This also avoids the problems with SC4.2 and unpatched SC5.
Submitted by: Andy Polyakov <appro@fy.chalmers.se>
[23-Dec-1998] down below; but in later
versions, these hyphens are gone.]
+ *) Bignum library bug fix. IRIX 6 passes "make test" now!
+ This also avoids the problems with SC4.2 and unpatched SC5.
+ [Andy Polyakov <appro@fy.chalmers.se>]
+
*) New functions sk_num, sk_value and sk_set to replace the previous macros.
These are required because of the typesafe stack would otherwise break
existing code. If old code used a structure member which used to be STACK
"debug-solaris-usparc-gcc","gcc:-O3 -g -mcpu=ultrasparc -Wall -DB_ENDIAN:-D_REENTRANT:-lsocket -lnsl:BN_LLONG RC4_CHAR DES_UNROLL BF_PTR:asm/sparcv8plus-gcc.o::",
# DO NOT use /xO[34] on sparc with SC3.0. It is broken, and will not pass the tests
-"solaris-sparc-cc","cc:-fast -O -Xa -DB_ENDIAN:-D_REENTRANT:-lsocket -lnsl:BN_LLONG RC4_CHAR DES_PTR DES_UNROLL BF_PTR:::",
+"solaris-sparc-sc3","cc:-fast -O -Xa -DB_ENDIAN:-D_REENTRANT:-lsocket -lnsl:BN_LLONG RC4_CHAR DES_PTR DES_UNROLL BF_PTR:::",
# SC4 is ok, better than gcc even on bn as long as you tell it -xarch=v8
# -fast slows things like DES down quite a lot
-# Don't use -xtarget=ultra with SC4.2. It is broken, and will break exptest.
-"solaris-sparc-sc4","cc:-xarch=v8 -xstrconst -xO5 -xdepend -Xa -DB_ENDIAN -DBN_DIV2W:-D_REENTRANT:-lsocket -lnsl:BN_LLONG RC4_CHAR DES_PTR DES_RISC1 DES_UNROLL BF_PTR:asm/sparcv8.o::",
-"solaris-usparc-sc4","cc:-xarch=v8plus -xstrconst -xO5 -xdepend -Xa -DB_ENDIAN -DBN_DIV2W:-D_REENTRANT:-lsocket -lnsl:BN_LLONG RC4_CHAR DES_PTR DES_RISC1 DES_UNROLL BF_PTR:asm/sparcv8plus.o::",
-# SC5.0 note: Compiler common patch 107357-01 or later is required!
-"solaris-usparc-sc5","cc:-xtarget=ultra -xarch=v8plus -xstrconst -xO5 -xdepend -Xa -DB_ENDIAN -DULTRASPARC -DBN_DIV2W:-D_REENTRANT:-lsocket -lnsl:BN_LLONG RC4_CHAR DES_PTR DES_RISC1 DES_UNROLL BF_PTR:asm/sparcv8plus.o:::asm/md5-sparcv8plus.o:",
-"solaris64-usparc-sc5","cc:-xtarget=ultra -xarch=v9 -xstrconst -xO5 -xdepend -Xa -DB_ENDIAN -DULTRASPARC:-D_REENTRANT:-lsocket -lnsl:SIXTY_FOUR_BIT_LONG RC4_CHAR DES_INT DES_PTR DES_RISC1 DES_UNROLL BF_PTR::::asm/md5-sparcv9.o:",
+"solaris-sparc-cc","cc:-xarch=v8 -xstrconst -xO5 -xdepend -Xa -DB_ENDIAN -DBN_DIV2W:-D_REENTRANT:-lsocket -lnsl:BN_LLONG RC4_CHAR DES_PTR DES_RISC1 DES_UNROLL BF_PTR:asm/sparcv8.o::",
+"solaris-usparc-cc","cc:-xtarget=ultra -xarch=v8plus -xstrconst -xO5 -xdepend -Xa -DB_ENDIAN -DULTRASPARC -DBN_DIV2W:-D_REENTRANT:-lsocket -lnsl:BN_LLONG RC4_CHAR DES_PTR DES_RISC1 DES_UNROLL BF_PTR:asm/sparcv8plus.o:::asm/md5-sparcv8plus.o:",
+"solaris64-usparc-cc","cc:-xtarget=ultra -xarch=v9 -xstrconst -xO5 -xdepend -Xa -DB_ENDIAN -DULTRASPARC:-D_REENTRANT:-lsocket -lnsl:SIXTY_FOUR_BIT_LONG RC4_CHAR DES_INT DES_PTR DES_RISC1 DES_UNROLL BF_PTR::::asm/md5-sparcv9.o:",
