Comment fix for mr <= 4
For micro kernels with more than 4 rows a comparison is done
against 4 to clamp the A and C pointers.
if mr < 4 then c3 and a3 are clamped.
if mr <= 4 then c4 and a4 are clamped.
The comment for mr <= 4 is corrected in this change.
PiperOrigin-RevId: 307435044
diff --git a/src/f16-gemm/6x8-aarch64-neonfp16arith-ld64.S.in b/src/f16-gemm/6x8-aarch64-neonfp16arith-ld64.S.in
index 3c1f765..348f970 100644
--- a/src/f16-gemm/6x8-aarch64-neonfp16arith-ld64.S.in
+++ b/src/f16-gemm/6x8-aarch64-neonfp16arith-ld64.S.in
@@ -82,7 +82,7 @@
ADD x12, x11, x4 // a4 = a3 + a_stride
ADD x13, x18, x7 // c4 = c3 + cm_stride
- // if mr <= 5
+ // if mr <= 4
CSEL x12, x11, x12, LS // a4 = a3
CSEL x13, x18, x13, LS // c4 = c3