[clang] 544d89e - [OpenCL] Add atomic_half type builtins
Sven van Haastregt via cfe-commits
cfe-commits at lists.llvm.org
Tue Oct 12 02:45:41 PDT 2021
Author: Sven van Haastregt
Date: 2021-10-12T10:45:30+01:00
New Revision: 544d89e847d42ce8856296752b0fb279aa89aace
URL: https://github.com/llvm/llvm-project/commit/544d89e847d42ce8856296752b0fb279aa89aace
DIFF: https://github.com/llvm/llvm-project/commit/544d89e847d42ce8856296752b0fb279aa89aace.diff
LOG: [OpenCL] Add atomic_half type builtins
Add atomic_half types and builtins operating on the types from the
cl_ext_float_atomics extension.
Patch by Haonan Yang.
Differential Revision: https://reviews.llvm.org/D109740
Added:
Modified:
clang/lib/Headers/opencl-c.h
clang/lib/Sema/OpenCLBuiltins.td
clang/lib/Sema/Sema.cpp
clang/test/SemaOpenCL/atomic-ops.cl
Removed:
################################################################################
diff --git a/clang/lib/Headers/opencl-c.h b/clang/lib/Headers/opencl-c.h
index 705a3231b577..562e0551ffdc 100644
--- a/clang/lib/Headers/opencl-c.h
+++ b/clang/lib/Headers/opencl-c.h
@@ -13640,6 +13640,118 @@ uintptr_t __ovld atomic_fetch_sub_explicit(volatile __local atomic_uintptr_t *ob
// The functionality added by cl_ext_float_atomics extension
#if defined(cl_ext_float_atomics)
+#if defined(__opencl_c_ext_fp16_global_atomic_load_store)
+void __ovld atomic_store(volatile __global atomic_half *object, half operand);
+void __ovld atomic_store_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order);
+void __ovld atomic_store_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+half __ovld atomic_load(volatile __global atomic_half *object);
+half __ovld atomic_load_explicit(volatile __global atomic_half *object,
+ memory_order order);
+half __ovld atomic_load_explicit(volatile __global atomic_half *object,
+ memory_order order, memory_scope scope);
+half __ovld atomic_exchange(volatile __global atomic_half *object,
+ half operand);
+half __ovld atomic_exchange_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_exchange_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+#endif // defined(__opencl_c_ext_fp16_global_atomic_load_store)
+
+#if defined(__opencl_c_ext_fp16_local_atomic_load_store)
+void __ovld atomic_store(volatile __local atomic_half *object, half operand);
+void __ovld atomic_store_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order);
+void __ovld atomic_store_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+half __ovld atomic_load(volatile __local atomic_half *object);
+half __ovld atomic_load_explicit(volatile __local atomic_half *object,
+ memory_order order);
+half __ovld atomic_load_explicit(volatile __local atomic_half *object,
+ memory_order order, memory_scope scope);
+half __ovld atomic_exchange(volatile __local atomic_half *object, half operand);
+half __ovld atomic_exchange_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_exchange_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+#endif // defined(__opencl_c_ext_fp16_local_atomic_load_store)
+
+#if defined(__opencl_c_ext_fp16_global_atomic_load_store) && \
+ defined(__opencl_c_ext_fp16_local_atomic_load_store)
+void __ovld atomic_store(volatile atomic_half *object, half operand);
+void __ovld atomic_store_explicit(volatile atomic_half *object, half operand,
+ memory_order order);
+void __ovld atomic_store_explicit(volatile atomic_half *object, half operand,
+ memory_order order, memory_scope scope);
+half __ovld atomic_load(volatile atomic_half *object);
+half __ovld atomic_load_explicit(volatile atomic_half *object,
