i386.c (x86_arch_always_fancy_math_387): New.

        * config/i386/i386.c (x86_arch_always_fancy_math_387): New.
        (override_options): Disable NO_FANCY_MATH_387 if the arch allows.
        * config/i386/i386.h (x86_arch_always_fancy_math_387): New.
        * config/i386/i386.md (sqrtxf2, sqrtextendsfxf2, sinxf2): Fix
        conditional.
        * docs/invoke.texi: Update -mno-fancy-math-387 docs.

From-SVN: r51584

7 files changed, 45 insertions, 6 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 3d0da36..101f6a8 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,12 @@
+2002-03-29  Richard Henderson  <rth@redhat.com>
+
+	* config/i386/i386.c (x86_arch_always_fancy_math_387): New.
+	(override_options): Disable NO_FANCY_MATH_387 if the arch allows.
+	* config/i386/i386.h (x86_arch_always_fancy_math_387): New.
+	* config/i386/i386.md (sqrtxf2, sqrtextendsfxf2, sinxf2): Fix
+	conditional.
+	* docs/invoke.texi: Update -mno-fancy-math-387 docs.
+
 2002-03-29  Dale Johannesen <dalej@apple.com>
 
 	* loop.c (combine_movables): Do allow combination of pseudos.
diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 95873e2..8e9df35 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -393,6 +393,7 @@ const int x86_accumulate_outgoing_args = m_ATHLON | m_PENT4 | m_PPRO;
 const int x86_prologue_using_move = m_ATHLON | m_PENT4 | m_PPRO;
 const int x86_epilogue_using_move = m_ATHLON | m_PENT4 | m_PPRO;
 const int x86_decompose_lea = m_PENT4;
+const int x86_arch_always_fancy_math_387 = m_PENT|m_PPRO|m_ATHLON|m_PENT4;
 
 /* In case the avreage insn count for single function invocation is
    lower than this constant, emit fast (but longer) prologue and
@@ -1118,6 +1119,11 @@ override_options ()
   if (flag_unsafe_math_optimizations)
     target_flags &= ~MASK_IEEE_FP;
 
+  /* If the architecture always has an FPU, turn off NO_FANCY_MATH_387,
+     since the insns won't need emulation.  */
+  if (x86_arch_always_fancy_math_387 & (1 << ix86_arch))
+    target_flags &= ~MASK_NO_FANCY_MATH_387;
+
   if (TARGET_64BIT)
     {
       if (TARGET_ALIGN_DOUBLE)
diff --git a/gcc/config/i386/i386.h b/gcc/config/i386/i386.h
index 9435913..ad2956d 100644
--- a/gcc/config/i386/i386.h
+++ b/gcc/config/i386/i386.h
@@ -222,6 +222,7 @@ extern const int x86_add_esp_4, x86_add_esp_8, x86_sub_esp_4, x86_sub_esp_8;
 extern const int x86_partial_reg_dependency, x86_memory_mismatch_stall;
 extern const int x86_accumulate_outgoing_args, x86_prologue_using_move;
 extern const int x86_epilogue_using_move, x86_decompose_lea;
+extern const int x86_arch_always_fancy_math_387;
 extern int x86_prefetch_sse;
 
 #define TARGET_USE_LEAVE (x86_use_leave & CPUMASK)
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index 27a352d..8b3ee78 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -14701,7 +14701,7 @@
   [(set (match_operand:XF 0 "register_operand" "=f")
 	(sqrt:XF (float_extend:XF
 		  (match_operand:DF 1 "register_operand" "0"))))]
-  "!TARGET_64BIT && TARGET_80387 && TARGET_NO_FANCY_MATH_387"
+  "!TARGET_64BIT && TARGET_80387 && !TARGET_NO_FANCY_MATH_387"
   "fsqrt"
   [(set_attr "type" "fpspc")
    (set_attr "mode" "XF")
@@ -14721,7 +14721,7 @@
   [(set (match_operand:XF 0 "register_operand" "=f")
 	(sqrt:XF (float_extend:XF
 		  (match_operand:SF 1 "register_operand" "0"))))]
-  "!TARGET_64BIT && TARGET_80387 && TARGET_NO_FANCY_MATH_387"
+  "!TARGET_64BIT && TARGET_80387 && !TARGET_NO_FANCY_MATH_387"
   "fsqrt"
   [(set_attr "type" "fpspc")
    (set_attr "mode" "XF")
@@ -14768,7 +14768,7 @@
 (define_insn "sinxf2"
   [(set (match_operand:XF 0 "register_operand" "=f")
 	(unspec:XF [(match_operand:XF 1 "register_operand" "0")] 1))]
-  "!TARGET_64BIT && TARGET_80387 && TARGET_NO_FANCY_MATH_387
+  "!TARGET_64BIT && TARGET_80387 && !TARGET_NO_FANCY_MATH_387
    && flag_unsafe_math_optimizations"
   "fsin"
   [(set_attr "type" "fpspc")
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index b1e6d5d..c26d183 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -7366,9 +7366,12 @@ in ordinary CPU registers instead.
 @opindex mno-fancy-math-387
 Some 387 emulators do not support the @code{sin}, @code{cos} and
 @code{sqrt} instructions for the 387.  Specify this option to avoid
-generating those instructions.  This option is the default on FreeBSD@.
-As of revision 2.6.1, these instructions are not generated unless you
-also use the @option{-funsafe-math-optimizations} switch.
+generating those instructions.  This option is the default on FreeBSD,
+OpenBSD and NetBSD@.  This option is overridden when @option{-march}
+indicates that the target cpu will always have an FPU and so the
+instruction will not need emulation.  As of revision 2.6.1, these
+instructions are not generated unless you also use the
+@option{-funsafe-math-optimizations} switch.
 
 @item -malign-double
 @itemx -mno-align-double
diff --git a/gcc/testsuite/gcc.dg/i386-387-1.c b/gcc/testsuite/gcc.dg/i386-387-1.c
new file mode 100644
index 0000000..2c9334b
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/i386-387-1.c
@@ -0,0 +1,10 @@
+/* Verify that -mno-fancy-math-387 works.  */
+/* { dg-do compile { target "i?86-*-*" } } */
+/* { dg-options "-O -ffast-math -mfpmath=387 -mno-fancy-math-387" } */
+/* { dg-final { scan-assembler "call\tsin" } } */
+/* { dg-final { scan-assembler "call\tcos" } } */
+/* { dg-final { scan-assembler "call\tsqrt" } } */
+
+double f1(double x) { return __builtin_sin(x); }
+double f2(double x) { return __builtin_cos(x); }
+double f3(double x) { return __builtin_sqrt(x); }
diff --git a/gcc/testsuite/gcc.dg/i386-387-2.c b/gcc/testsuite/gcc.dg/i386-387-2.c
new file mode 100644
index 0000000..2456ca0
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/i386-387-2.c
@@ -0,0 +1,10 @@
+/* Verify that -march overrides -mno-fancy-math-387.  */
+/* { dg-do compile { target "i?86-*-*" } } */
+/* { dg-options "-O -ffast-math -mfpmath=387 -march=i686 -mno-fancy-math-387" } */
+/* { dg-final { scan-assembler "fsin" } } */
+/* { dg-final { scan-assembler "fcos" } } */
+/* { dg-final { scan-assembler "fsqrt" } } */
+
+double f1(double x) { return __builtin_sin(x); }
+double f2(double x) { return __builtin_cos(x); }
+double f3(double x) { return __builtin_sqrt(x); }