From 501728a354fc70d47ef4f69b5f181b8a38b325c9 Mon Sep 17 00:00:00 2001 From: Martin Kroeker Date: Wed, 20 Aug 2025 06:24:38 -0700 Subject: [PATCH] adjust register 20 accesses to 21 after moving x18 --- kernel/arm64/sgemm_direct_sme1_2VLx2VL.S | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/kernel/arm64/sgemm_direct_sme1_2VLx2VL.S b/kernel/arm64/sgemm_direct_sme1_2VLx2VL.S index ebbd0cadd..afb662c1f 100644 --- a/kernel/arm64/sgemm_direct_sme1_2VLx2VL.S +++ b/kernel/arm64/sgemm_direct_sme1_2VLx2VL.S @@ -62,7 +62,7 @@ add C2, N, C4 //N + SVLs add C3, C5, C4 //K*SVLs + SVLs whilelt p2.s, M_cntr, M //Tile 0,1 predicate (M dimension) - sub w20, w20, #2 //SVLs-2 + sub w21, w21, #2 //SVLs-2 .M_Loop: incw M_cntr @@ -199,7 +199,7 @@ process_K_less_than_equal_2: st1w {za1h.s[w13, #0]}, p5, [Cptr1] st1w {za2h.s[w13, #0]}, p6, [Cptr0, C6, lsl #2] st1w {za3h.s[w13, #0]}, p7, [Cptr1, C6, lsl #2] - cmp w13, w20 + cmp w13, w21 b.mi .Loop_store_ZA psel p4, p0, p2.s[w13, 1] psel p5, p1, p2.s[w13, 1]