Changeset ffc3b26

Timestamp:

Feb 1, 2017, 11:30:06 AM (7 years ago)

Author:

Thierry Delisle <tdelisle@…>

Branches:

ADT, aaron-thesis, arm-eh, ast-experimental, cleanup-dtors, deferred_resn, demangler, enum, forall-pointer-decay, jacob/cs343-translation, jenkins-sandbox, master, new-ast, new-ast-unique-expr, new-env, no_list, persistent-indexer, pthread-emulation, qualifiedEnum, resolv-new, with_gc

Children:

8761006c

Parents:

dd0b961

Message:

x86_64 now properly save x87 FPU and SSE control words on context switch

Location:

src/libcfa/concurrency

Files:

• CtxSwitch-x86_64.S (modified) (2 diffs)
• invoke.c (modified) (2 diffs)

src/libcfa/concurrency/CtxSwitch-x86_64.S

@@ -49 +49 @@
         // Save volatile registers on the stack.
 
+        subq   $8,%rsp
+        stmxcsr 0(%rsp)         // 4 bytes
+        fnstcw  4(%rsp)         // 2 bytes
         pushq %r15
         pushq %r14
@@ -72 +75 @@
         popq %r14
         popq %r15
+        fldcw   4(%rsp)
+        ldmxcsr 0(%rsp)
+        addq $8,%rsp
 
         // Return to thread.

src/libcfa/concurrency/invoke.c

@@ -109 +109 @@
       struct FakeStack {
             void *fixedRegisters[5];                    // fixed registers rbx, r12, r13, r14, r15
-            void *rturn;                                        // where to go on return from uSwitch
+            uint32_t mxcr;                                  // SSE Status and Control bits (control bits are preserved across function calls)
+            uint16_t fcw;                                   // X97 FPU control word (preserved across function calls)
+            void *rturn;                                      // where to go on return from uSwitch
             void *dummyReturn;                          // NULL return address to provide proper alignment
       };
@@ -120 +122 @@
       ((struct FakeStack *)(((struct machine_context_t *)stack->context)->SP))->fixedRegisters[0] = this;
       ((struct FakeStack *)(((struct machine_context_t *)stack->context)->SP))->fixedRegisters[1] = invoke;
+      ((struct FakeStack *)(((struct machine_context_t *)stack->context)->SP))->mxcr = 0; //SSE control word must be zeroed out to be valid
+      // ((struct FakeStack *)(((struct machine_context_t *)stack->context)->SP))->fcw = 0x34;
+      // ((struct FakeStack *)(((struct machine_context_t *)stack->context)->SP))->padding = 0x56;
 #else
       #error Only __i386__ and __x86_64__ is supported for threads in cfa