Also, add a missing include of opensslconf.h so that we properly
get the OPENSSL_SYSNAME_ULTRASPARC define even in the 32-bit case.
These changes give a pretty reasonable speed boost.
On a SPARC T4-2, without these changes:
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
des cbc 39148.28k 40071.83k 40669.53k 40685.80k 40864.43k
des ede3 14358.54k 14528.02k 14611.88k 14632.96k 14641.83k
and with them:
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
des cbc 54334.03k 56280.92k 57101.01k 57549.14k 57619.80k
des ede3 20010.98k 20682.62k 20834.90k 20873.56k 20886.87k
Signed-off-by: David S. Miller <da...@davemloft.net>
---
Configure | 4 +-
crypto/des/asm/des_enc.m4 | 62 ++++++------------
crypto/des/des_sparccore.c | 155 ++++++++++++++++++++++++++++++++++++++++++++
3 files changed, 176 insertions(+), 45 deletions(-)
create mode 100644 crypto/des/des_sparccore.c
diff --git a/Configure b/Configure
index b4cbb56..15d961a 100755
--- a/Configure
+++ b/Configure
@@ -130,8 +130,8 @@ my $x86_elf_asm="$x86_asm:elf";
my $x86_64_asm="x86_64cpuid.o:x86_64-gcc.o x86_64-mont.o x86_64-mont5.o
x86_64-gf2m.o modexp512-x86_64.o::aes-x86_64.o vpaes-x86_64.o bsaes-x86_64.o
aesni-x86_64.o aesni-sha1-x86_64.o::md5-x86_64.o:sha1-x86_64.o sha256-x86_64.o
sha512-x86_64.o::rc4-x86_64.o rc4-md5-x86_64.o:::wp-x86_64.o:cmll-x86_64.o
cmll_misc.o:ghash-x86_64.o:e_padlock-x86_64.o";
my $ia64_asm="ia64cpuid.o:bn-ia64.o ia64-mont.o::aes_core.o aes_cbc.o
aes-ia64.o::md5-ia64.o:sha1-ia64.o sha256-ia64.o sha512-ia64.o::rc4-ia64.o
rc4_skey.o:::::ghash-ia64.o::void";
-my $sparcv9_asm="sparcv9cap.o sparccpuid.o:bn-sparcv9.o sparcv9-mont.o
sparcv9a-mont.o:des_enc-sparc.o fcrypt_b.o:aes_sparccore.o
aes-sparcv9.o::md5-sparcv9.o:sha1-sparcv9.o sha256-sparcv9.o
sha512-sparcv9.o::::::cmll-sparcv9.o cmll_sparccore.o:ghash-sparcv9.o::void";
-my $sparcv8_asm=":sparcv8.o:des_enc-sparc.o fcrypt_b.o:::::::::::::void";
+my $sparcv9_asm="sparcv9cap.o sparccpuid.o:bn-sparcv9.o sparcv9-mont.o
sparcv9a-mont.o:des_enc-sparc.o des_sparccore.o fcrypt_b.o:aes_sparccore.o
aes-sparcv9.o::md5-sparcv9.o:sha1-sparcv9.o sha256-sparcv9.o
sha512-sparcv9.o::::::cmll-sparcv9.o cmll_sparccore.o:ghash-sparcv9.o::void";
+my $sparcv8_asm=":sparcv8.o:des_enc-sparc.o des_sparccore.o
fcrypt_b.o:::::::::::::void";
my $alpha_asm="alphacpuid.o:bn_asm.o
alpha-mont.o:::::sha1-alpha.o:::::::ghash-alpha.o::void";
my $mips64_asm=":bn-mips.o mips-mont.o::aes_cbc.o aes-mips.o:::sha1-mips.o
sha256-mips.o sha512-mips.o::::::::";
my $mips32_asm=$mips64_asm; $mips32_asm =~ s/\s*sha512\-mips\.o//;
diff --git a/crypto/des/asm/des_enc.m4 b/crypto/des/asm/des_enc.m4
index 5d323d0..a69273c 100644
--- a/crypto/des/asm/des_enc.m4
+++ b/crypto/des/asm/des_enc.m4
@@ -46,6 +46,8 @@
.ident "des_enc.m4 2.1"
.file "des_enc-sparc.S"
+#include <openssl/opensslconf.h>
+
#ifdef OPENSSL_FIPSCANISTER
#include <openssl/fipssyms.h>
#endif
@@ -790,18 +792,10 @@ define(load_little_endian, {
! first in memory to rightmost in register
#ifdef OPENSSL_SYSNAME_ULTRASPARC
- andcc $1, 3, global0
- bne,pn %icc, $5
- nop
-
lda [$1] 0x88, $2
add $1, 4, $4
-
- ba,pt %icc, $5a
lda [$4] 0x88, $3
-#endif
-
-$5:
+#else
ldub [$1+3], $2