# Sunos configs, assuming sparc for the gcc one.
##"sunos-cc", "cc:-O4 -DNOPROTO -DNOCONST:(unknown)::DES_UNROLL:::",
# 3 times faster, use if at all possible.
#"irix-gcc","gcc:-O2 -mips2::SIXTY_FOUR_BIT BN_LLONG RC4_INDEX RC4_CHAR:::",
"irix-gcc","gcc:-O2 -DTERMIOS -DB_ENDIAN:(unknown)::BN_LLONG MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC2 DES_PTR BF_PTR:::",
-"irix64-gcc","gcc:-mips3 -O2 -DTERMIOS -DB_ENDIAN:(unknown)::MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC2 DES_PTR BF_PTR SIXTY_FOUR_BIT:::",
"irix-cc", "cc:-O2 -use_readonly_const -DTERMIOS -DB_ENDIAN:(unknown)::BN_LLONG DES_PTR DES_RISC2 DES_UNROLL BF_PTR:::",
-"irix64-cc", "cc:-O2 -use_readonly_const -DTERMIOS -DB_ENDIAN:(unknown)::DES_PTR DES_RISC2 DES_UNROLL BF_PTR SIXTY_FOUR_BIT:::",
+"irix-mips3-gcc","gcc:-mips3 -O2 -DTERMIOS -DB_ENDIAN:(unknown)::MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC2 DES_PTR BF_PTR SIXTY_FOUR_BIT:::",
+"irix-mips3-cc", "cc:-n32 -mips3 -O2 -use_readonly_const -DTERMIOS -DB_ENDIAN:(unknown)::DES_PTR DES_RISC2 DES_UNROLL BF_PTR SIXTY_FOUR_BIT:::",
"debug-irix-cc", "cc:-w2 -g -DCRYPTO_MDEBUG -DTERMIOS -DB_ENDIAN:(unknown):::::",
-# This is the n64 mode build.
-"irix-n64-cc", "cc:-64 -O2 -use_readonly_const -DTERMIOS:(unknown)::DES_RISC2 DES_UNROLL SIXTY_FOUR_BIT:::",
+# This is the n64 mode build. (Untested!)
+"irix64-mips4-cc", "cc:-64 -mips4 -O2 -use_readonly_const -DTERMIOS:(unknown)::DES_RISC2 DES_UNROLL SIXTY_FOUR_BIT:::",
# HPUX 9.X config.
# Don't use the bundled cc. It is broken. Use HP ANSI C if possible, or gcc.
OpenSSL STATUS Last modified at
- ______________ $Date: 1999/05/18 08:52:01 $
+ ______________ $Date: 1999/05/20 01:42:57 $
DEVELOPMENT STATE
o OpenSSL 0.9.2b: Released on March 22th, 1999
o OpenSSL 0.9.1c: Released on December 23th, 1998
- [ Proposed new numbering scheme: <major>.<minor>[<patchlevel>]
- 0.9.1c is 0913
- 1.0 is 010000
- 1.0 a is 010001
- 1.8 z is 01081a ]
-
RELEASE SHOWSTOPPERS
o BSD/OS: assembler functions must not have leading underscores
- o exptest and rsa_oaep_test fail with irix64-*
- (Don Badrak <dbadrak@geo.census.gov>: "Re: Problems to compile openssl
- on IRIX 6.2", openssl-users)
AVAILABLE PATCHES
if [ "$SYSTEM" = "SunOS" ]
then
case `cc -V 2>&1` in
- *4*) CC=sc4;;
- *5*) CC=sc5;;
- *) CC=cc;;
+ *4*) CC=cc;;
+ *5*) CC=cc;;
+ *) CC=sc3;;
esac
fi
fi
/* This is where the long long data type is 64 bits, but long is 32.
* For machines where there are 64bit registers, this is the mode to use.
