[libclc] Refine __clc_fp*_subnormals_supported and __clc_flush_denormal_if_not_supported

libclc/CMakeLists.txt

@@ -41,8 +41,6 @@ set( LIBCLC_MIN_LLVM 3.9.0 )
 set( LIBCLC_TARGETS_TO_BUILD "all"
     CACHE STRING "Semicolon-separated list of libclc targets to build, or 'all'." )
 
-option( ENABLE_RUNTIME_SUBNORMAL "Enable runtime linking of subnormal support." OFF )
-
 option(
   LIBCLC_USE_SPIRV_BACKEND "Build SPIR-V targets with the SPIR-V backend." OFF
 )
@@ -231,19 +229,6 @@ set( tahiti_aliases pitcairn verde oland hainan bonaire kabini kaveri hawaii
 configure_file( libclc.pc.in libclc.pc @ONLY )
 install( FILES ${CMAKE_CURRENT_BINARY_DIR}/libclc.pc DESTINATION "${CMAKE_INSTALL_DATADIR}/pkgconfig" )
 
-if( ENABLE_RUNTIME_SUBNORMAL )
-  foreach( file IN ITEMS subnormal_use_default subnormal_disable )
-    link_bc(
-       TARGET ${file}
-       INPUTS ${CMAKE_CURRENT_SOURCE_DIR}/opencl/lib/generic/${file}.ll
-    )
-    install(
-      FILES $<TARGET_PROPERTY:${file},TARGET_FILE>
-      DESTINATION "${CMAKE_INSTALL_DATADIR}/clc"
-    )
-  endforeach()
-endif()
-
 find_package( Python3 REQUIRED COMPONENTS Interpreter )
 file( TO_CMAKE_PATH ${CMAKE_CURRENT_SOURCE_DIR}/utils/gen_convert.py script_loc )
 add_custom_command(
@@ -371,9 +356,6 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
       list( APPEND opencl_gen_files clspv-convert.cl )
     else()
       list( APPEND opencl_gen_files convert.cl )
-      if ( NOT ENABLE_RUNTIME_SUBNORMAL )
-        list( APPEND opencl_lib_files opencl/lib/generic/subnormal_use_default.ll )
-      endif()
     endif()
   endif()
 
@@ -430,6 +412,7 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
       # Error on undefined macros
       -Werror=undef
       -fdiscard-value-names
+      -Xclang -fdenormal-fp-math=dynamic
     )
 
     if( NOT "${cpu}" STREQUAL "" )

libclc/clc/include/clc/math/clc_subnormal_config.h

@@ -10,7 +10,6 @@
 
 #include <clc/clcfunc.h>
 
-_CLC_DECL bool __clc_subnormals_disabled();
 _CLC_DECL bool __clc_fp16_subnormals_supported();
 _CLC_DECL bool __clc_fp32_subnormals_supported();
 _CLC_DECL bool __clc_fp64_subnormals_supported();

libclc/clc/include/clc/math/math.h

@@ -11,7 +11,6 @@
 
 #include <clc/clc_as_type.h>
 #include <clc/clcfunc.h>
-#include <clc/math/clc_subnormal_config.h>
 
 #define SNAN 0x001
 #define QNAN 0x002
@@ -24,15 +23,6 @@
 #define PNOR 0x100
 #define PINF 0x200
 
-#if (defined __AMDGCN__ || defined __R600__) && !defined __HAS_FMAF__
-#define __CLC_HAVE_HW_FMA32() (0)
-#elif defined(CLC_SPIRV)
-bool __attribute__((noinline)) __clc_runtime_has_hw_fma32(void);
-#define __CLC_HAVE_HW_FMA32() __clc_runtime_has_hw_fma32()
-#else
-#define __CLC_HAVE_HW_FMA32() (1)
-#endif
-
 #define HAVE_BITALIGN() (0)
 #define HAVE_FAST_FMA32() (0)
 
@@ -65,14 +55,12 @@ bool __attribute__((noinline)) __clc_runtime_has_hw_fma32(void);
 
 #define LOG_MAGIC_NUM_SP32 (1 + NUMEXPBITS_SP32 - EXPBIAS_SP32)
 
-_CLC_OVERLOAD _CLC_INLINE float __clc_flush_denormal_if_not_supported(float x) {
-  int ix = __clc_as_int(x);
-  if (!__clc_fp32_subnormals_supported() && ((ix & EXPBITS_SP32) == 0) &&
-      ((ix & MANTBITS_SP32) != 0)) {
-    ix &= SIGNBIT_SP32;
-    x = __clc_as_float(ix);
-  }
-  return x;
+_CLC_OVERLOAD _CLC_INLINE float __clc_soft_flush_denormal(float x) {
+  // Avoid calling __clc_fp32_subnormals_supported here: it uses
+  // llvm.canonicalize, which quiets sNaN.
+  return __builtin_elementwise_abs(x) < 0x1p-149f
+             ? __builtin_elementwise_copysign(0.0f, x)
+             : x;
 }
 
 #ifdef cl_khr_fp64

libclc/clc/lib/clspv/SOURCES

@@ -1,2 +1 @@
-math/clc_sw_fma.cl
 integer/clc_mul_hi.cl