[libc][math] Fix floating-point test support on x86_64 Apple machines

[ddcc]

Summary

Provide platform-specific x87 FPU definitions and operations

[lntue]

libc/src/__support/FPUtil/x86_64/FEnvImpl.h

624–640

Do you mind combining the #ifdef to make it a bit easier to read:

#ifdef __APPLE__
  x87_status.status_word |= (fpstate->status_word & 0x3F);
  // We can set the x87 control word as is as there no sensitive bits.
  x87_status.control_word = fpstate->control_word;
#else
  x87_status.status_word |= (fpstate->x87_status.status_word & 0x3F);
  // Copy other non-sensitive parts of the status word.
  for (int i = 0; i < 5; i++) {
    x87_status._[i] = fpstate->x87_status._[i];
  // We can set the x87 control word as is as there no sensitive bits.
  x87_status.control_word = fpstate->x87_status.control_word;
#endif // __APPLE__

[ddcc]

Combine ifdef section

[ddcc]

libc/src/__support/FPUtil/x86_64/FEnvImpl.h

220–221

Just wanted to point out that this change will affect all x86_64 platforms. But based on the comment, I believe this function is supposed to have been fetching the exception status bits from bot hteh x87 FPU and the SSE unit anyway?

[sivachandra]

libc/src/__support/FPUtil/x86_64/FEnvImpl.h

220–221

Thanks for catching this! If the tests are passing, go ahead and submit. If you can actually craft a test case to catch regressions, even better.

[ddcc]

libc/src/__support/FPUtil/x86_64/FEnvImpl.h

220–221

Do you know if the pre-merge checks run the tests at all? I don't have a non-Apple x86_64 machine easily accessible

[lntue]

libc/src/__support/FPUtil/x86_64/FEnvImpl.h

220–221

I think the pre-merge checks are run on a Linux x86_64 machine, probably in overlay mode. Post-commit bots will run them in full build mode.

[ddcc]

libc/src/__support/FPUtil/x86_64/FEnvImpl.h

220–221

Ok, ran the tests in overlay mode locally on a Linux machine, and all passed.

libc/src/__support/FPUtil/x86_64/FEnvImpl.h

LIBC_INLINE int clear_except(int excepts) {
 
   uint32_t mxcsr = internal::get_mxcsr();
   mxcsr &= ~internal::get_status_value_for_except(excepts);
   internal::write_mxcsr(mxcsr);
   return 0;
 }
 
 LIBC_INLINE int test_except(int excepts) {
-  uint16_t status_value = internal::get_status_value_for_except(excepts);
+  uint16_t status_word = internal::get_x87_status_word();
+  uint32_t mxcsr = internal::get_mxcsr();
   // Check both x87 status word and MXCSR.
+  uint16_t status_value = internal::get_status_value_for_except(excepts);
   return internal::exception_status_to_macro(
-      static_cast<uint16_t>(status_value & internal::get_mxcsr()));
+      static_cast<uint16_t>(status_value & (status_word | mxcsr)));
 }
 
 // Sets the exception flags but does not trigger the exception handler.
 LIBC_INLINE int set_except(int excepts) {
   uint16_t status_value = internal::get_status_value_for_except(excepts);
   internal::X87StateDescriptor state;
   internal::get_x87_state_descriptor(state);
   state.status_word |= status_value;
mxcsr_control = (mxcsr_control &
                   mxcsr_value;
   internal::write_mxcsr(mxcsr_control);
 
   return 0;
 }
 
 namespace internal {
 
-#ifdef _WIN32
+#if defined(_WIN32)
 // MSVC fenv.h defines a very simple representation of the floating point state
 // which just consists of control and status words of the x87 unit.
 struct FPState {
   uint32_t control_word;
   uint32_t status_word;
 };
+#elif defined(__APPLE__)
+struct FPState {
+  uint16_t control_word;
+  uint16_t status_word;
+  uint32_t mxcsr;
+  uint8_t reserved[8];
+};
 #else
 struct FPState {
   X87StateDescriptor x87_status;
   uint32_t mxcsr;
 };
 #endif // _WIN32
 
 } // namespace internal
LIBC_INLINE int set_env(const fenv_t *envp) {
   mxcsr |= static_cast<uint16_t>((state->control_word & 0x400) << 5);
 
   internal::write_mxcsr(mxcsr);
   return 0;
 }
 #else
 LIBC_INLINE int get_env(fenv_t *envp) {
   internal::FPState *state = reinterpret_cast<internal::FPState *>(envp);
+#ifdef __APPLE__
+  internal::X87StateDescriptor x87_status;
+  internal::get_x87_state_descriptor(x87_status);
+  state->control_word = x87_status.control_word;
+  state->status_word = x87_status.status_word;
+#else
   internal::get_x87_state_descriptor(state->x87_status);
+#endif // __APPLE__
   state->mxcsr = internal::get_mxcsr();
   return 0;
 }
 
 LIBC_INLINE int set_env(const fenv_t *envp) {
   // envp contains everything including pieces like the current
   // top of FPU stack. We cannot arbitrarily change them. So, we first
   // read the current status and update only those pieces which are
if (envp == FE_DFL_ENV) {
     return 0;
   }
 
   const internal::FPState *fpstate =
       reinterpret_cast<const internal::FPState *>(envp);
 
   // Copy the exception status flags from envp.
   x87_status.status_word &= ~uint16_t(0x3F);
+#ifdef __APPLE__
+  x87_status.status_word |= (fpstate->status_word & 0x3F);
+  // We can set the x87 control word as is as there no sensitive bits.
+  x87_status.control_word = fpstate->control_word;
+#else
   x87_status.status_word |= (fpstate->x87_status.status_word & 0x3F);
   // Copy other non-sensitive parts of the status word.
   for (int i = 0; i < 5; i++)
     x87_status._[i] = fpstate->x87_status._[i];
   // We can set the x87 control word as is as there no sensitive bits.
   x87_status.control_word = fpstate->x87_status.control_word;
+#endif // __APPLE__
   internal::write_x87_state_descriptor(x87_status);
 
   // We can write the MXCSR state as is as there are no sensitive bits.
   internal::write_mxcsr(fpstate->mxcsr);
   return 0;
 }
 #endif
 
 } // namespace fputil
 } // namespace __llvm_libc
 
 #endif // LLVM_LIBC_SRC_SUPPORT_FPUTIL_X86_64_FENVIMPL_H