From: Chris Metcalf <cmetcalf@tilera.com>
Date: Thu, 17 May 2012 15:27:36 +0000 (-0400)
Subject: tile: use ieee754/dbl-64 version of fma() and fmaf()
X-Git-Tag: glibc-2.16-ports-before-merge~107
X-Git-Url: https://sourceware.org/git/?a=commitdiff_plain;h=ac9182e7532b7a903afa1b86545dd9f1cc308a74;p=glibc.git

tile: use ieee754/dbl-64 version of fma() and fmaf()

It turns out that even if you stub out the rounding and exception
support and use the ieee754 version, it's still much better than
the generic version that just uses normal multiply and add.
The resulting functions have only 1 ULP of error according to the tests.
---

diff --git a/ChangeLog.tile b/ChangeLog.tile
index b800112196..7d3c7bde7e 100644
--- a/ChangeLog.tile
+++ b/ChangeLog.tile
@@ -1,3 +1,8 @@
+2012-05-17  Chris Metcalf  <cmetcalf@tilera.com>
+
+	* sysdeps/tile/s_fma.c: Use ieee754/dbl-64 without rounding or
+	exception support; it's still much better than the generic version.
+
 2012-05-16  Chris Metcalf  <cmetcalf@tilera.com>
 
 	* sysdeps/unix/sysv/linux/tile/ucontext_i: Fix tilegx32 offset bug.
diff --git a/sysdeps/tile/s_fma.c b/sysdeps/tile/s_fma.c
index 355b4edf6a..5be586562e 100644
--- a/sysdeps/tile/s_fma.c
+++ b/sysdeps/tile/s_fma.c
@@ -1,3 +1,8 @@
-/* Although tile uses ieee754/dbl-64, it does not support the
-   rounding modes required to use the standard dbl-64 s_fma.c.  */
-#include <math/s_fma.c>
+/* tile does not support the rounding modes required by the ieee754/dbl-64
+   implementation, but it's still better than the generic implementation.  */
+
+#define libc_feholdexcept_setround(e, x) do { (void) (e); } while (0)
+#define libc_feupdateenv_test(e, x) ((void) (e), 0)
+#define libc_fetestexcept(x) 0
+
+#include <sysdeps/ieee754/dbl-64/s_fma.c>
diff --git a/sysdeps/tile/s_fmaf.c b/sysdeps/tile/s_fmaf.c
index caa8e40654..92d66b2578 100644
--- a/sysdeps/tile/s_fmaf.c
+++ b/sysdeps/tile/s_fmaf.c
@@ -1,3 +1,8 @@
-/* Although tile uses ieee754/dbl-64, it does not support the
-   rounding modes required to use the standard dbl-64 s_fmaf.c.  */
-#include <math/s_fmaf.c>
+/* tile does not support the rounding modes required by the ieee754/dbl-64
+   implementation, but it's still better than the generic implementation.  */
+
+#define libc_feholdexcept_setround(e, x) do { (void) (e); } while (0)
+#define libc_feupdateenv_test(e, x) ((void) (e), 0)
+#define libc_fetestexcept(x) 0
+
+#include <sysdeps/ieee754/dbl-64/s_fmaf.c>