p256-x86_64-asm.pl: minor sqr_montx cleanup.
Drop some redundant instructions in reduction in ecp_nistz256_sqr_montx.
(Imported from upstream's 8fc063dcc9668589fd95533d25932396d60987f9.)
I believe this is a no-op for us as we do not currently enable the
ADX-based optimizations.
Change-Id: I34a5f5ffb965d59c67f6b9f0ca7937e49ba6e820
Reviewed-on: https://boringssl-review.googlesource.com/16884
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
diff --git a/crypto/fipsmodule/ec/asm/p256-x86_64-asm.pl b/crypto/fipsmodule/ec/asm/p256-x86_64-asm.pl
index 4dadd4a..1ac3d21 100755
--- a/crypto/fipsmodule/ec/asm/p256-x86_64-asm.pl
+++ b/crypto/fipsmodule/ec/asm/p256-x86_64-asm.pl
@@ -863,19 +863,18 @@
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
@@ -885,8 +884,7 @@
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
@@ -896,8 +894,7 @@
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
@@ -907,12 +904,12 @@
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
@@ -921,8 +918,7 @@
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]
