GPU OpenCL: Improve compiler command line arguments

Author: davidrohr

GPU/Common/GPUCommonMath.h

@@ -399,7 +399,7 @@ GPUdi() T GPUCommonMath::MaxWithRef(T x, T y, T z, T w, S refX, S refY, S refZ,
 
 GPUdi() float GPUCommonMath::InvSqrt(float _x)
 {
-#ifdef GPUCA_NO_FAST_MATH
+#if defined(GPUCA_NO_FAST_MATH) || defined(__OPENCL__)
   return 1.f / Sqrt(_x);
 #elif defined(__CUDACC__) || defined(__HIPCC__)
   return __frsqrt_rn(_x);

GPU/GPUTracking/Base/opencl2/CMakeLists.txt

@@ -23,9 +23,11 @@ endif()
 set(CL_SRC ${GPUDIR}/Base/opencl-common/GPUReconstructionOCL.cl)
 set(CL_BIN ${CMAKE_CURRENT_BINARY_DIR}/GPUReconstructionOCL2Code)
 
-set(OCL_FLAGS -ferror-limit=1000 -Dcl_clang_storage_class_specifiers -Wno-invalid-constexpr -Wno-unused-command-line-argument -cl-std=CLC++2021)
+set(OCL_FLAGS -Dcl_clang_storage_class_specifiers -cl-std=CLC++2021)
 if(NOT DEFINED GPUCA_NO_FAST_MATH OR NOT ${GPUCA_NO_FAST_MATH})
-  set(OCL_FLAGS ${OCL_FLAGS} -Xclang -fdenormal-fp-math-f32=ieee -cl-mad-enable -cl-no-signed-zeros)
+  set(OCL_FLAGS ${OCL_FLAGS} -cl-denorms-are-zero -cl-mad-enable -cl-no-signed-zeros -cl-fast-relaxed-math)
+else()
+set(OCL_FLAGS ${OCL_FLAGS} -cl-fp32-correctly-rounded-divide-sqrt)
 endif()
 set(OCL_DEFINECL "-D$<JOIN:$<TARGET_PROPERTY:O2::GPUTracking,COMPILE_DEFINITIONS>,$<SEMICOLON>-D>"
             "-I$<JOIN:$<FILTER:$<TARGET_PROPERTY:O2::GPUTracking,INCLUDE_DIRECTORIES>,EXCLUDE,^/usr/include/?>,$<SEMICOLON>-I>"
@@ -47,6 +49,7 @@ if(OPENCL2_ENABLED_SPIRV) # BUILD OpenCL2 intermediate code for SPIR-V target
               -O0
               --target=spirv64
               -fno-integrated-objemitter
+              -ferror-limit=1000 -Wno-invalid-constexpr -Wno-unused-command-line-argument
               ${OCL_FLAGS}
               ${OCL_DEFINECL}
               -o ${CL_BIN}.spirv -c ${CL_SRC}
@@ -64,6 +67,7 @@ if(OPENCL2_ENABLED) # BUILD OpenCL2 source code for runtime compilation target
   add_custom_command(
       OUTPUT ${CL_BIN}.src
       COMMAND ${LLVM_CLANG}
+              -Wno-unused-command-line-argument
               ${OCL_FLAGS}
               ${OCL_DEFINECL}
               -cl-no-stdinc