+ memory_order order);
+half __ovld atomic_load_explicit(volatile atomic_half *object,
+ memory_order order, memory_scope scope);
+half __ovld atomic_exchange(volatile atomic_half *object, half operand);
+half __ovld atomic_exchange_explicit(volatile atomic_half *object, half operand,
+ memory_order order);
+half __ovld atomic_exchange_explicit(volatile atomic_half *object, half operand,
+ memory_order order, memory_scope scope);
+#endif // defined(__opencl_c_ext_fp16_global_atomic_load_store) &&
+ // defined(__opencl_c_ext_fp16_local_atomic_load_store)
+
+#if defined(__opencl_c_ext_fp16_global_atomic_min_max)
+half __ovld atomic_fetch_min(volatile __global atomic_half *object,
+ half operand);
+half __ovld atomic_fetch_max(volatile __global atomic_half *object,
+ half operand);
+half __ovld atomic_fetch_min_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_max_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_min_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+half __ovld atomic_fetch_max_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+#endif // defined(__opencl_c_ext_fp16_global_atomic_min_max)
+
+#if defined(__opencl_c_ext_fp16_local_atomic_min_max)
+half __ovld atomic_fetch_min(volatile __local atomic_half *object,
+ half operand);
+half __ovld atomic_fetch_max(volatile __local atomic_half *object,
+ half operand);
+half __ovld atomic_fetch_min_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_max_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_min_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+half __ovld atomic_fetch_max_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+#endif // defined(__opencl_c_ext_fp16_local_atomic_min_max)
+
+#if defined(__opencl_c_ext_fp16_global_atomic_min_max) && \
+ defined(__opencl_c_ext_fp16_local_atomic_min_max)
+half __ovld atomic_fetch_min(volatile atomic_half *object, half operand);
+half __ovld atomic_fetch_max(volatile atomic_half *object, half operand);
+half __ovld atomic_fetch_min_explicit(volatile atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_max_explicit(volatile atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_min_explicit(volatile atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+half __ovld atomic_fetch_max_explicit(volatile atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+#endif // defined(__opencl_c_ext_fp16_global_atomic_min_max) && \
+ defined(__opencl_c_ext_fp16_local_atomic_min_max)
+
#if defined(__opencl_c_ext_fp32_global_atomic_min_max)
float __ovld atomic_fetch_min(volatile __global atomic_float *object,
float operand);
@@ -13742,6 +13854,57 @@ double __ovld atomic_fetch_max_explicit(volatile atomic_double *object,
#endif // defined(__opencl_c_ext_fp64_global_atomic_min_max) && \
defined(__opencl_c_ext_fp64_local_atomic_min_max)
+#if defined(__opencl_c_ext_fp16_global_atomic_add)
+half __ovld atomic_fetch_add(volatile __global atomic_half *object,
+ half operand);
+half __ovld atomic_fetch_sub(volatile __global atomic_half *object,
+ half operand);
+half __ovld atomic_fetch_add_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_sub_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_add_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+half __ovld atomic_fetch_sub_explicit(volatile __global atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+#endif // defined(__opencl_c_ext_fp16_global_atomic_add)
+
+#if defined(__opencl_c_ext_fp16_local_atomic_add)
+half __ovld atomic_fetch_add(volatile __local atomic_half *object,
+ half operand);
