diff --git a/vp8/encoder/x86/fwalsh_sse2.asm b/vp8/encoder/x86/fwalsh_sse2.asm
index 145b59d45..39439f0d8 100644
--- a/vp8/encoder/x86/fwalsh_sse2.asm
+++ b/vp8/encoder/x86/fwalsh_sse2.asm
@@ -17,6 +17,7 @@ sym(vp8_short_walsh4x4_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 3
+    SAVE_XMM
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -143,6 +144,7 @@ sym(vp8_short_walsh4x4_sse2):
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
diff --git a/vp8/encoder/x86/subtract_sse2.asm b/vp8/encoder/x86/subtract_sse2.asm
index 27661e07f..3fb23d097 100644
--- a/vp8/encoder/x86/subtract_sse2.asm
+++ b/vp8/encoder/x86/subtract_sse2.asm
@@ -77,6 +77,7 @@ sym(vp8_subtract_mby_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 4
+    SAVE_XMM
     GET_GOT     rbx
     push rsi
     push rdi
@@ -138,6 +139,7 @@ submby_loop:
     pop rsi
     ; begin epilog
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret