obtained gcc-4.6.4.tar.bz2 from upstream website;upstream

verified gcc-4.6.4.tar.bz2.sig;
imported gcc-4.6.4 source tree from verified upstream tarball.

downloading a git-generated archive based on the 'upstream' tag
should provide you with a source tree that is binary identical
to the one extracted from the above tarball.

if you have obtained the source via the command 'git clone',
however, do note that line-endings of files in your working
directory might differ from line-endings of the respective
files in the upstream repository.

1 files changed, 60 insertions, 0 deletions

diff --git a/gcc/testsuite/gcc.target/alpha/pr19518.c b/gcc/testsuite/gcc.target/alpha/pr19518.c
new file mode 100644
index 000000000..42c58b5a5
--- /dev/null
+++ b/gcc/testsuite/gcc.target/alpha/pr19518.c
@@ -0,0 +1,60 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mcpu=ev67" } */
+
+typedef short INT16;
+typedef unsigned int CARD32;
+typedef unsigned short CARD16;
+typedef unsigned char CARD8;
+typedef struct _Picture *PicturePtr;
+typedef int FbStride;
+typedef unsigned long __m64;
+extern __m64 load8888 (__m64);
+static __inline __m64 _mm_adds_pu8(__m64 __m1, __m64 __m2)
+{
+    return __m1 + __builtin_alpha_minsb8(__m2, ~__m1);
+}
+static __inline __m64 _mm_packs_pu16(__m64 __m1, __m64 __m2)
+{
+    __m1 = __builtin_alpha_minuw4(__m1, 0x00ff00ff00ff00ff);
+    __m2 = __builtin_alpha_minuw4(__m2, 0x00ff00ff00ff00ff);
+    return __m1 | (__m2 << 32);
+}
+typedef unsigned long long ullong;
+static __inline__ __m64 pix_multiply(__m64 a)
+{
+    if (a)
+	return a;
+}
+static __inline__ __m64 over(__m64 src, __m64 srca, __m64 dest)
+{
+    return _mm_adds_pu8(src, pix_multiply(dest));
+}
+
+void fbCompositeSolid_nx8888mmx(CARD8 op, PicturePtr pSrc, PicturePtr pMask,
+				INT16 yDst, CARD16 width, CARD16 height)
+{
+    CARD32 src;
+    CARD32 *dstLine, *dst;
+    CARD16 w;
+    FbStride dstStride;
+    __m64 vsrc, vsrca;
+    vsrc = load8888(src);
+    while (height--) {
+	dst = dstLine;
+	dstLine += dstStride;
+	while (w && (unsigned long) dst & 7) {
+	    *dst = _mm_packs_pu16(_mm_adds_pu8(vsrc, load8888(*dst)),
+				  _mm_setzero_si64());
+	    dst++;
+	}
+	while (w >= 2) {
+	    __m64 dest0, dest1;
+	    *(__m64 *) dst = _mm_packs_pu16(dest0, dest1);
+	    w -= 2;
+	}
+	while (w) {
+	    *dst = _mm_packs_pu16(_mm_adds_pu8(vsrc, pix_multiply(0)), 0);
+	    w--;
+	}
+    }
+}