+half __ovld atomic_fetch_sub(volatile __local atomic_half *object,
+ half operand);
+half __ovld atomic_fetch_add_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_sub_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_add_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+half __ovld atomic_fetch_sub_explicit(volatile __local atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+#endif // defined(__opencl_c_ext_fp16_local_atomic_add)
+
+#if defined(__opencl_c_ext_fp16_global_atomic_add) && \
+ defined(__opencl_c_ext_fp16_local_atomic_add)
+half __ovld atomic_fetch_add(volatile atomic_half *object, half operand);
+half __ovld atomic_fetch_sub(volatile atomic_half *object, half operand);
+half __ovld atomic_fetch_add_explicit(volatile atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_sub_explicit(volatile atomic_half *object,
+ half operand, memory_order order);
+half __ovld atomic_fetch_add_explicit(volatile atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+half __ovld atomic_fetch_sub_explicit(volatile atomic_half *object,
+ half operand, memory_order order,
+ memory_scope scope);
+#endif // defined(__opencl_c_ext_fp16_global_atomic_add) && \
+ defined(__opencl_c_ext_fp16_local_atomic_add)
+
#if defined(__opencl_c_ext_fp32_global_atomic_add)
float __ovld atomic_fetch_add(volatile __global atomic_float *object,
float operand);
diff --git a/clang/lib/Sema/OpenCLBuiltins.td b/clang/lib/Sema/OpenCLBuiltins.td
index 2501e50b8968..8cf7ec58eff5 100644
--- a/clang/lib/Sema/OpenCLBuiltins.td
+++ b/clang/lib/Sema/OpenCLBuiltins.td
@@ -85,16 +85,25 @@ def FuncExtKhrGlMsaaSharing : FunctionExtension<"cl_khr_gl_msaa_sha
def FuncExtOpenCLCPipes : FunctionExtension<"__opencl_c_pipes">;
def FuncExtOpenCLCWGCollectiveFunctions : FunctionExtension<"__opencl_c_work_group_collective_functions">;
+def FuncExtFloatAtomicsFp16GlobalLoadStore : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp16_global_atomic_load_store">;
+def FuncExtFloatAtomicsFp16LocalLoadStore : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp16_local_atomic_load_store">;
+def FuncExtFloatAtomicsFp16GenericLoadStore : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp16_global_atomic_load_store __opencl_c_ext_fp16_local_atomic_load_store">;
+def FuncExtFloatAtomicsFp16GlobalAdd : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp16_global_atomic_add">;
def FuncExtFloatAtomicsFp32GlobalAdd : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp32_global_atomic_add">;
def FuncExtFloatAtomicsFp64GlobalAdd : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp64_global_atomic_add">;
+def FuncExtFloatAtomicsFp16LocalAdd : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp16_local_atomic_add">;
def FuncExtFloatAtomicsFp32LocalAdd : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp32_local_atomic_add">;
def FuncExtFloatAtomicsFp64LocalAdd : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp64_local_atomic_add">;
+def FuncExtFloatAtomicsFp16GenericAdd : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp16_local_atomic_add __opencl_c_ext_fp16_global_atomic_add">;
def FuncExtFloatAtomicsFp32GenericAdd : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp32_local_atomic_add __opencl_c_ext_fp32_global_atomic_add">;
def FuncExtFloatAtomicsFp64GenericAdd : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp64_local_atomic_add __opencl_c_ext_fp64_global_atomic_add">;
+def FuncExtFloatAtomicsFp16GlobalMinMax : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp16_global_atomic_min_max">;
