[OpenCL] Add work-group and miscellaneous vector builtin functions
Add the work-group and miscellaneous vector builtin functions from the
OpenCL C specification.
Patch by Pierre Gondois and Sven van Haastregt.
diff --git a/clang/lib/Sema/OpenCLBuiltins.td b/clang/lib/Sema/OpenCLBuiltins.td
index 0bd4c51..353e0c1 100644
--- a/clang/lib/Sema/OpenCLBuiltins.td
+++ b/clang/lib/Sema/OpenCLBuiltins.td
@@ -274,14 +274,21 @@
def VecAndScalar: IntList<"VecAndScalar", [1, 2, 3, 4, 8, 16]>;
def VecNoScalar : IntList<"VecNoScalar", [2, 3, 4, 8, 16]>;
def Vec1 : IntList<"Vec1", [1]>;
+def Vec2 : IntList<"Vec2", [2]>;
+def Vec4 : IntList<"Vec4", [4]>;
+def Vec8 : IntList<"Vec8", [8]>;
+def Vec16 : IntList<"Vec16", [16]>;
def Vec1234 : IntList<"Vec1234", [1, 2, 3, 4]>;
// Type lists.
-def TLAll : TypeList<"TLAll", [Char, UChar, Short, UShort, Int, UInt, Long, ULong, Float, Double, Half]>;
+def TLAll : TypeList<"TLAll", [Char, UChar, Short, UShort, Int, UInt, Long, ULong, Float, Double, Half]>;
+def TLAllUnsigned : TypeList<"TLAllUnsigned", [UChar, UChar, UShort, UShort, UInt, UInt, ULong, ULong, UInt, ULong, UShort]>;
def TLFloat : TypeList<"TLFloat", [Float, Double, Half]>;
def TLSignedInts : TypeList<"TLSignedInts", [Char, Short, Int, Long]>;
def TLUnsignedInts : TypeList<"TLUnsignedInts", [UChar, UShort, UInt, ULong]>;
+def TLIntLongFloats : TypeList<"TLIntLongFloats", [Int, UInt, Long, ULong, Float, Double, Half]>;
+
// All unsigned integer types twice, to facilitate unsigned return types for e.g.
// uchar abs(char) and
// uchar abs(uchar).
@@ -306,6 +313,8 @@
def UGenTypeN : GenericType<"UGenTypeN", TLUnsignedInts, VecAndScalar>;
// Float
def FGenTypeN : GenericType<"FGenTypeN", TLFloat, VecAndScalar>;
+// (u)int, (u)long, and all floats
+def IntLongFloatGenType1 : GenericType<"IntLongFloatGenType1", TLIntLongFloats, Vec1>;
// GenType definitions for every single base type (e.g. fp32 only).
// Names are like: GenTypeFloatVecAndScalar.
@@ -868,6 +877,31 @@
}
//--------------------------------------------------------------------
+// OpenCL v1.1 s6.11.12, v1.2 s6.12.12, v2.0 s6.13.12 - Miscellaneous Vector Functions
+// --- Table 19 ---
+foreach name = ["shuffle"] in {
+ foreach VSize1 = [Vec2, Vec4, Vec8, Vec16] in {
+ foreach VSize2 = [Vec2, Vec4, Vec8, Vec16] in {
+ def : Builtin<name, [GenericType<"TLAll" # VSize1.Name, TLAll, VSize1>,
+ GenericType<"TLAll" # VSize2.Name, TLAll, VSize2>,
+ GenericType<"TLAllUnsigned" # VSize1.Name, TLAllUnsigned, VSize1>],
+ Attr.Const>;
+ }
+ }
+}
+foreach name = ["shuffle2"] in {
+ foreach VSize1 = [Vec2, Vec4, Vec8, Vec16] in {
+ foreach VSize2 = [Vec2, Vec4, Vec8, Vec16] in {
+ def : Builtin<name, [GenericType<"TLAll" # VSize1.Name, TLAll, VSize1>,
+ GenericType<"TLAll" # VSize2.Name, TLAll, VSize2>,
+ GenericType<"TLAll" # VSize2.Name, TLAll, VSize2>,
+ GenericType<"TLAllUnsigned" # VSize1.Name, TLAllUnsigned, VSize1>],
+ Attr.Const>;
+ }
+ }
+}
+
+//--------------------------------------------------------------------
// OpenCL v1.1 s6.11.3, v1.2 s6.12.14, v2.0 s6.13.14: Image Read and Write Functions
// OpenCL Extension v2.0 s5.1.8 and s6.1.8: Image Read and Write Functions
// --- Table 22: Image Read Functions with Samplers ---
@@ -1020,6 +1054,27 @@
}
+//--------------------------------------------------------------------
+// OpenCL v2.0 s6.13.15 - Work-group Functions
+// --- Table 26 ---
+let MinVersion = CL20 in {
+ foreach name = ["work_group_all", "work_group_any"] in {
+ def : Builtin<name, [Int, Int], Attr.Convergent>;
+ }
+ foreach name = ["work_group_broadcast"] in {
+ def : Builtin<name, [IntLongFloatGenType1, IntLongFloatGenType1, Size], Attr.Convergent>;
+ def : Builtin<name, [IntLongFloatGenType1, IntLongFloatGenType1, Size, Size], Attr.Convergent>;
+ def : Builtin<name, [IntLongFloatGenType1, IntLongFloatGenType1, Size, Size, Size], Attr.Convergent>;
+ }
+ foreach op = ["add", "min", "max"] in {
+ foreach name = ["work_group_reduce_", "work_group_scan_exclusive_",
+ "work_group_scan_inclusive_"] in {
+ def : Builtin<name # op, [IntLongFloatGenType1, IntLongFloatGenType1], Attr.Convergent>;
+ }
+ }
+}
+
+
// OpenCL v2.0 s9.17.3: Additions to section 6.13.1: Work-Item Functions
let MinVersion = CL20 in {
let Extension = "cl_khr_subgroups" in {