Split sources for amdgcn and r600

Most files remain in a common amdgpu directory.

Also switches barriers to to use convergent,
and use llvm.amdgcn.s.barrier.

This now requires 3.9/trunk to build amdgcn.

llvm-svn: 260777

3 files changed, 110 insertions, 0 deletions

diff --git a/libclc/amdgpu/lib/math/ldexp.cl b/libclc/amdgpu/lib/math/ldexp.cl
new file mode 100644
index 00000000000..80439ce3a43
--- /dev/null
+++ b/libclc/amdgpu/lib/math/ldexp.cl
@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include <clc/clc.h>
+
+#include "../../../generic/lib/clcmacro.h"
+
+#ifdef __HAS_LDEXPF__
+#define BUILTINF __builtin_amdgpu_ldexpf
+#else
+#include "math/clc_ldexp.h"
+#define BUILTINF __clc_ldexp
+#endif
+
+// This defines all the ldexp(floatN, intN) variants.
+_CLC_DEFINE_BINARY_BUILTIN(float, ldexp, BUILTINF, float, int);
+
+#ifdef cl_khr_fp64
+  #pragma OPENCL EXTENSION cl_khr_fp64 : enable
+    // This defines all the ldexp(doubleN, intN) variants.
+  _CLC_DEFINE_BINARY_BUILTIN(double, ldexp, __builtin_amdgpu_ldexp, double, int);
+#endif
+
+// This defines all the ldexp(GENTYPE, int);
+#define __CLC_BODY <../../../generic/lib/math/ldexp.inc>
+#include <clc/math/gentype.inc>
+
+#undef BUILTINF
diff --git a/libclc/amdgpu/lib/math/nextafter.cl b/libclc/amdgpu/lib/math/nextafter.cl
new file mode 100644
index 00000000000..4611c81ae91
--- /dev/null
+++ b/libclc/amdgpu/lib/math/nextafter.cl
@@ -0,0 +1,4 @@
+#include <clc/clc.h>
+#include "../lib/clcmacro.h"
+
+_CLC_DEFINE_BINARY_BUILTIN(float, nextafter, __clc_nextafter, float, float)
diff --git a/libclc/amdgpu/lib/math/sqrt.cl b/libclc/amdgpu/lib/math/sqrt.cl
new file mode 100644
index 00000000000..3e5b17c5db0
--- /dev/null
+++ b/libclc/amdgpu/lib/math/sqrt.cl
@@ -0,0 +1,59 @@
+/*
+ * Copyright (c) 2015 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include <clc/clc.h>
+#include "../../../generic/lib/clcmacro.h"
+#include "math/clc_sqrt.h"
+
+_CLC_DEFINE_UNARY_BUILTIN(float, sqrt, __clc_sqrt, float)
+
+#ifdef cl_khr_fp64
+
+#pragma OPENCL EXTENSION cl_khr_fp64 : enable
+
+
+_CLC_OVERLOAD _CLC_DEF double sqrt(double x) {
+
+  uint vcc = x < 0x1p-767;
+  uint exp0 = vcc ? 0x100 : 0;
+  unsigned exp1 = vcc ? 0xffffff80 : 0;
+
+  double v01 = ldexp(x, exp0);
+  double v23 = __builtin_amdgpu_rsq(v01);
+  double v45 = v01 * v23;
+  v23 = v23 * 0.5;
+
+  double v67 = fma(-v23, v45, 0.5);
+  v45 = fma(v45, v67, v45);
+  double v89 = fma(-v45, v45, v01);
+  v23 = fma(v23, v67, v23);
+  v45 = fma(v89, v23, v45);
+  v67 = fma(-v45, v45, v01);
+  v23 = fma(v67, v23, v45);
+
+  v23 = ldexp(v23, exp1);
+  return ((x == __builtin_inf()) || (x == 0.0)) ? v01 : v23;
+}
+
+_CLC_UNARY_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, double, sqrt, double);
+
+#endif