UPDATE : valgrind version update and add files for SDK package

[sdk/tools/upstream/valgrind.git] / coregrind / m_syswrap / syswrap-amd64-darwin.c

diff --git a/coregrind/m_syswrap/syswrap-amd64-darwin.c b/coregrind/m_syswrap/syswrap-amd64-darwin.c
index 0863abf..018c4bb 100644 (file)
--- a/coregrind/m_syswrap/syswrap-amd64-darwin.c
+++ b/coregrind/m_syswrap/syswrap-amd64-darwin.c
@@ -7,7 +7,7 @@
    This file is part of Valgrind, a dynamic binary instrumentation
    framework.
 
-   Copyright (C) 2005-2011 Apple Inc.
+   Copyright (C) 2005-2013 Apple Inc.
       Greg Parker  gparker@apple.com
 
    This program is free software; you can redistribute it and/or
@@ -97,8 +97,23 @@ static void x86_float_state64_from_vex(x86_float_state64_t *mach,
                                        VexGuestAMD64State *vex)
 {
    // DDD: #warning GrP fixme fp state
-
-   VG_(memcpy)(&mach->__fpu_xmm0, &vex->guest_XMM0, 16 * sizeof(mach->__fpu_xmm0));
+   // JRS: what about the YMMHI bits?  Are they important?
+   VG_(memcpy)(&mach->__fpu_xmm0,  &vex->guest_YMM0,   sizeof(mach->__fpu_xmm0));
+   VG_(memcpy)(&mach->__fpu_xmm1,  &vex->guest_YMM1,   sizeof(mach->__fpu_xmm1));
+   VG_(memcpy)(&mach->__fpu_xmm2,  &vex->guest_YMM2,   sizeof(mach->__fpu_xmm2));
+   VG_(memcpy)(&mach->__fpu_xmm3,  &vex->guest_YMM3,   sizeof(mach->__fpu_xmm3));
+   VG_(memcpy)(&mach->__fpu_xmm4,  &vex->guest_YMM4,   sizeof(mach->__fpu_xmm4));
+   VG_(memcpy)(&mach->__fpu_xmm5,  &vex->guest_YMM5,   sizeof(mach->__fpu_xmm5));
+   VG_(memcpy)(&mach->__fpu_xmm6,  &vex->guest_YMM6,   sizeof(mach->__fpu_xmm6));
+   VG_(memcpy)(&mach->__fpu_xmm7,  &vex->guest_YMM7,   sizeof(mach->__fpu_xmm7));
+   VG_(memcpy)(&mach->__fpu_xmm8,  &vex->guest_YMM8,   sizeof(mach->__fpu_xmm8));
+   VG_(memcpy)(&mach->__fpu_xmm9,  &vex->guest_YMM9,   sizeof(mach->__fpu_xmm9));
+   VG_(memcpy)(&mach->__fpu_xmm10, &vex->guest_YMM10,  sizeof(mach->__fpu_xmm10));
+   VG_(memcpy)(&mach->__fpu_xmm11, &vex->guest_YMM11,  sizeof(mach->__fpu_xmm11));
+   VG_(memcpy)(&mach->__fpu_xmm12, &vex->guest_YMM12,  sizeof(mach->__fpu_xmm12));
+   VG_(memcpy)(&mach->__fpu_xmm13, &vex->guest_YMM13,  sizeof(mach->__fpu_xmm13));
+   VG_(memcpy)(&mach->__fpu_xmm14, &vex->guest_YMM14,  sizeof(mach->__fpu_xmm14));
+   VG_(memcpy)(&mach->__fpu_xmm15, &vex->guest_YMM15,  sizeof(mach->__fpu_xmm15));
 }
 
