diff --git a/.gitlab/ci/rules.yml b/.gitlab/ci/rules.yml
index 9fab5014d9..302c1585f3 100644
--- a/.gitlab/ci/rules.yml
+++ b/.gitlab/ci/rules.yml
@@ -144,6 +144,7 @@
   - "components/bt/controller/lib_esp32c6/esp32c6-bt-lib"
   - "components/bt/controller/lib_esp32c5/esp32c5-bt-lib"
   - "components/bt/esp_ble_mesh/lib/lib"
+  - "components/ulp/RVfplib"
   - ".gitmodules"
 
 .patterns-idf-pytest-plugin: &patterns-idf-pytest-plugin
diff --git a/.gitmodules b/.gitmodules
index f306088493..7ee4cba25b 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -147,3 +147,13 @@
 [submodule "components/bt/esp_ble_mesh/lib/lib"]
 	path = components/bt/esp_ble_mesh/lib/lib
 	url = ../../espressif/esp-ble-mesh-lib.git
+
+[submodule "components/ulp/RVfplib"]
+	path = components/ulp/RVfplib
+	url = ../../pulp-platform/RVfplib.git
+	sbom-version = 9de634e25a7130f1d08a2434100aabd12581841b
+	sbom-cpe = cpe:2.3:a:RVfplib:librvfp:{}:*:*:*:*:*:*:*
+	sbom-supplier = Organization: pulp-platform community <https://pulp-platform.org>
+	sbom-url = https://github.com/pulp-platform/RVfplib
+	sbom-description = Optimized RISC-V FP emulation for 32-bit processors
+	sbom-hash = 9de634e25a7130f1d08a2434100aabd12581841b
diff --git a/components/ulp/Kconfig b/components/ulp/Kconfig
index 94167bd5df..f0d1c08b32 100644
--- a/components/ulp/Kconfig
+++ b/components/ulp/Kconfig
@@ -108,4 +108,22 @@ menu "Ultra Low Power (ULP) Co-processor"
 
     endmenu
 
+    menu "ULP float library source"
+        depends on ULP_COPROC_TYPE_LP_CORE || ULP_COPROC_TYPE_RISCV
+
+        config ULP_LP_CORE_FLOAT_LIB_FROM_RVFPLIB
+            bool
+            prompt "librvfp"
+            default "n"
+            help
+                In the soft-fp part of libgcc, RISC-V version is written in C,
+                and handles all edge cases in IEEE754, which makes it larger
+                and performance is slow.
+
+                RVfplib is an optimized RISC-V library for FP arithmetic on 32-bit
+                integer processors, for single and double-precision FP.
+                RVfplib is "fast", but it has a few exceptions from IEEE 754 compliance.
+
+    endmenu
+
 endmenu  # Ultra Low Power (ULP) Co-processor
diff --git a/components/ulp/RVfplib b/components/ulp/RVfplib
new file mode 160000
index 0000000000..9de634e25a
--- /dev/null
+++ b/components/ulp/RVfplib
@@ -0,0 +1 @@
+Subproject commit 9de634e25a7130f1d08a2434100aabd12581841b
diff --git a/components/ulp/cmake/CMakeLists.txt b/components/ulp/cmake/CMakeLists.txt
index 05bddaa68d..b105ade907 100644
--- a/components/ulp/cmake/CMakeLists.txt
+++ b/components/ulp/cmake/CMakeLists.txt
@@ -171,6 +171,15 @@ else()
 
 endif()
 
+set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
+
+if(CONFIG_ULP_LP_CORE_FLOAT_LIB_FROM_RVFPLIB)
+    file(GLOB ULP_S_LIBRVFP_SOURCES "${IDF_PATH}/components/ulp/RVfplib/src/asm/ieee754/*.S")
+    add_library(rvfp STATIC "${ULP_S_LIBRVFP_SOURCES}")
+    target_compile_definitions(rvfp PRIVATE "SIZE=1")
+    target_link_libraries(${ULP_APP_NAME} rvfp)
+endif()
+
 if(ULP_COCPU_IS_LP_CORE)
     set(ULP_BASE_ADDR "0x0")
 else()
diff --git a/components/ulp/project_include.cmake b/components/ulp/project_include.cmake
index b3e35dca96..032ecaa90e 100644
--- a/components/ulp/project_include.cmake
+++ b/components/ulp/project_include.cmake
@@ -71,6 +71,7 @@ function(ulp_embed_binary app_name s_sources exp_dep_srcs)
                             -DULP_COCPU_IS_RISCV=${ULP_IS_RISCV}
                             -DULP_COCPU_IS_LP_CORE=${ULP_IS_LP_CORE_RISCV}
                             -DCONFIG_ESP_ROM_HAS_LP_ROM=${CONFIG_ESP_ROM_HAS_LP_ROM}
+                            -DCONFIG_ULP_LP_CORE_FLOAT_LIB_FROM_RVFPLIB=${CONFIG_ULP_LP_CORE_FLOAT_LIB_FROM_RVFPLIB}
                             ${extra_cmake_args}
                 BUILD_COMMAND ${CMAKE_COMMAND} --build ${CMAKE_CURRENT_BINARY_DIR}/${app_name} --target build
                 BUILD_BYPRODUCTS ${ulp_artifacts} ${ulp_artifacts_extras} ${ulp_ps_sources}