diff --git a/arch/x86/core/Kconfig.ia32 b/arch/x86/core/Kconfig.ia32
index 593ba0da84a..864318a5f8f 100644
--- a/arch/x86/core/Kconfig.ia32
+++ b/arch/x86/core/Kconfig.ia32
@@ -86,7 +86,8 @@ config X86_PAE
 	  more memory for paging structures.
 
 menu "Architecture Floating Point Options"
-depends on CPU_HAS_FPU
+
+if CPU_HAS_FPU
 
 config SSE
 	bool "SSE registers"
@@ -138,6 +139,15 @@ config LAZY_FPU_SHARING
 	  malware to read the contents of all floating point registers, see
 	  CVE-2018-3665.
 
+endif # CPU_HAS_FPU
+
+config X86_FP_USE_SOFT_FLOAT
+	bool
+	prompt "Use Software Floating Point Operations" if !(NEWLIB_LIBC && !FPU)
+	default y if NEWLIB_LIBC && !FPU
+	help
+	  Enable using software floating point operations.
+
 endmenu
 
 config DCACHE_LINE_SIZE
diff --git a/arch/x86/core/ia32.cmake b/arch/x86/core/ia32.cmake
index f10276052fe..87964172e98 100644
--- a/arch/x86/core/ia32.cmake
+++ b/arch/x86/core/ia32.cmake
@@ -33,3 +33,8 @@ zephyr_library_sources_ifdef(
 
 # Last since we declare default exception handlers here
 zephyr_library_sources(ia32/fatal.c)
+
+zephyr_library_sources_ifdef(
+  CONFIG_X86_FP_USE_SOFT_FLOAT
+  ia32/soft_float_stubs.c
+)
diff --git a/arch/x86/core/ia32/soft_float_stubs.c b/arch/x86/core/ia32/soft_float_stubs.c
new file mode 100644
index 00000000000..fe3e81a1b32
--- /dev/null
+++ b/arch/x86/core/ia32/soft_float_stubs.c
@@ -0,0 +1,280 @@
+/*
+ * Copyright (c) 2021 Intel Corporation
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include <kernel.h>
+#include <toolchain.h>
+
+/**
+ * @file
+ * @brief Provide soft float function stubs for long double operations.
+ *
+ * GCC soft float does not support long double so these need to be
+ * stubbed out.
+ *
+ * The function names come from the GCC public documentation.
+ */
+
+extern void abort(void);
+
+__weak void __addtf3(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __addxf3(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __subtf3(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __subxf3(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __multf3(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __mulxf3(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __divtf3(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __divxf3(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __negtf2(long double a)
+{
+	k_oops();
+}
+
+__weak void __negxf2(long double a)
+{
+	k_oops();
+}
+
+__weak void __extendsftf2(float a)
+{
+	k_oops();
+}
+
+__weak void __extendsfxf2(float a)
+{
+	k_oops();
+}
+
+__weak void __extenddftf2(double a)
+{
+	k_oops();
+}
+
+__weak void __extenddfxf2(double a)
+{
+	k_oops();
+}
+
+__weak void __truncxfdf2(long double a)
+{
+	k_oops();
+}
+
+__weak void __trunctfdf2(long double a)
+{
+	k_oops();
+}
+
+__weak void __truncxfsf2(long double a)
+{
+	k_oops();
+}
+
+__weak void __trunctfsf2(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixtfsi(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixxfsi(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixtfdi(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixxfdi(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixtfti(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixxfti(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixunstfsi(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixunsxfsi(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixunstfdi(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixunsxfdi(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixunstfti(long double a)
+{
+	k_oops();
+}
+
+__weak void __fixunsxfti(long double a)
+{
+	k_oops();
+}
+
+__weak void __floatsitf(int i)
+{
+	k_oops();
+}
+
+__weak void __floatsixf(int i)
+{
+	k_oops();
+}
+
+__weak void __floatditf(long i)
+{
+	k_oops();
+}
+
+__weak void __floatdixf(long i)
+{
+	k_oops();
+}
+
+__weak void __floattitf(long long i)
+{
+	k_oops();
+}
+
+__weak void __floattixf(long long i)
+{
+	k_oops();
+}
+
+__weak void __floatunsitf(unsigned int i)
+{
+	k_oops();
+}
+
+__weak void __floatunsixf(unsigned int i)
+{
+	k_oops();
+}
+
+__weak void __floatunditf(unsigned long i)
+{
+	k_oops();
+}
+
+__weak void __floatundixf(unsigned long i)
+{
+	k_oops();
+}
+
+__weak void __floatuntitf(unsigned long long i)
+{
+	k_oops();
+}
+
+__weak void __floatuntixf(unsigned long long i)
+{
+	k_oops();
+}
+
+__weak void __cmptf2(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __unordtf2(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __eqtf2(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __netf2(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __getf2(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __lttf2(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __letf2(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __gttf2(long double a, long double b)
+{
+	k_oops();
+}
+
+__weak void __powitf2(long double a, int b)
+{
+	k_oops();
+}
+
+__weak void __powixf2(long double a, int b)
+{
+	k_oops();
+}
diff --git a/cmake/compiler/gcc/target_x86.cmake b/cmake/compiler/gcc/target_x86.cmake
index 4684945e751..8dac67a0068 100644
--- a/cmake/compiler/gcc/target_x86.cmake
+++ b/cmake/compiler/gcc/target_x86.cmake
@@ -8,6 +8,11 @@ else()
   string(PREPEND CMAKE_ASM_FLAGS "-m32 ")
   string(PREPEND CMAKE_C_FLAGS   "-m32 ")
   string(PREPEND CMAKE_CXX_FLAGS "-m32 ")
+
+  if(CONFIG_X86_FP_USE_SOFT_FLOAT)
+    list(APPEND TOOLCHAIN_C_FLAGS  -msoft-float)
+    list(APPEND TOOLCHAIN_LD_FLAGS -msoft-float)
+  endif()
 endif()
 
 # GNU Assembler, by default on non-Linux targets, treats slashes as