From: Mark Wooding <mdw@distorted.org.uk>
Date: Thu, 7 Nov 2019 01:54:57 +0000 (+0000)
Subject: math/mpx-mul4-amd64-sse2.S: Improve the end-of-loop condition testing.
X-Git-Tag: 2.6.0~13
X-Git-Url: https://git.distorted.org.uk/~mdw/catacomb/commitdiff_plain/59aa39f8a0065d8bc3b2a353706f41f0dd20be12

math/mpx-mul4-amd64-sse2.S: Improve the end-of-loop condition testing.

Previously, I waited until `rdi' was set up for the next iteration
before comparing it against the limit.  But in fact, `DV' already has
the right value, so we can compare earlier.
---

diff --git a/math/mpx-mul4-amd64-sse2.S b/math/mpx-mul4-amd64-sse2.S
index da3e6d61..1c205a73 100644
--- a/math/mpx-mul4-amd64-sse2.S
+++ b/math/mpx-mul4-amd64-sse2.S
@@ -1270,10 +1270,10 @@ FUNC(mpxmont_redc4_amd64_sse2)
 	jb	7b
 
 	// All done for this iteration.  Start the next.
-8:	mov	rdi, DV			// -> Z = dv[i]
-	mov	rbx, NV			// -> X = nv[0]
-	cmp	rdi, DVLO		// all done yet?
+	cmp	DV, DVLO		// all done yet?
 	jae	9f
+	mov	rdi, DV			// -> Z = dv[i]
+	mov	rbx, NV			// -> X = nv[0]
 	add	DV, 16
 	call	mont4
 	add	rdi, 16