[commit: ghc] master: Reset FPU precision back to MSVCRT defaults (6f7d827)

[git at git.haskell.org]

Repository : ssh://git@git.haskell.org/ghc

On branch  : master
Link       : http://ghc.haskell.org/trac/ghc/changeset/6f7d8279cea4aa1082fb07adf5da507297e21ee8/ghc

>---------------------------------------------------------------

commit 6f7d8279cea4aa1082fb07adf5da507297e21ee8
Author: Tamar Christina <tamar at zhox.com>
Date:   Wed Dec 14 16:45:35 2016 -0500

    Reset FPU precision back to MSVCRT defaults
    
    Mingw-w64 does a stupid thing. They set the FPU precision to extended
    mode by default.  The reasoning is that it's for compatibility with GNU
    Linux ported libraries. However the problem is this is incompatible with
    the standard Windows double precision mode.  In fact, if we create a new
    OS thread then Windows will reset the FPU to double precision mode.  So
    we end up with a weird state where the main thread by default has a
    different precision than any child threads.
    
    Test Plan: ./validate new test T7289
    
    Reviewers: simonmar, austin, bgamari, erikd
    
    Reviewed By: simonmar
    
    Subscribers: thomie, #ghc_windows_task_force
    
    Differential Revision: https://phabricator.haskell.org/D2819
    
    GHC Trac Issues: #7289


>---------------------------------------------------------------

6f7d8279cea4aa1082fb07adf5da507297e21ee8
 rts/RtsStartup.c                                   | 34 +++++++++++++++++++---
 .../should_compile => rts/T7289}/Makefile          |  0
 testsuite/tests/rts/T7289/T7289.hs                 |  9 ++++++
 testsuite/tests/rts/T7289/T7289.stdout             |  2 ++
 testsuite/tests/rts/T7289/all.T                    |  6 ++++
 testsuite/tests/rts/T7289/fp.c                     | 26 +++++++++++++++++
 6 files changed, 73 insertions(+), 4 deletions(-)

diff --git a/rts/RtsStartup.c b/rts/RtsStartup.c
index dd4efa6..955ad13 100644
--- a/rts/RtsStartup.c
+++ b/rts/RtsStartup.c
@@ -46,7 +46,9 @@
 #include "win32/AsyncIO.h"
 #endif
 
-#if !defined(mingw32_HOST_OS)
+#if defined(mingw32_HOST_OS)
+#include <fenv.h>
+#else
 #include "posix/TTY.h"
 #endif
 
@@ -69,10 +71,18 @@ static void flushStdHandles(void);
 
 #define X86_INIT_FPU 0
 
-#if X86_INIT_FPU
 static void
 x86_init_fpu ( void )
 {
+#if defined(mingw32_HOST_OS) && !X86_INIT_FPU
+    /* Mingw-w64 does a stupid thing. They set the FPU precision to extended mode by default.
+    The reasoning is that it's for compatibility with GNU Linux ported libraries. However the
+    problem is this is incompatible with the standard Windows double precision mode.  In fact,
+    if we create a new OS thread then Windows will reset the FPU to double precision mode.
+    So we end up with a weird state where the main thread by default has a different precision
+    than any child threads. */
+    fesetenv(FE_PC53_ENV);
+#elif X86_INIT_FPU
   __volatile unsigned short int fpu_cw;
 
   // Grab the control word
@@ -87,8 +97,26 @@ x86_init_fpu ( void )
 
   // Store the new control word back
   __asm __volatile ("fldcw %0" : : "m" (fpu_cw));
+#else
+    return;
+#endif
+}
+
+#if defined(mingw32_HOST_OS)
+/* And now we have to override the build in ones in Mingw-W64's CRT. */
+void _fpreset(void)
+{
+    x86_init_fpu();
+}
+
+#ifdef __GNUC__
+void __attribute__((alias("_fpreset"))) fpreset(void);
+#else
+void fpreset(void) {
+    _fpreset();
 }
 #endif
+#endif
 
 /* -----------------------------------------------------------------------------
    Starting up the RTS
@@ -244,9 +272,7 @@ hs_init_ghc(int *argc, char **argv[], RtsConfig rts_config)
     startupAsyncIO();
 #endif
 
-#if X86_INIT_FPU
     x86_init_fpu();
-#endif
 
     startupHpc();
 
diff --git a/testsuite/tests/annotations/should_compile/Makefile b/testsuite/tests/rts/T7289/Makefile
similarity index 100%
copy from testsuite/tests/annotations/should_compile/Makefile
copy to testsuite/tests/rts/T7289/Makefile
diff --git a/testsuite/tests/rts/T7289/T7289.hs b/testsuite/tests/rts/T7289/T7289.hs
new file mode 100644
index 0000000..1751333
--- /dev/null
+++ b/testsuite/tests/rts/T7289/T7289.hs
@@ -0,0 +1,9 @@
+module Main where
+
+import Control.Concurrent
+
+foreign import ccall "showControlBits" checkfpu :: IO ()
+
+main
+ = do checkfpu
+      forkOS checkfpu
diff --git a/testsuite/tests/rts/T7289/T7289.stdout b/testsuite/tests/rts/T7289/T7289.stdout
new file mode 100644
index 0000000..7a74a81
--- /dev/null
+++ b/testsuite/tests/rts/T7289/T7289.stdout
@@ -0,0 +1,2 @@
+FPU: 0x027f
+FPU: 0x027f
diff --git a/testsuite/tests/rts/T7289/all.T b/testsuite/tests/rts/T7289/all.T
new file mode 100644
index 0000000..7ef59cc
--- /dev/null
+++ b/testsuite/tests/rts/T7289/all.T
@@ -0,0 +1,6 @@
+test('T7289', [ extra_clean(['fp.o', 'testfp.o', 'testfp.hi'])
+               , extra_files(['fp.c'])
+               , unless(opsys('mingw32'), skip)
+               , only_ways(['threaded1'])
+               ],
+               compile_and_run, ['fp.c'])
diff --git a/testsuite/tests/rts/T7289/fp.c b/testsuite/tests/rts/T7289/fp.c
new file mode 100644
index 0000000..12f1b39
--- /dev/null
+++ b/testsuite/tests/rts/T7289/fp.c
@@ -0,0 +1,26 @@
+#include <stdio.h>
+#include <xmmintrin.h>
+#include <float.h>
+
+static unsigned int
+getFPUStateX86 (void)
+{
+    unsigned int control = 0;
+#if defined(_MSC_VER)
+    control = _controlfp(0, 0);
+#else
+    __asm__ __volatile__("fnstcw %0" : "=m" (control));
+#endif
+    return control;
+}
+
+static unsigned int
+getSSEStateX86 (void)
+{
+    return _mm_getcsr();
+}
+
+extern void showControlBits (void)
+{
+    printf("FPU: 0x%04x\n", getFPUStateX86());
+}