ldub [$1+2], $4
@@ -830,8 +824,7 @@ $5:
ldub [$1+0+4], $4
sll $3, 8, $3
or $3, $4, $3
-$5a:
-
+#endif
})
@@ -853,19 +846,12 @@ define(load_little_endian_inc, {
! first in memory to rightmost in register
#ifdef OPENSSL_SYSNAME_ULTRASPARC
- andcc $1, 3, global0
- bne,pn %icc, $5
- nop
-
lda [$1] 0x88, $2
add $1, 4, $1
lda [$1] 0x88, $3
- ba,pt %icc, $5a
add $1, 4, $1
-#endif
-
-$5:
+#else
ldub [$1+3], $2
ldub [$1+2], $4
@@ -894,8 +880,7 @@ $5:
ldub [$1+0+4-8], $4
sll $3, 8, $3
or $3, $4, $3
-$5a:
-
+#endif
})
@@ -988,18 +973,11 @@ define(store_little_endian, {
! rightmost in register to first in memory
#ifdef OPENSSL_SYSNAME_ULTRASPARC
- andcc $1, 3, global0
- bne,pn %icc, $5
- nop
-
sta $2, [$1] 0x88
add $1, 4, $4
- ba,pt %icc, $5a
sta $3, [$4] 0x88
-#endif
-
-$5:
+#else
and $2, 255, $4
stub $4, [$1+0]
@@ -1028,9 +1006,7 @@ $5:
srl $3, 24, $4
stub $4, [$1+3+4]
-
-$5a:
-
+#endif
})
@@ -1441,15 +1417,15 @@ DES_decrypt3:
.DES_decrypt3.end:
.size DES_decrypt3,.DES_decrypt3.end-DES_decrypt3
-! void DES_ncbc_encrypt(input, output, length, schedule, ivec, enc)
+! void DES_ncbc_encrypt_asm(input, output, length, schedule, ivec, enc)
! *****************************************************************
.align 32
- .global DES_ncbc_encrypt
- .type DES_ncbc_encrypt,#function
+ .global DES_ncbc_encrypt_asm
+ .type DES_ncbc_encrypt_asm,#function
-DES_ncbc_encrypt:
+DES_ncbc_encrypt_asm:
save %sp, FRAME, %sp
@@ -1662,8 +1638,8 @@ DES_ncbc_encrypt:
store_n_bytes(local7, in2, global4, out4, local3, local4, .SNB1,
.ncbc.dec.store.iv)
-.DES_ncbc_encrypt.end:
- .size DES_ncbc_encrypt, .DES_ncbc_encrypt.end-DES_ncbc_encrypt
+.DES_ncbc_encrypt_asm.end:
+ .size DES_ncbc_encrypt_asm,
.DES_ncbc_encrypt_asm.end-DES_ncbc_encrypt_asm
! void DES_ede3_cbc_encrypt(input, output, lenght, ks1, ks2, ks3, ivec, enc)
@@ -1671,10 +1647,10 @@ DES_ncbc_encrypt:
.align 32
- .global DES_ede3_cbc_encrypt
- .type DES_ede3_cbc_encrypt,#function
+ .global DES_ede3_cbc_encrypt_asm
+ .type DES_ede3_cbc_encrypt_asm,#function
-DES_ede3_cbc_encrypt:
+DES_ede3_cbc_encrypt_asm:
save %sp, FRAME, %sp
@@ -1900,8 +1876,8 @@ DES_ede3_cbc_encrypt:
store_n_bytes(local7, in2, global4, out4, local3, local4, .SNB2,
.ede3.dec.store.iv)
-.DES_ede3_cbc_encrypt.end:
- .size
DES_ede3_cbc_encrypt,.DES_ede3_cbc_encrypt.end-DES_ede3_cbc_encrypt
+.DES_ede3_cbc_encrypt_asm.end:
+ .size
DES_ede3_cbc_encrypt_asm,.DES_ede3_cbc_encrypt_asm.end-DES_ede3_cbc_encrypt_asm
.align 256
.type .des_and,#object
diff --git a/crypto/des/des_sparccore.c b/crypto/des/des_sparccore.c
new file mode 100644
index 0000000..4aaa3c7
--- /dev/null
+++ b/crypto/des/des_sparccore.c
@@ -0,0 +1,155 @@
+#include <openssl/crypto.h>
+
+#include "des_locl.h"
+
+extern void DES_ncbc_encrypt_asm(const DES_LONG *in, DES_LONG *out, long
length,
+ DES_key_schedule *ks, DES_LONG *ivec, int enc);
+
+static void DES_ncbc_encrypt_slow(const unsigned char *in, unsigned char *out,
long length,
+ DES_key_schedule *ks, DES_cblock *ivec, int
enc)
+{
+ DES_LONG tmp_buf[2];
+ DES_LONG ivb[2];
+
+ memcpy(ivb, ivec, sizeof(*ivec));
+
+ while (length > 0) {
+ int this_len = length;
+
+ if (this_len > 8)
+ this_len = 8;
+ memcpy(tmp_buf, in, this_len);