* IRIX, on R4000 and above should use this mode, along with the relevent
- * assember code :-). Do NOT define BN_ULLONG.
+ * assember code :-). Do NOT define BN_LLONG.
*/
#ifdef SIXTY_FOUR_BIT
-#define BN_LLONG
-/* #define BN_ULLONG unsigned long long */
+#undef BN_LLONG
+#undef BN_ULLONG
#define BN_ULONG unsigned long long
#define BN_LONG long long
#define BN_BITS 128
int BN_num_bits_word(BN_ULONG l)
{
- static char bits[256]={
+ static const char bits[256]={
0,1,2,2,3,3,3,3,4,4,4,4,4,4,4,4,
5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,
6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
BIGNUM *bn_expand2(BIGNUM *b, int words)
{
- BN_ULONG *A,*B,*a;
- int i,j;
+ BN_ULONG *A,*a;
+ const BN_ULONG *B;
+ int i;
bn_check_top(b);
BNerr(BN_F_BN_EXPAND2,ERR_R_MALLOC_FAILURE);
return(NULL);
}
-memset(A,0x5c,sizeof(BN_ULONG)*(words+1));
#if 1
B=b->d;
/* Check if the previous number needs to be copied */
if (B != NULL)
{
+#if 0
/* This lot is an unrolled loop to copy b->top
* BN_ULONGs from B to A
*/
+/*
+ * I have nothing against unrolling but it's usually done for
+ * several reasons, namely:
+ * - minimize percentage of decision making code, i.e. branches;
+ * - avoid cache trashing;
+ * - make it possible to schedule loads earlier;
+ * Now let's examine the code below. The cornerstone of C is
+ * "programmer is always right" and that's what we love it for:-)
+ * For this very reason C compilers have to be paranoid when it
+ * comes to data aliasing and assume the worst. Yeah, but what
+ * does it mean in real life? This means that loop body below will
+ * be compiled to sequence of loads immediately followed by stores
+ * as compiler assumes the worst, something in A==B+1 style. As a
+ * result CPU pipeline is going to starve for incoming data. Secondly
+ * if A and B happen to share same cache line such code is going to
+ * cause severe cache trashing. Both factors have severe impact on
+ * performance of modern CPUs and this is the reason why this
+ * particulare piece of code is #ifdefed away and replaced by more
+ * "friendly" version found in #else section below. This comment
+ * also applies to BN_copy function.
+ *
+ * <appro@fy.chalmers.se>
+ */
for (i=b->top&(~7); i>0; i-=8)
{
A[0]=B[0]; A[1]=B[1]; A[2]=B[2]; A[3]=B[3];
*/
;
}
+#else
+ for (i=b->top>>2; i>0; i--,A+=4,B+=4)
+ {
+ /*
+ * The fact that the loop is unrolled
+ * 4-wise is a tribute to Intel. It's
+ * the one that doesn't have enough
+ * registers to accomodate more data.
+ * I'd unroll it 8-wise otherwise:-)
+ *
+ * <appro@fy.chalmers.se>
+ */
+ BN_ULONG a0,a1,a2,a3;
+ a0=B[0]; a1=B[1]; a2=B[2]; a3=B[3];
+ A[0]=a0; A[1]=a1; A[2]=a2; A[3]=a3;
+ }
+ switch (b->top&3)
+ {
+ case 3: A[2]=B[2];
+ case 2: A[1]=B[1];
+ case 1: A[0]=B[0];
+ case 0: ; /* ultrix cc workaround, see above */
+ }
+#endif
Free(b->d);
}
/* Now need to zero any data between b->top and b->max */
