);
- t = l; l = r; r = t ^ F1(r, Km[3], Kr[3]);
- t = l; l = r; r = t ^ F2(r, Km[4], Kr[4]);
- t = l; l = r; r = t ^ F3(r, Km[5], Kr[5]);
- t = l; l = r; r = t ^ F1(r, Km[6], Kr[6]);
- t = l; l = r; r = t ^ F2(r, Km[7], Kr[7]);
- t = l; l = r; r = t ^ F3(r, Km[8], Kr[8]);
- t = l; l = r; r = t ^ F1(r, Km[9], Kr[9]);
- t = l; l = r; r = t ^ F2(r, Km[10], Kr[10]);
- t = l; l = r; r = t ^ F3(r, Km[11], Kr[11]);
t = l; l = r; r = t ^ F1(r, Km[12], Kr[12]);
t = l; l = r; r = t ^ F2(r, Km[13], Kr[13]);
t = l; l = r; r = t ^ F3(r, Km[14], Kr[14]);
t = l; l = r; r = t ^ F1(r, Km[15], Kr[15]);
- } else {
- t = l; l = r; r = t ^ F1(r, Km[0], Kr[0]);
- t = l; l = r; r = t ^ F2(r, Km[1], Kr[1]);
- t = l; l = r; r = t ^ F3(r, Km[2], Kr[2]);
- t = l; l = r; r = t ^ F1(r, Km[3], Kr[3]);
- t = l; l = r; r = t ^ F2(r, Km[4], Kr[4]);
- t = l; l = r; r = t ^ F3(r, Km[5], Kr[5]);
- t = l; l = r; r = t ^ F1(r, Km[6], Kr[6]);
- t = l; l = r; r = t ^ F2(r, Km[7], Kr[7]);
- t = l; l = r; r = t ^ F3(r, Km[8], Kr[8]);
- t = l; l = r; r = t ^ F1(r, Km[9], Kr[9]);
- t = l; l = r; r = t ^ F2(r, Km[10], Kr[10]);
- t = l; l = r; r = t ^ F3(r, Km[11], Kr[11]);
}
/* c1...c64 <-- (R16,L16). (Exchange final blocks L16, R16 and
@@ -663,32 +650,19 @@ static void cast5_decrypt(struct crypto_tfm *tfm, u8 *outbuf, const u8 *inbuf)
t = l; l = r; r = t ^ F3(r, Km[14], Kr[14]);
t = l; l = r; r = t ^ F2(r, Km[13], Kr[13]);
t = l; l = r; r = t ^ F1(r, Km[12], Kr[12]);
- t = l; l = r; r = t ^ F3(r, Km[11], Kr[11]);
- t = l; l = r; r = t ^ F2(r, Km[10], Kr[10]);
- t = l; l = r; r = t ^ F1(r, Km[9], Kr[9]);
- t = l; l = r; r = t ^ F3(r, Km[8], Kr[8]);
- t = l; l = r; r = t ^ F2(r, Km[7], Kr[7]);
- t = l; l = r; r = t ^ F1(r, Km[6], Kr[6]);
- t = l; l = r; r = t ^ F3(r, Km[5], Kr[5]);
- t = l; l = r; r = t ^ F2(r, Km[4], Kr[4]);
- t = l; l = r; r = t ^ F1(r, Km[3], Kr[3]);
- t = l; l = r; r = t ^ F3(r, Km[2], Kr[2]);
- t = l; l = r; r = t ^ F2(r, Km[1], Kr[1]);
- t = l; l = r; r = t ^ F1(r, Km[0], Kr[0]);
- } else {
- t = l; l = r; r = t ^ F3(r, Km[11], Kr[11]);
- t = l; l = r; r = t ^ F2(r, Km[10], Kr[10]);
- t = l; l = r; r = t ^ F1(r, Km[9], Kr[9]);
- t = l; l = r; r = t ^ F3(r, Km[8], Kr[8]);
- t = l; l = r; r = t ^ F2(r, Km[7], Kr[7]);
- t = l; l = r; r = t ^ F1(r, Km[6], Kr[6]);
- t = l; l = r; r = t ^ F3(r, Km[5], Kr[5]);
- t = l; l = r; r = t ^ F2(r, Km[4], Kr[4]);
- t = l; l = r; r = t ^ F1(r, Km[3], Kr[3]);
- t = l; l = r; r = t ^ F3(r, Km[2], Kr[2]);
- t = l; l = r; r = t ^ F2(r, Km[1], Kr[1]);
- t = l; l = r; r = t ^ F1(r, Km[0], Kr[0]);
}
+ t = l; l = r; r = t ^ F3(r, Km[11], Kr[11]);
+ t = l; l = r; r = t ^ F2(r, Km[10], Kr[10]);
+ t = l; l = r; r = t ^ F1(r, Km[9], Kr[9]);
+ t = l; l = r; r = t ^ F3(r, Km[8], Kr[8]);
+ t = l; l = r; r = t ^ F2(r, Km[7], Kr[7]);
+ t = l; l = r; r = t ^ F1(r, Km[6], Kr[6]);