+
+ DES_ncbc_encrypt_asm(tmp_buf, tmp_buf, this_len, ks,
+ ivb, enc);
+
+ memcpy(out, tmp_buf, this_len);
+
+ length -= 8;
+ }
+
+ memcpy(ivec, ivb, sizeof(*ivec));
+}
+
+void DES_ncbc_encrypt(const unsigned char *in, unsigned char *out, long length,
+ DES_key_schedule *ks, DES_cblock *ivec, int enc)
+{
+ const DES_LONG *aligned_in;
+ DES_LONG *aligned_ivec;
+ DES_LONG *aligned_out;
+ DES_LONG ivb[2];
+
+ aligned_out = (DES_LONG *) out;
+ if ((unsigned long) out & 0x3) {
+ aligned_out = OPENSSL_malloc(length);
+ if (!aligned_out) {
+ DES_ncbc_encrypt_slow(in, out, length, ks, ivec, enc);
+ return;
+ }
+ }
+
+ aligned_in = (const DES_LONG *) in;
+ if ((unsigned long) in & 0x3) {
+ memcpy(aligned_out, in, length);
+ aligned_in = (const DES_LONG *) aligned_out;
+ }
+
+ aligned_ivec = (DES_LONG *) ivec;
+ if ((unsigned long) ivec & 0x3) {
+ memcpy(ivb, ivec, sizeof(*ivec));
+ aligned_ivec = ivb;
+ }
+
+ DES_ncbc_encrypt_asm(aligned_in, aligned_out, length, ks,
+ aligned_ivec, enc);
+
+ if ((unsigned long) out & 0x3) {
+ memcpy(out, aligned_out, length);
+ OPENSSL_free(aligned_out);
+ }
+
+ if (aligned_ivec == ivb)
+ memcpy(ivec, ivb, sizeof(*ivec));
+}
+
+extern void DES_ede3_cbc_encrypt_asm(const DES_LONG *in, DES_LONG *out, long
length,
+ DES_key_schedule *ks1,
+ DES_key_schedule *ks2,
+ DES_key_schedule *ks3,
+ DES_LONG *ivec, int enc);
+
+static void DES_ede3_cbc_encrypt_slow(const unsigned char *in, unsigned char
*out,
+ long length,
+ DES_key_schedule *ks1,
+ DES_key_schedule *ks2,
+ DES_key_schedule *ks3,
+ DES_cblock *ivec, int enc)
+{
+ DES_LONG tmp_buf[2];
+ DES_LONG ivb[2];
+
+ memcpy(ivb, ivec, sizeof(*ivec));
+
+ while (length > 0) {
+ int this_len = length;
+
+ if (this_len > 8)
+ this_len = 8;
+ memcpy(tmp_buf, in, this_len);
+
+ DES_ede3_cbc_encrypt_asm(tmp_buf, tmp_buf, this_len,
+ ks1, ks2, ks3, ivb, enc);
+
+ memcpy(out, tmp_buf, this_len);
+
+ length -= 8;
+ }
+
+ memcpy(ivec, ivb, sizeof(*ivec));
+}
+
+void DES_ede3_cbc_encrypt(const unsigned char *in, unsigned char *out,
+ long length, DES_key_schedule *ks1,
+ DES_key_schedule *ks2, DES_key_schedule *ks3,
+ DES_cblock *ivec, int enc)
+{
+ const DES_LONG *aligned_in;
+ DES_LONG *aligned_ivec;
+ DES_LONG *aligned_out;
+ DES_LONG ivb[2];
+
+ aligned_out = (DES_LONG *) out;
+ if ((unsigned long) out & 0x3) {
+ aligned_out = OPENSSL_malloc(length);
+ if (!aligned_out) {
+ DES_ede3_cbc_encrypt_slow(in, out, length,
+ ks1, ks2, ks3,
+ ivec, enc);
+ return;
+ }
+ }
+
+ aligned_in = (const DES_LONG *) in;
+ if ((unsigned long) in & 0x3) {
+ memcpy(aligned_out, in, length);
+ aligned_in = (const DES_LONG *) aligned_out;
+ }
+
+ aligned_ivec = (DES_LONG *) ivec;
+ if ((unsigned long) ivec & 0x3) {
+ memcpy(ivb, ivec, sizeof(*ivec));
+ aligned_ivec = ivb;
+ }
+
+ DES_ede3_cbc_encrypt_asm(aligned_in, aligned_out, length,
+ ks1, ks2, ks3,
+ aligned_ivec, enc);
+
+ if ((unsigned long) out & 0x3) {
+ memcpy(out, aligned_out, length);
+ OPENSSL_free(aligned_out);
+ }
+
+ if (aligned_ivec == ivb)
+ memcpy(ivec, ivb, sizeof(*ivec));
+}
--
1.7.10.4
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List openssl-dev@openssl.org
Automated List Manager majord...@openssl.org