 
@@ -159,8 +174,23 @@ static void x86_float_state64_to_vex(const x86_float_state64_t *mach,
                                      VexGuestAMD64State *vex)
 {
    // DDD: #warning GrP fixme fp state
-
-   VG_(memcpy)(&vex->guest_XMM0, &mach->__fpu_xmm0, 16 * sizeof(mach->__fpu_xmm0));
+   // JRS: what about the YMMHI bits?  Are they important?
+   VG_(memcpy)(&vex->guest_YMM0,  &mach->__fpu_xmm0,  sizeof(mach->__fpu_xmm0));
+   VG_(memcpy)(&vex->guest_YMM1,  &mach->__fpu_xmm1,  sizeof(mach->__fpu_xmm1));
+   VG_(memcpy)(&vex->guest_YMM2,  &mach->__fpu_xmm2,  sizeof(mach->__fpu_xmm2));
+   VG_(memcpy)(&vex->guest_YMM3,  &mach->__fpu_xmm3,  sizeof(mach->__fpu_xmm3));
+   VG_(memcpy)(&vex->guest_YMM4,  &mach->__fpu_xmm4,  sizeof(mach->__fpu_xmm4));
+   VG_(memcpy)(&vex->guest_YMM5,  &mach->__fpu_xmm5,  sizeof(mach->__fpu_xmm5));
+   VG_(memcpy)(&vex->guest_YMM6,  &mach->__fpu_xmm6,  sizeof(mach->__fpu_xmm6));
+   VG_(memcpy)(&vex->guest_YMM7,  &mach->__fpu_xmm7,  sizeof(mach->__fpu_xmm7));
+   VG_(memcpy)(&vex->guest_YMM8,  &mach->__fpu_xmm8,  sizeof(mach->__fpu_xmm8));
+   VG_(memcpy)(&vex->guest_YMM9,  &mach->__fpu_xmm9,  sizeof(mach->__fpu_xmm9));
+   VG_(memcpy)(&vex->guest_YMM10, &mach->__fpu_xmm10, sizeof(mach->__fpu_xmm10));
+   VG_(memcpy)(&vex->guest_YMM11, &mach->__fpu_xmm11, sizeof(mach->__fpu_xmm11));
+   VG_(memcpy)(&vex->guest_YMM12, &mach->__fpu_xmm12, sizeof(mach->__fpu_xmm12));
+   VG_(memcpy)(&vex->guest_YMM13, &mach->__fpu_xmm13, sizeof(mach->__fpu_xmm13));
+   VG_(memcpy)(&vex->guest_YMM14, &mach->__fpu_xmm14, sizeof(mach->__fpu_xmm14));
+   VG_(memcpy)(&vex->guest_YMM15, &mach->__fpu_xmm15, sizeof(mach->__fpu_xmm15));
 }
 
 
@@ -402,23 +432,34 @@ void wqthread_hijack(Addr self, Addr kport, Addr stackaddr, Addr workitem,
       lock. */
    VG_(acquire_BigLock_LL)("wqthread_hijack");
 
-   if (0) VG_(printf)("wqthread_hijack: self %#lx, kport %#lx, "
-                      "stackaddr %#lx, workitem %#lx, reuse %d, sp %#lx\n", 
-                      self, kport, stackaddr, workitem, reuse, sp);
+   if (0) VG_(printf)(
+             "wqthread_hijack: self %#lx, kport %#lx, "
+            "stackaddr %#lx, workitem %#lx, reuse/flags %x, sp %#lx\n", 
+            self, kport, stackaddr, workitem, reuse, sp);
 
    /* Start the thread with all signals blocked.  VG_(scheduler) will
       set the mask correctly when we finally get there. */
    VG_(sigfillset)(&blockall);
    VG_(sigprocmask)(VKI_SIG_SETMASK, &blockall, NULL);
 
-   if (reuse) {
+   /* For 10.7 and earlier, |reuse| appeared to be used as a simple
+      boolean.  In 10.8 and later its name changed to |flags| and has
+      various other bits OR-d into it too, so it's necessary to fish
+      out just the relevant parts.  Hence: */
+#  if DARWIN_VERS <= DARWIN_10_7
+   Bool is_reuse = reuse != 0;
+#  elif DARWIN_VERS == DARWIN_10_8
+   Bool is_reuse = (reuse & 0x20000 /* == WQ_FLAG_THREAD_REUSE */) != 0;
+#  endif
+
+   if (is_reuse) {
 
      /* For whatever reason, tst->os_state.pthread appear to have a
         constant offset of 96 on 10.7, but zero on 10.6 and 10.5.  No
         idea why. */
 #      if DARWIN_VERS <= DARWIN_10_6
        UWord magic_delta = 0;
-#      elif DARWIN_VERS == DARWIN_10_7
+#      elif DARWIN_VERS >= DARWIN_10_7
        UWord magic_delta = 0x60;
 #      endif
 
@@ -463,7 +504,7 @@ void wqthread_hijack(Addr self, Addr kport, Addr stackaddr, Addr workitem,
    stacksize = 512*1024;  // wq stacks are always DEFAULT_STACK_SIZE
    stack = VG_PGROUNDUP(sp) - stacksize;
 
-   if (reuse) {
+   if (is_reuse) {
       // Continue V's thread back in the scheduler. 
       // The client thread is of course in another location entirely.