+ t = l; l = r; r = t ^ F3(r, Km[5], Kr[5]);
+ t = l; l = r; r = t ^ F2(r, Km[4], Kr[4]);
+ t = l; l = r; r = t ^ F1(r, Km[3], Kr[3]);
+ t = l; l = r; r = t ^ F3(r, Km[2], Kr[2]);
+ t = l; l = r; r = t ^ F2(r, Km[1], Kr[1]);
+ t = l; l = r; r = t ^ F1(r, Km[0], Kr[0]);
dst[0] = cpu_to_be32(r);
dst[1] = cpu_to_be32(l);
--
1.7.2.2
* Nicolas Kaiser <[email protected]>:
> I noticed that by factoring out common rounds from the
> branches of the if-statements in the encryption and
> decryption functions, the executable file size goes down
> significantly, for crypto/cast5.ko from 26688 bytes
> to 24336 bytes (amd64).
>
> On my test system, I saw a slight speedup.
Hi there!
How would I add a simple cast5 cipher speed test to tcrypt.c, like this?
case 14:
test_cipher_speed("ecb(cast5)", ENCRYPT, sec, NULL, 0,
speed_template_8);
test_cipher_speed("ecb(cast5)", DECRYPT, sec, NULL, 0,
speed_template_8);
test_cipher_speed("cbc(cast5)", ENCRYPT, sec, NULL, 0,
speed_template_8);
test_cipher_speed("cbc(cast5)", DECRYPT, sec, NULL, 0,
speed_template_8);
break;
Does that make sense? Here are my results:
# insmod crypto/tcrypt.ko mode=14 sec=30
Before:
testing speed of ecb(cast5) encryption
test 0 (64 bit key, 16 byte blocks): 118241789 operations in 30 seconds (1891868624 bytes)
test 1 (64 bit key, 64 byte blocks): 40502820 operations in 30 seconds (2592180480 bytes)
test 2 (64 bit key, 256 byte blocks): 11111145 operations in 30 seconds (2844453120 bytes)
test 3 (64 bit key, 1024 byte blocks): 2858271 operations in 30 seconds (2926869504 bytes)
test 4 (64 bit key, 8192 byte blocks): 359327 operations in 30 seconds (2943606784 bytes)
testing speed of ecb(cast5) decryption
test 0 (64 bit key, 16 byte blocks): 118443232 operations in 30 seconds (1895091712 bytes)
test 1 (64 bit key, 64 byte blocks): 40060571 operations in 30 seconds (2563876544 bytes)
test 2 (64 bit key, 256 byte blocks): 10944717 operations in 30 seconds (2801847552 bytes)
test 3 (64 bit key, 1024 byte blocks): 2811154 operations in 30 seconds (2878621696 bytes)
test 4 (64 bit key, 8192 byte blocks): 353043 operations in 30 seconds (2892128256 bytes)
testing speed of cbc(cast5) encryption
test 0 (64 bit key, 16 byte blocks): 110614202 operations in 30 seconds (1769827232 bytes)
test 1 (64 bit key, 64 byte blocks): 37260496 operations in 30 seconds (2384671744 bytes)
test 2 (64 bit key, 256 byte blocks): 10323268 operations in 30 seconds (2642756608 bytes)
test 3 (64 bit key, 1024 byte blocks): 2650259 operations in 30 seconds (2713865216 bytes)
test 4 (64 bit key, 8192 byte blocks): 333281 operations in 30 seconds (2730237952 bytes)
testing speed of cbc(cast5) decryption
test 0 (64 bit key, 16 byte blocks): 102455488 operations in 30 seconds (1639287808 bytes)
test 1 (64 bit key, 64 byte blocks): 35507599 operations in 30 seconds (2272486336 bytes)
