diff --git a/CMakeLists.txt b/CMakeLists.txt
index 804998c..f9db75a 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -9,19 +9,22 @@ set(CMAKE_CXX_STANDARD_REQUIRED ON)
 option(STREFLOP_AUTO "Autodetect the most appropriate Streflop implementation" ON)
 option(STREFLOP_SSE "Use SSE instead of x87 extended precision" OFF)
 option(STREFLOP_NEON "Use ARM NEON (auto-enabled on ARM64 when SSE/SOFT are off)" OFF)
+option(STREFLOP_WASM "Use WebAssembly fixed round-to-nearest arithmetic" OFF)
 option(STREFLOP_SOFT "Use soft float" OFF)
 option(STREFLOP_NO_DENORMALS "Squash denormals to zero" OFF)
 option(BUILD_TESTS "Build test executables" OFF)
 
 # Autodetect STREFLOP mode when STREFLOP_AUTO is ON and no explicit mode is selected
-if(STREFLOP_AUTO AND NOT STREFLOP_SSE AND NOT STREFLOP_NEON AND NOT STREFLOP_SOFT AND NOT STREFLOP_NO_DENORMALS AND NOT BUILD_TESTS)
+if(STREFLOP_AUTO AND NOT STREFLOP_SSE AND NOT STREFLOP_NEON AND NOT STREFLOP_SOFT AND NOT STREFLOP_WASM AND NOT STREFLOP_NO_DENORMALS AND NOT BUILD_TESTS)
 	# Use HOST processor as fallback when cross-compiling (CMAKE_SYSTEM_PROCESSOR may be empty)
 	if(NOT CMAKE_SYSTEM_PROCESSOR AND CMAKE_HOST_SYSTEM_PROCESSOR)
 		set(_streflop_processor "${CMAKE_HOST_SYSTEM_PROCESSOR}")
 	else()
 		set(_streflop_processor "${CMAKE_SYSTEM_PROCESSOR}")
 	endif()
-	if(_streflop_processor MATCHES "arm64|aarch64|armv8|ARM64|AARCH64")
+	if(EMSCRIPTEN)
+		set(STREFLOP_WASM ON CACHE BOOL "Use WebAssembly arithmetic" FORCE)
+	elseif(_streflop_processor MATCHES "arm64|aarch64|armv8|ARM64|AARCH64")
 		set(STREFLOP_NEON ON CACHE BOOL "Use ARM NEON (autodetected)" FORCE)
 		message(STATUS "STREFLOP: Autodetected ARM64 platform - enabling NEON mode")
 	elseif(_streflop_processor MATCHES "x86|i686|amd64|AMD64|x86_64")
@@ -264,7 +267,7 @@ SET_SOURCE_FILES_PROPERTIES(${libm_flt32_source} PROPERTIES COMPILE_FLAGS "-DLIB
 SET_SOURCE_FILES_PROPERTIES(${libm_dbl64_source} PROPERTIES COMPILE_FLAGS "-DLIBM_COMPILING_DBL64 ${streflop_cxx_flags} ${libm_extra_flags}")
 
 # Conditionally add long double sources (not for SSE or NEON mode — no x87 extended precision)
-if(NOT STREFLOP_SSE AND NOT STREFLOP_NEON)
+if(NOT STREFLOP_SSE AND NOT STREFLOP_NEON AND NOT STREFLOP_WASM)
 	SET_SOURCE_FILES_PROPERTIES(${libm_ldbl96_source} PROPERTIES COMPILE_FLAGS "-DLIBM_COMPILING_LDBL96 ${streflop_cxx_flags} ${libm_extra_flags}")
 	SET(full_libm_source ${libm_flt32_source} ${libm_dbl64_source} ${libm_ldbl96_source})
 else()
@@ -322,6 +325,9 @@ target_include_directories(streflop PUBLIC
 # Set FPU type compile definitions and flags
 if(STREFLOP_SOFT)
 	target_compile_definitions(streflop PUBLIC STREFLOP_SOFT)
+elseif(STREFLOP_WASM)
+	target_compile_definitions(streflop PUBLIC STREFLOP_WASM)
+	target_compile_options(streflop PRIVATE -ffp-contract=off -fno-fast-math)
 elseif(STREFLOP_SSE)
 	target_compile_definitions(streflop PUBLIC STREFLOP_SSE)
 	if(NOT MSVC)
diff --git a/FPUSettings.h b/FPUSettings.h
index 5883b36..b4d09bf 100644
--- a/FPUSettings.h
+++ b/FPUSettings.h
@@ -625,6 +625,22 @@ template<> inline void streflop_init<Double>() {
 #error "Extended precision not supported on ARM NEON"
 #endif
 
