diff options
| author | Andy Polyakov <appro@openssl.org> | 2017-06-03 21:08:57 +0200 |
|---|---|---|
| committer | Andy Polyakov <appro@openssl.org> | 2017-06-05 14:26:38 +0200 |
| commit | 7fb37eed05059e1b135ef5818fb9f2b3b93b8daa (patch) | |
| tree | 4348c302e9e5e97a6207616b43e69ee7819394ac /crypto/ec | |
| parent | c8ec34109cab8c92685958ddfef0776a4b3b8460 (diff) | |
ec/asm/ecp_nistz256-x86_64.pl: minor sqr_montx cleanup.
Drop some redundant instructions in reduction in ecp_nistz256_sqr_montx.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit 8fc063dcc9668589fd95533d25932396d60987f9)
Diffstat (limited to 'crypto/ec')
| -rwxr-xr-x | crypto/ec/asm/ecp_nistz256-x86_64.pl | 22 |
1 files changed, 9 insertions, 13 deletions
diff --git a/crypto/ec/asm/ecp_nistz256-x86_64.pl b/crypto/ec/asm/ecp_nistz256-x86_64.pl index 16b6639b54..714e852a18 100755 --- a/crypto/ec/asm/ecp_nistz256-x86_64.pl +++ b/crypto/ec/asm/ecp_nistz256-x86_64.pl @@ -1185,19 +1185,18 @@ __ecp_nistz256_sqr_montx: adox $t1, $acc5 .byte 0x67,0x67 mulx %rdx, $t0, $t4 - mov $acc0, %rdx + mov .Lpoly+8*3(%rip), %rdx adox $t0, $acc6 shlx $a_ptr, $acc0, $t0 adox $t4, $acc7 shrx $a_ptr, $acc0, $t4 - mov .Lpoly+8*3(%rip), $t1 + mov %rdx,$t1 # reduction step 1 add $t0, $acc1 adc $t4, $acc2 - mulx $t1, $t0, $acc0 - mov $acc1, %rdx + mulx $acc0, $t0, $acc0 adc $t0, $acc3 shlx $a_ptr, $acc1, $t0 adc \$0, $acc0 @@ -1207,8 +1206,7 @@ __ecp_nistz256_sqr_montx: add $t0, $acc2 adc $t4, $acc3 - mulx $t1, $t0, $acc1 - mov $acc2, %rdx + mulx $acc1, $t0, $acc1 adc $t0, $acc0 shlx $a_ptr, $acc2, $t0 adc \$0, $acc1 @@ -1218,8 +1216,7 @@ __ecp_nistz256_sqr_montx: add $t0, $acc3 adc $t4, $acc0 - mulx $t1, $t0, $acc2 - mov $acc3, %rdx + mulx $acc2, $t0, $acc2 adc $t0, $acc1 shlx $a_ptr, $acc3, $t0 adc \$0, $acc2 @@ -1229,12 +1226,12 @@ __ecp_nistz256_sqr_montx: add $t0, $acc0 adc $t4, $acc1 - mulx $t1, $t0, $acc3 + mulx $acc3, $t0, $acc3 adc $t0, $acc2 adc \$0, $acc3 - xor $t3, $t3 # cf=0 - adc $acc0, $acc4 # accumulate upper half + xor $t3, $t3 + add $acc0, $acc4 # accumulate upper half mov .Lpoly+8*1(%rip), $a_ptr adc $acc1, $acc5 mov $acc4, $acc0 @@ -1243,8 +1240,7 @@ __ecp_nistz256_sqr_montx: mov $acc5, $acc1 adc \$0, $t3 - xor %eax, %eax # cf=0 - sbb \$-1, $acc4 # .Lpoly[0] + sub \$-1, $acc4 # .Lpoly[0] mov $acc6, $acc2 sbb $a_ptr, $acc5 # .Lpoly[1] sbb \$0, $acc6 # .Lpoly[2] |