HIP: use build-in atomicAdd(double)

psychocoderHPC · sbastrakov · commit 0b96515b1c8a · 2022-07-27T10:02:16.000+02:00
HIP is providing for HIP 4.2+ native atomicAdd interfaces for double precision.
Use always the build-in atomicAdd so speedup atomic add.
diff --git a/include/alpaka/atomic/AtomicUniformCudaHipBuiltIn.hpp b/include/alpaka/atomic/AtomicUniformCudaHipBuiltIn.hpp
@@ -124,7 +124,7 @@ namespace alpaka
                 double* const addr,
                 double const& value) -> double
             {
-#        if BOOST_ARCH_PTX >= BOOST_VERSION_NUMBER(6, 0, 0)
+#        if BOOST_ARCH_PTX >= BOOST_VERSION_NUMBER(6, 0, 0) || BOOST_LANG_HIP
                 return ::atomicAdd(addr, value);
 #        else
                 // Code from: http://docs.nvidia.com/cuda/cuda-c-programming-guide/#atomic-functions

Original file line number	Diff line number	Diff line change
`@@ -124,7 +124,7 @@ namespace alpaka`
`124`	`124`	`double* const addr,`
`125`	`125`	`double const& value) -> double`
`126`	`126`	`{`
`127`		`-# if BOOST_ARCH_PTX >= BOOST_VERSION_NUMBER(6, 0, 0)`
	`127`	`+# if BOOST_ARCH_PTX >= BOOST_VERSION_NUMBER(6, 0, 0) \|\| BOOST_LANG_HIP`
`128`	`128`	`return ::atomicAdd(addr, value);`
`129`	`129`	`# else`
`130`	`130`	`// Code from: http://docs.nvidia.com/cuda/cuda-c-programming-guide/#atomic-functions`