+#elif defined(STREFLOP_WASM)
+
+// WebAssembly has fixed round-to-nearest/ties-to-even, gradual underflow,
+// and no observable FP exception flags or configurable FP traps. Unsupported
+// environment requests fail instead of pretending to change the hardware.
+struct fpenv_t { int rounding; };
+extern fpenv_t FE_DFL_ENV;
+inline int fegetround() { return FE_TONEAREST; }
+inline int fesetround(FPU_RoundMode mode) { return mode == FE_TONEAREST ? 0 : -1; }
+inline int feraiseexcept(FPU_Exceptions excepts) { return excepts == 0 ? 0 : -1; }
+inline int feclearexcept(int) { return 0; }
+inline int fegetenv(fpenv_t* env) { env->rounding = FE_TONEAREST; return 0; }
+inline int fesetenv(const fpenv_t* env) { return env->rounding == FE_TONEAREST ? 0 : -1; }
+inline int feholdexcept(fpenv_t* env) { return fegetenv(env); }
+template<typename T> inline void streflop_init() {}
+
 #elif defined(STREFLOP_SOFT)
 /// Raise exception for these flags
 inline int feraiseexcept(FPU_Exceptions excepts) {
diff --git a/SMath.cpp b/SMath.cpp
index 5d317ed..81b5d65 100644
--- a/SMath.cpp
+++ b/SMath.cpp
@@ -35,7 +35,7 @@ namespace streflop {
     fpenv_t FE_DFL_ENV = 0;
 #elif defined(STREFLOP_SSE)
     fpenv_t FE_DFL_ENV = {0,0};
-#elif defined(STREFLOP_NEON)
+#elif defined(STREFLOP_NEON) || defined(STREFLOP_WASM)
     fpenv_t FE_DFL_ENV = {0};
 #elif defined(STREFLOP_SOFT)
     fpenv_t FE_DFL_ENV = {42,0,0};
diff --git a/SMath.h b/SMath.h
index 36187b8..620876b 100644
--- a/SMath.h
+++ b/SMath.h
@@ -19,6 +19,15 @@
 // just in case, should already be included
 #include "streflop.h"
 
+// Emscripten's compatibility math header defines these as macros.
+// They must not rename the namespace-scoped float overloads below.
+#ifdef isnanf
+#undef isnanf
+#endif
+#ifdef isinff
+#undef isinff
+#endif
+
 // Names from the libm conversion
 namespace streflop_libm {
 using streflop::Simple;
diff --git a/streflop.h b/streflop.h
index a39c1f1..3336ccf 100644
--- a/streflop.h
+++ b/streflop.h
@@ -12,21 +12,13 @@
 #ifndef STREFLOP_H
 #define STREFLOP_H
 
-// protect against bad defines
-#if   defined(STREFLOP_SSE) && defined(STREFLOP_NEON)
-#error You have to define exactly one of STREFLOP_SSE STREFLOP_NEON STREFLOP_X87 STREFLOP_SOFT, but you defined both STREFLOP_SSE and STREFLOP_NEON
-#elif defined(STREFLOP_SSE) && defined(STREFLOP_X87)
-#error You have to define exactly one of STREFLOP_SSE STREFLOP_NEON STREFLOP_X87 STREFLOP_SOFT, but you defined both STREFLOP_SSE and STREFLOP_X87
-#elif defined(STREFLOP_SSE) && defined(STREFLOP_SOFT)
-#error You have to define exactly one of STREFLOP_SSE STREFLOP_NEON STREFLOP_X87 STREFLOP_SOFT, but you defined both STREFLOP_SSE and STREFLOP_SOFT
-#elif defined(STREFLOP_NEON) && defined(STREFLOP_X87)
-#error You have to define exactly one of STREFLOP_SSE STREFLOP_NEON STREFLOP_X87 STREFLOP_SOFT, but you defined both STREFLOP_NEON and STREFLOP_X87
-#elif defined(STREFLOP_NEON) && defined(STREFLOP_SOFT)
-#error You have to define exactly one of STREFLOP_SSE STREFLOP_NEON STREFLOP_X87 STREFLOP_SOFT, but you defined both STREFLOP_NEON and STREFLOP_SOFT
-#elif defined(STREFLOP_X87) && defined(STREFLOP_SOFT)
-#error You have to define exactly one of STREFLOP_SSE STREFLOP_NEON STREFLOP_X87 STREFLOP_SOFT, but you defined both STREFLOP_X87 and STREFLOP_SOFT
-#elif !defined(STREFLOP_SSE) && !defined(STREFLOP_NEON) && !defined(STREFLOP_X87) && !defined(STREFLOP_SOFT)
-#error You have to define exactly one of STREFLOP_SSE STREFLOP_NEON STREFLOP_X87 STREFLOP_SOFT, but you defined none
+// Exactly one backend is required.
+#if (defined(STREFLOP_SSE) + defined(STREFLOP_NEON) + defined(STREFLOP_X87) + defined(STREFLOP_SOFT) + defined(STREFLOP_WASM)) != 1
+#error STREFLOP: Select exactly one floating-point backend.
+#endif
+
+#if defined(STREFLOP_WASM) && (!defined(__wasm__) || defined(STREFLOP_NO_DENORMALS))
+#error STREFLOP_WASM requires WebAssembly with gradual underflow.
 #endif
 
 // First, define the numerical types
@@ -40,9 +32,9 @@ namespace streflop {
     typedef double Double;
     #undef Extended
 
-#elif defined(STREFLOP_NEON)
+#elif defined(STREFLOP_NEON) || defined(STREFLOP_WASM)
 
-    // NEON always uses native types, denormals are handled by FPU flags
+    // NEON and WASM use native IEEE binary32/binary64 types.
     typedef float Simple;
     typedef double Double;
     #undef Extended
diff --git a/streflop_cond.h b/streflop_cond.h
index f199a2e..48589ef 100644
--- a/streflop_cond.h
+++ b/streflop_cond.h
@@ -9,7 +9,7 @@
 #ifndef STREFLOP_COND_H
 #define STREFLOP_COND_H
 
-#if (!defined(NOT_USING_STREFLOP) && (defined(STREFLOP_SSE) || defined(STREFLOP_NEON) || defined(STREFLOP_X87) || defined(STREFLOP_SOFT)))
+#if (!defined(NOT_USING_STREFLOP) && (defined(STREFLOP_SSE) || defined(STREFLOP_NEON) || defined(STREFLOP_X87) || defined(STREFLOP_SOFT) || defined(STREFLOP_WASM)))
 #define STREFLOP_ENABLED 1
 #endif