X-Git-Url: https://git.distorted.org.uk/~mdw/catacomb/blobdiff_plain/92edc356a312bc64abca0c30bc03d4b6676f3d39..f79c8756d054b2979f79d8277affb988c4f39c49:/math/mpx-mul4-amd64-sse2.S

diff --git a/math/mpx-mul4-amd64-sse2.S b/math/mpx-mul4-amd64-sse2.S
index a37aba69..da3e6d61 100644
--- a/math/mpx-mul4-amd64-sse2.S
+++ b/math/mpx-mul4-amd64-sse2.S
@@ -1155,7 +1155,7 @@ FUNC(mpxmont_redc4_amd64_sse2)
 	// outer loop dv	r10	rcx
 	// outer loop dv limit	r11	r11
 	// nv base		rdx	r8
-	// nv limit		r9	r12*
+	// nv limit		r9	r10*
 	// n			rcx	r9
 	// c			rcx	r9
 
@@ -1183,14 +1183,13 @@ FUNC(mpxmont_redc4_amd64_sse2)
 #  define DV rcx
 #  define DVLO r11
 #  define NV r8
-#  define NVL r12
+#  define NVL r10
 #  define N r9
 #  define C r9d
 
 	pushreg	rbx
 	pushreg	rdi
-	pushreg	r12
-	stalloc	160
+	stalloc	168
 
 	savexmm	xmm6,    0
 	savexmm	xmm7,   16
@@ -1260,7 +1259,9 @@ FUNC(mpxmont_redc4_amd64_sse2)
 	cmp	rdi, DVL4
 	jb	0b
 
-	// Deal with the tail end.
+	// Deal with the tail end.  Note that the actual destination length
+	// won't be an exacty number of blocks of four, so it's safe to just
+	// drop through here.
 7:	add	[rdi], C
 	mov	C, 0
 	add	rdi, 4
@@ -1268,9 +1269,7 @@ FUNC(mpxmont_redc4_amd64_sse2)
 	cmp	rdi, DVL
 	jb	7b
 
-	// All done for this iteration.  Start the next.  (This must have at
-	// least one follow-on iteration, or we'd not have started this outer
-	// loop.)
+	// 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?
@@ -1300,8 +1299,7 @@ FUNC(mpxmont_redc4_amd64_sse2)
 	rstrxmm	xmm14, 128
 	rstrxmm	xmm15, 144
 
-	stfree	160
-	popreg	r12
+	stfree	168
 	popreg	rdi
 	popreg	rbx
 #endif