def FuncExtFloatAtomicsFp32GlobalMinMax : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp32_global_atomic_min_max">;
def FuncExtFloatAtomicsFp64GlobalMinMax : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp64_global_atomic_min_max">;
+def FuncExtFloatAtomicsFp16LocalMinMax : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp16_local_atomic_min_max">;
def FuncExtFloatAtomicsFp32LocalMinMax : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp32_local_atomic_min_max">;
def FuncExtFloatAtomicsFp64LocalMinMax : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp64_local_atomic_min_max">;
+def FuncExtFloatAtomicsFp16GenericMinMax : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp16_local_atomic_min_max __opencl_c_ext_fp16_global_atomic_min_max">;
def FuncExtFloatAtomicsFp32GenericMinMax : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp32_local_atomic_min_max __opencl_c_ext_fp32_global_atomic_min_max">;
def FuncExtFloatAtomicsFp64GenericMinMax : FunctionExtension<"cl_ext_float_atomics __opencl_c_ext_fp64_local_atomic_min_max __opencl_c_ext_fp64_global_atomic_min_max">;
@@ -362,6 +371,7 @@ def AtomicLong : Type<"atomic_long", QualType<"Context.getAtomicType(
def AtomicULong : Type<"atomic_ulong", QualType<"Context.getAtomicType(Context.UnsignedLongTy)">>;
def AtomicFloat : Type<"atomic_float", QualType<"Context.getAtomicType(Context.FloatTy)">>;
def AtomicDouble : Type<"atomic_double", QualType<"Context.getAtomicType(Context.DoubleTy)">>;
+def AtomicHalf : Type<"atomic_half", QualType<"Context.getAtomicType(Context.HalfTy)">>;
def AtomicIntPtr : Type<"atomic_intptr_t", QualType<"Context.getAtomicType(Context.getIntPtrType())">>;
def AtomicUIntPtr : Type<"atomic_uintptr_t", QualType<"Context.getAtomicType(Context.getUIntPtrType())">>;
def AtomicSize : Type<"atomic_size_t", QualType<"Context.getAtomicType(Context.getSizeType())">>;
@@ -1106,7 +1116,75 @@ let MinVersion = CL20 in {
// The functionality added by cl_ext_float_atomics extension
let MinVersion = CL20 in {
+ let Extension = FuncExtFloatAtomicsFp16GlobalLoadStore in {
+ def : Builtin<"atomic_store",
+ [Void, PointerType<VolatileType<AtomicHalf>, GlobalAS>, AtomicHalf]>;
+ def : Builtin<"atomic_store_explicit",
+ [Void, PointerType<VolatileType<AtomicHalf>, GlobalAS>, AtomicHalf, MemoryOrder]>;
+ def : Builtin<"atomic_store_explicit",
+ [Void, PointerType<VolatileType<AtomicHalf>, GlobalAS>, AtomicHalf, MemoryOrder, MemoryScope]>;
+ def : Builtin<"atomic_load",
+ [Half, PointerType<VolatileType<AtomicHalf>, GlobalAS>]>;
+ def : Builtin<"atomic_load_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GlobalAS>, MemoryOrder]>;
+ def : Builtin<"atomic_load_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GlobalAS>, MemoryOrder, MemoryScope]>;
+ def : Builtin<"atomic_exchange",
+ [Half, PointerType<VolatileType<AtomicHalf>, GlobalAS>, Half]>;
+ def : Builtin<"atomic_exchange_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GlobalAS>, Half, MemoryOrder]>;
+ def : Builtin<"atomic_exchange_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GlobalAS>, Half, MemoryOrder, MemoryScope]>;
+ }
+ let Extension = FuncExtFloatAtomicsFp16LocalLoadStore in {
+ def : Builtin<"atomic_store",
+ [Void, PointerType<VolatileType<AtomicHalf>, LocalAS>, AtomicHalf]>;
+ def : Builtin<"atomic_store_explicit",
+ [Void, PointerType<VolatileType<AtomicHalf>, LocalAS>, AtomicHalf, MemoryOrder]>;
+ def : Builtin<"atomic_store_explicit",
+ [Void, PointerType<VolatileType<AtomicHalf>, LocalAS>, AtomicHalf, MemoryOrder, MemoryScope]>;
+ def : Builtin<"atomic_load",
+ [Half, PointerType<VolatileType<AtomicHalf>, LocalAS>]>;
+ def : Builtin<"atomic_load_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, LocalAS>, MemoryOrder]>;
+ def : Builtin<"atomic_load_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, LocalAS>, MemoryOrder, MemoryScope]>;
+ def : Builtin<"atomic_exchange",
+ [Half, PointerType<VolatileType<AtomicHalf>, LocalAS>, Half]>;
+ def : Builtin<"atomic_exchange_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, LocalAS>, Half, MemoryOrder]>;
+ def : Builtin<"atomic_exchange_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, LocalAS>, Half, MemoryOrder, MemoryScope]>;
+ }
+ let Extension = FuncExtFloatAtomicsFp16GenericLoadStore in {
+ def : Builtin<"atomic_store",
+ [Void, PointerType<VolatileType<AtomicHalf>, GenericAS>, AtomicHalf]>;
+ def : Builtin<"atomic_store_explicit",
+ [Void, PointerType<VolatileType<AtomicHalf>, GenericAS>, AtomicHalf, MemoryOrder]>;
+ def : Builtin<"atomic_store_explicit",
+ [Void, PointerType<VolatileType<AtomicHalf>, GenericAS>, AtomicHalf, MemoryOrder, MemoryScope]>;
+ def : Builtin<"atomic_load",
+ [Half, PointerType<VolatileType<AtomicHalf>, GenericAS>]>;
+ def : Builtin<"atomic_load_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GenericAS>, MemoryOrder]>;
+ def : Builtin<"atomic_load_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GenericAS>, MemoryOrder, MemoryScope]>;
+ def : Builtin<"atomic_exchange",
+ [Half, PointerType<VolatileType<AtomicHalf>, GenericAS>, Half]>;
+ def : Builtin<"atomic_exchange_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GenericAS>, Half, MemoryOrder]>;
+ def : Builtin<"atomic_exchange_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GenericAS>, Half, MemoryOrder, MemoryScope]>;
+ }
foreach ModOp = ["add", "sub"] in {
+ let Extension = FuncExtFloatAtomicsFp16GlobalAdd in {
+ def : Builtin<"atomic_fetch_" # ModOp,
+ [Half, PointerType<VolatileType<AtomicFloat>, GlobalAS>, Half]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicFloat>, GlobalAS>, Half, MemoryOrder]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicFloat>, GlobalAS>, Half, MemoryOrder, MemoryScope]>;
+ }
let Extension = FuncExtFloatAtomicsFp32GlobalAdd in {
def : Builtin<"atomic_fetch_" # ModOp,
[Float, PointerType<VolatileType<AtomicFloat>, GlobalAS>, Float]>;
@@ -1123,6 +1201,14 @@ let MinVersion = CL20 in {
def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
[Double, PointerType<VolatileType<AtomicDouble>, GlobalAS>, Double, MemoryOrder, MemoryScope]>;
}
+ let Extension = FuncExtFloatAtomicsFp16LocalAdd in {
+ def : Builtin<"atomic_fetch_" # ModOp,
+ [Half, PointerType<VolatileType<AtomicFloat>, LocalAS>, Half]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicFloat>, LocalAS>, Half, MemoryOrder]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicFloat>, LocalAS>, Half, MemoryOrder, MemoryScope]>;
+ }
let Extension = FuncExtFloatAtomicsFp32LocalAdd in {
def : Builtin<"atomic_fetch_" # ModOp,
[Float, PointerType<VolatileType<AtomicFloat>, LocalAS>, Float]>;
@@ -1139,6 +1225,14 @@ let MinVersion = CL20 in {
def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
[Double, PointerType<VolatileType<AtomicDouble>, LocalAS>, Double, MemoryOrder, MemoryScope]>;
}
+ let Extension = FuncExtFloatAtomicsFp16GenericAdd in {
+ def : Builtin<"atomic_fetch_" # ModOp,
+ [Half, PointerType<VolatileType<AtomicFloat>, GenericAS>, Half]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicFloat>, GenericAS>, Half, MemoryOrder]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicFloat>, GenericAS>, Half, MemoryOrder, MemoryScope]>;
+ }
let Extension = FuncExtFloatAtomicsFp32GenericAdd in {
def : Builtin<"atomic_fetch_" # ModOp,
[Float, PointerType<VolatileType<AtomicFloat>, GenericAS>, Float]>;
@@ -1157,6 +1251,14 @@ let MinVersion = CL20 in {
}
}
foreach ModOp = ["min", "max"] in {
+ let Extension = FuncExtFloatAtomicsFp16GlobalMinMax in {
+ def : Builtin<"atomic_fetch_" # ModOp,
+ [Half, PointerType<VolatileType<AtomicHalf>, GlobalAS>, Half]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GlobalAS>, Half, MemoryOrder]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GlobalAS>, Half, MemoryOrder, MemoryScope]>;
+ }
let Extension = FuncExtFloatAtomicsFp32GlobalMinMax in {
def : Builtin<"atomic_fetch_" # ModOp,
[Float, PointerType<VolatileType<AtomicFloat>, GlobalAS>, Float]>;
@@ -1173,6 +1275,14 @@ let MinVersion = CL20 in {
def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
[Double, PointerType<VolatileType<AtomicDouble>, GlobalAS>, Double, MemoryOrder, MemoryScope]>;
}
+ let Extension = FuncExtFloatAtomicsFp16LocalMinMax in {
+ def : Builtin<"atomic_fetch_" # ModOp,
+ [Half, PointerType<VolatileType<AtomicHalf>, LocalAS>, Half]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, LocalAS>, Half, MemoryOrder]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, LocalAS>, Half, MemoryOrder, MemoryScope]>;
+ }
let Extension = FuncExtFloatAtomicsFp32LocalMinMax in {
def : Builtin<"atomic_fetch_" # ModOp,
[Float, PointerType<VolatileType<AtomicFloat>, LocalAS>, Float]>;
@@ -1189,6 +1299,14 @@ let MinVersion = CL20 in {
def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
[Double, PointerType<VolatileType<AtomicDouble>, LocalAS>, Double, MemoryOrder, MemoryScope]>;
}
+ let Extension = FuncExtFloatAtomicsFp16GenericMinMax in {
+ def : Builtin<"atomic_fetch_" # ModOp,
+ [Half, PointerType<VolatileType<AtomicHalf>, GenericAS>, Half]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GenericAS>, Half, MemoryOrder]>;
+ def : Builtin<"atomic_fetch_" # ModOp # "_explicit",
+ [Half, PointerType<VolatileType<AtomicHalf>, GenericAS>, Half, MemoryOrder, MemoryScope]>;
+ }
let Extension = FuncExtFloatAtomicsFp32GenericMinMax in {
def : Builtin<"atomic_fetch_" # ModOp,
[Float, PointerType<VolatileType<AtomicFloat>, GenericAS>, Float]>;
diff --git a/clang/lib/Sema/Sema.cpp b/clang/lib/Sema/Sema.cpp
index 6430671a8531..88833f233adc 100644
--- a/clang/lib/Sema/Sema.cpp
+++ b/clang/lib/Sema/Sema.cpp
@@ -367,6 +367,11 @@ void Sema::Initialize() {
AddPointerSizeDependentTypes();
}
+ if (getOpenCLOptions().isSupported("cl_khr_fp16", getLangOpts())) {
+ auto AtomicHalfT = Context.getAtomicType(Context.HalfTy);
+ addImplicitTypedef("atomic_half", AtomicHalfT);
+ }
+
std::vector<QualType> Atomic64BitTypes;
if (getOpenCLOptions().isSupported("cl_khr_int64_base_atomics",
getLangOpts()) &&
diff --git a/clang/test/SemaOpenCL/atomic-ops.cl b/clang/test/SemaOpenCL/atomic-ops.cl
index 728c07540a4b..8a56fa73297d 100644
--- a/clang/test/SemaOpenCL/atomic-ops.cl
+++ b/clang/test/SemaOpenCL/atomic-ops.cl
@@ -19,7 +19,7 @@ int8 i64;
atomic_int gn;
void f(atomic_int *i, const atomic_int *ci,
- atomic_intptr_t *p, atomic_float *f, atomic_double *d, atomic_half *h, // expected-error {{unknown type name 'atomic_half'}}
+ atomic_intptr_t *p, atomic_float *f, atomic_double *d, atomic_half *h,
int *I, const int *CI,
intptr_t *P, float *D, struct S *s1, struct S *s2,
global atomic_int *i_g, local atomic_int *i_l, private atomic_int *i_p,
More information about the cfe-commits
mailing list