test 2 (64 bit key, 256 byte blocks): 9962653 operations in 30 seconds (2550439168 bytes)
test 3 (64 bit key, 1024 byte blocks): 2569557 operations in 30 seconds (2631226368 bytes)
test 4 (64 bit key, 8192 byte blocks): 323085 operations in 30 seconds (2646712320 bytes)
that's in sum:
87993190 bytes/s
86877105 bytes/s
81609058 bytes/s
78267680 bytes/s
After:
testing speed of ecb(cast5) encryption
test 0 (64 bit key, 16 byte blocks): 120033632 operations in 30 seconds (1920538112 bytes)
test 1 (64 bit key, 64 byte blocks): 41484473 operations in 30 seconds (2655006272 bytes)
test 2 (64 bit key, 256 byte blocks): 11418041 operations in 30 seconds (2923018496 bytes)
test 3 (64 bit key, 1024 byte blocks): 2940940 operations in 30 seconds (3011522560 bytes)
test 4 (64 bit key, 8192 byte blocks): 369831 operations in 30 seconds (3029655552 bytes)
testing speed of ecb(cast5) decryption
test 0 (64 bit key, 16 byte blocks): 122215696 operations in 30 seconds (1955451136 bytes)
test 1 (64 bit key, 64 byte blocks): 41845591 operations in 30 seconds (2678117824 bytes)
test 2 (64 bit key, 256 byte blocks): 11478322 operations in 30 seconds (2938450432 bytes)
test 3 (64 bit key, 1024 byte blocks): 2952595 operations in 30 seconds (3023457280 bytes)
test 4 (64 bit key, 8192 byte blocks): 371305 operations in 30 seconds (3041730560 bytes)
testing speed of cbc(cast5) encryption
test 0 (64 bit key, 16 byte blocks): 112301700 operations in 30 seconds (1796827200 bytes)
test 1 (64 bit key, 64 byte blocks): 38131586 operations in 30 seconds (2440421504 bytes)
test 2 (64 bit key, 256 byte blocks): 10609299 operations in 30 seconds (2715980544 bytes)
test 3 (64 bit key, 1024 byte blocks): 2728723 operations in 30 seconds (2794212352 bytes)
test 4 (64 bit key, 8192 byte blocks): 343249 operations in 30 seconds (2811895808 bytes)
testing speed of cbc(cast5) decryption
test 0 (64 bit key, 16 byte blocks): 106721453 operations in 30 seconds (1707543248 bytes)
test 1 (64 bit key, 64 byte blocks): 37573333 operations in 30 seconds (2404693312 bytes)
test 2 (64 bit key, 256 byte blocks): 10624511 operations in 30 seconds (2719874816 bytes)
test 3 (64 bit key, 1024 byte blocks): 2745305 operations in 30 seconds (2811192320 bytes)
test 4 (64 bit key, 8192 byte blocks): 345642 operations in 30 seconds (2831499264 bytes)
that's in sum:
90264940 bytes/s
90914715 bytes/s
83728916 bytes/s
83165353 bytes/s
Best regards,
Nicolas Kaiser
On Wed, Oct 27, 2010 at 12:03:24PM +0200, Nicolas Kaiser wrote:
> I noticed that by factoring out common rounds from the
> branches of the if-statements in the encryption and
> decryption functions, the executable file size goes down
> significantly, for crypto/cast5.ko from 26688 bytes
> to 24336 bytes (amd64).
Patch applied. Thanks a lot!
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt