diff options
| author | Matt Johnston <matt@ucc.asn.au> | 2011-11-30 23:03:47 +0800 |
|---|---|---|
| committer | Matt Johnston <matt@ucc.asn.au> | 2011-11-30 23:03:47 +0800 |
| commit | fa4450bee6fd8e613aa7cfc62214f4373205d422 (patch) | |
| tree | 3e0f6d97af660956d6bcdf9fb36a113326ddbb2c | |
| parent | a5cc5391e0eac9f3fbbdd38df57e8aaec5bcccae (diff) | |
| download | dropbear-fa4450bee6fd8e613aa7cfc62214f4373205d422.tar.gz | |
- More asm constraint fixes. Now seems to build OK on 32-bit OS X.
| -rw-r--r-- | tomsfastmath/src/sqr/fp_sqr_comba.c | 26 |
1 files changed, 18 insertions, 8 deletions
diff --git a/tomsfastmath/src/sqr/fp_sqr_comba.c b/tomsfastmath/src/sqr/fp_sqr_comba.c index c75f773..0c79f62 100644 --- a/tomsfastmath/src/sqr/fp_sqr_comba.c +++ b/tomsfastmath/src/sqr/fp_sqr_comba.c @@ -35,15 +35,18 @@ #define COMBA_FINI #define SQRADD(i, j) \ + do { fp_digit ti = (i); \ asm( \ "movl %6,%%eax \n\t" \ "mull %%eax \n\t" \ "addl %%eax,%0 \n\t" \ "adcl %%edx,%1 \n\t" \ "adcl $0,%2 \n\t" \ - :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i) :"%eax","%edx","%cc"); + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(ti) :"%eax","%edx","%cc"); } while (0); #define SQRADD2(i, j) \ + do { fp_digit ti = (i); \ + fp_digit tj = (j); \ asm( \ "movl %6,%%eax \n\t" \ "mull %7 \n\t" \ @@ -53,27 +56,34 @@ asm( \ "addl %%eax,%0 \n\t" \ "adcl %%edx,%1 \n\t" \ "adcl $0,%2 \n\t" \ - :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i), "m"(j) :"%eax","%edx","%cc"); + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(ti), "m"(tj) :"%eax","%edx","%cc"); } while (0); #define SQRADDSC(i, j) \ + do { \ asm( \ - "movl %6,%%eax \n\t" \ - "mull %7 \n\t" \ + "movl %2,%%eax \n\t" \ + "mull %3 \n\t" \ "movl %%eax,%0 \n\t" \ "movl %%edx,%1 \n\t" \ - "xorl %2,%2 \n\t" \ - :"=r"(sc0), "=r"(sc1), "=r"(sc2): "0"(sc0), "1"(sc1), "2"(sc2), "g"(i), "g"(j) :"%eax","%edx","%cc"); + :"=r"(sc0), "=r"(sc1): "g"(i), "g"(j) :"%eax","%edx","%cc"); \ + sc2 = 0; } while (0); #define SQRADDAC(i, j) \ + do { fp_digit ti = (i); \ + fp_digit tj = (j); \ asm( \ "movl %6,%%eax \n\t" \ "mull %7 \n\t" \ "addl %%eax,%0 \n\t" \ "adcl %%edx,%1 \n\t" \ "adcl $0,%2 \n\t" \ - :"=r"(sc0), "=r"(sc1), "=r"(sc2): "0"(sc0), "1"(sc1), "2"(sc2), "g"(i), "g"(j) :"%eax","%edx","%cc"); + :"=r"(sc0), "=r"(sc1), "=r"(sc2): "0"(sc0), "1"(sc1), "2"(sc2), "m"(ti), "m"(tj) :"%eax","%edx","%cc"); \ + } while (0); #define SQRADDDB \ + do { fp_digit tsc0 = (sc0); \ + fp_digit tsc1 = (sc1); \ + fp_digit tsc2 = (sc2); \ asm( \ "addl %6,%0 \n\t" \ "adcl %7,%1 \n\t" \ @@ -81,7 +91,7 @@ asm( \ "addl %6,%0 \n\t" \ "adcl %7,%1 \n\t" \ "adcl %8,%2 \n\t" \ - :"=r"(c0), "=r"(c1), "=r"(c2) : "0"(c0), "1"(c1), "2"(c2), "r"(sc0), "r"(sc1), "r"(sc2) : "%cc"); + :"=r"(c0), "=r"(c1), "=r"(c2) : "0"(c0), "1"(c1), "2"(c2), "m"(tsc0), "m"(tsc1), "m"(tsc2) : "%cc"); } while (0); #elif defined(TFM_X86_64) /* x86-64 optimized */ |