- B= &(b->d[b->top]);
- j=(b->max - b->top) & ~7;
- for (i=0; i<j; i+=8)
+ A= &(b->d[b->top]);
+ for (i=(b->max - b->top)>>3; i>0; i--,A+=8)
{
- B[0]=0; B[1]=0; B[2]=0; B[3]=0;
- B[4]=0; B[5]=0; B[6]=0; B[7]=0;
- B+=8;
- }
- j=(b->max - b->top) & 7;
- for (i=0; i<j; i++)
- {
- B[0]=0;
- B++;
+ A[0]=0; A[1]=0; A[2]=0; A[3]=0;
+ A[4]=0; A[5]=0; A[6]=0; A[7]=0;
}
+ for (i=(b->max - b->top)&7; i>0; i--,A++)
+ A[0]=0;
#else
- memcpy(a->d,b->d,sizeof(b->d[0])*b->top);
+ memset(A,0,sizeof(BN_ULONG)*(words+1));
+ memcpy(A,b->d,sizeof(b->d[0])*b->top);
+ b->d=a;
+ b->max=words;
#endif
/* memset(&(p[b->max]),0,((words+1)-b->max)*sizeof(BN_ULONG)); */
BIGNUM *BN_copy(BIGNUM *a, BIGNUM *b)
{
int i;
- BN_ULONG *A,*B;
+ BN_ULONG *A;
+ const BN_ULONG *B;
bn_check_top(b);
#if 1
A=a->d;
B=b->d;
- for (i=b->top&(~7); i>0; i-=8)
+ for (i=b->top>>2; i>0; i--,A+=4,B+=4)
{
- A[0]=B[0];
- A[1]=B[1];
- A[2]=B[2];
- A[3]=B[3];
- A[4]=B[4];
- A[5]=B[5];
- A[6]=B[6];
- A[7]=B[7];
- A+=8;
- B+=8;
+ BN_ULONG a0,a1,a2,a3;
+ a0=B[0]; a1=B[1]; a2=B[2]; a3=B[3];
+ A[0]=a0; A[1]=a1; A[2]=a2; A[3]=a3;
}
- switch (b->top&7)
+ switch (b->top&3)
{
- case 7:
- A[6]=B[6];
- case 6:
- A[5]=B[5];
- case 5:
- A[4]=B[4];
- case 4:
- A[3]=B[3];
- case 3:
- A[2]=B[2];
- case 2:
- A[1]=B[1];
- case 1:
- A[0]=B[0];
- case 0:
- /* I need the 'case 0' entry for utrix cc.
- * If the optimiser is turned on, it does the
- * switch table by doing
- * a=top&7
- * a--;
- * goto jump_table[a];
- * If top is 0, this makes us jump to 0xffffffc which is
- * rather bad :-(.
- * eric 23-Apr-1998
- */
- ;
+ case 3: A[2]=B[2];
+ case 2: A[1]=B[1];
+ case 1: A[0]=B[0];
+ case 0: ; /* ultrix cc workaround, see comments in bn_expand2 */
}
#else
memcpy(a->d,b->d,sizeof(b->d[0])*b->top);
#ifndef SIXTY_FOUR_BIT /* the data item > unsigned long */
ret<<=BN_BITS4; /* stops the compiler complaining */
ret<<=BN_BITS4;
+#else
+ ret=0;
#endif
ret|=a->d[i];
}
#ifndef SIXTY_FOUR_BIT /* the data item > unsigned long */
w>>=BN_BITS4;
w>>=BN_BITS4;
+#else
+ w=0;
#endif
a->d[i]=(BN_ULONG)w&BN_MASK2;
if (a->d[i] != 0) a->top=i+1;
a->top=i+1;
}
- a->d[i]|=(1L<<j);
+ a->d[i]|=(((BN_ULONG)1)<<j);
return(1);
}
j=n%BN_BITS2;
if (a->top <= i) return(0);
- a->d[i]&=(~(1L<<j));
+ a->d[i]&=(~(((BN_ULONG)1)<<j));
bn_fix_top(a);
return(1);
}
loop: for (i=1; i<NUMPRIMES; i++)
{
/* check that rnd is a prime */
- if (BN_mod_word(rnd,(BN_LONG)primes[i]) <= 1)
+ if (BN_mod_word(rnd,(BN_ULONG)primes[i]) <= 1)
{
if (!BN_add(rnd,rnd,add)) goto err;
goto loop;
/* check that p and q are prime */
/* check that for p and q
* gcd(p-1,primes) == 1 (except for 2) */
- if ( (BN_mod_word(p,(BN_LONG)primes[i]) == 0) ||
- (BN_mod_word(q,(BN_LONG)primes[i]) == 0))
+ if ( (BN_mod_word(p,(BN_ULONG)primes[i]) == 0) ||
+ (BN_mod_word(q,(BN_ULONG)primes[i]) == 0))
{
if (!BN_add(p,p,padd)) goto err;
if (!BN_add(q,q,qadd)) goto err;