diff --git a/src/bthread/context.cpp b/src/bthread/context.cpp
index 7f913adfc1..467445c964 100644
--- a/src/bthread/context.cpp
+++ b/src/bthread/context.cpp
@@ -347,20 +347,40 @@ __asm (
 "    pushq  %r14  \n"
 "    pushq  %r13  \n"
 "    pushq  %r12  \n"
-"    leaq  -0x8(%rsp), %rsp\n"
+"    leaq  -0xa8(%rsp), %rsp\n"
+"    movups %xmm6,  0x00(%rsp)\n"
+"    movups %xmm7,  0x10(%rsp)\n"
+"    movups %xmm8,  0x20(%rsp)\n"
+"    movups %xmm9,  0x30(%rsp)\n"
+"    movups %xmm10, 0x40(%rsp)\n"
+"    movups %xmm11, 0x50(%rsp)\n"
+"    movups %xmm12, 0x60(%rsp)\n"
+"    movups %xmm13, 0x70(%rsp)\n"
+"    movups %xmm14, 0x80(%rsp)\n"
+"    movups %xmm15, 0x90(%rsp)\n"
 "    cmp  $0, %rcx\n"
 "    je  1f\n"
-"    stmxcsr  (%rsp)\n"
-"    fnstcw   0x4(%rsp)\n"
+"    stmxcsr  0xa0(%rsp)\n"
+"    fnstcw   0xa4(%rsp)\n"
 "1:\n"
 "    movq  %rsp, (%rdi)\n"
 "    movq  %rsi, %rsp\n"
+"    movups 0x00(%rsp), %xmm6\n"
+"    movups 0x10(%rsp), %xmm7\n"
+"    movups 0x20(%rsp), %xmm8\n"
+"    movups 0x30(%rsp), %xmm9\n"
+"    movups 0x40(%rsp), %xmm10\n"
+"    movups 0x50(%rsp), %xmm11\n"
+"    movups 0x60(%rsp), %xmm12\n"
+"    movups 0x70(%rsp), %xmm13\n"
+"    movups 0x80(%rsp), %xmm14\n"
+"    movups 0x90(%rsp), %xmm15\n"
 "    cmp  $0, %rcx\n"
 "    je  2f\n"
-"    ldmxcsr  (%rsp)\n"
-"    fldcw  0x4(%rsp)\n"
+"    ldmxcsr  0xa0(%rsp)\n"
+"    fldcw    0xa4(%rsp)\n"
 "2:\n"
-"    leaq  0x8(%rsp), %rsp\n"
+"    leaq  0xa8(%rsp), %rsp\n"
 "    popq  %r12  \n"
 "    popq  %r13  \n"
 "    popq  %r14  \n"
@@ -387,12 +407,12 @@ __asm (
 "bthread_make_fcontext:\n"
 "    movq  %rdi, %rax\n"
 "    andq  $-16, %rax\n"
-"    leaq  -0x48(%rax), %rax\n"
-"    movq  %rdx, 0x38(%rax)\n"
-"    stmxcsr  (%rax)\n"
-"    fnstcw   0x4(%rax)\n"
+"    leaq  -0xe8(%rax), %rax\n"
+"    movq  %rdx, 0xd8(%rax)\n"
+"    stmxcsr  0xa0(%rax)\n"
+"    fnstcw   0xa4(%rax)\n"
 "    leaq  finish(%rip), %rcx\n"
-"    movq  %rcx, 0x40(%rax)\n"
+"    movq  %rcx, 0xe0(%rax)\n"
 "    ret \n"
 "finish:\n"
 "    xorq  %rdi, %rdi\n"
diff --git a/src/bthread/task_tracer.cpp b/src/bthread/task_tracer.cpp
index 1cc5c05a2e..a9e4910e0f 100644
--- a/src/bthread/task_tracer.cpp
+++ b/src/bthread/task_tracer.cpp
@@ -280,16 +280,16 @@ unw_cursor_t TaskTracer::MakeCursor(bthread_fcontext_t fcontext) {
 
     // Only need RBP, RIP, RSP on x86_64.
     // The base pointer (RBP).
-    if (unw_set_reg(&cursor, UNW_X86_64_RBP, regs[6]) != 0) {
+    if (unw_set_reg(&cursor, UNW_X86_64_RBP, regs[26]) != 0) {
         LOG(ERROR) << "Fail to set RBP";
     }
     // The instruction pointer (RIP).
-    if (unw_set_reg(&cursor, UNW_REG_IP, regs[7]) != 0) {
+    if (unw_set_reg(&cursor, UNW_REG_IP, regs[27]) != 0) {
         LOG(ERROR) << "Fail to set RIP";
     }
 #if UNW_VERSION_MAJOR >= 1 && UNW_VERSION_MINOR >= 7
     // The stack pointer (RSP).
-    if (unw_set_reg(&cursor, UNW_REG_SP, regs[8]) != 0) {
+    if (unw_set_reg(&cursor, UNW_REG_SP, regs[28]) != 0) {
         LOG(ERROR) << "Fail to set RSP";
     }
 #endif