diff --git a/src/soft_float/float_from_int/__floatsitf.c b/src/soft_float/float_from_int/__floatsitf.c
new file mode 100644
index 0000000..a1d23f1
--- /dev/null
+++ b/src/soft_float/float_from_int/__floatsitf.c
@@ -0,0 +1,54 @@
+/* SPDX-FileCopyrightText: © 2023 Decompollaborate */
+/* SPDX-License-Identifier: MIT */
+
+#include "gcc_vr4300/types.h"
+#include "gcc_vr4300/export.h"
+
+#if ABI_N32 || ABI_N64
+EXPORT(__floatsitf);
+
+/**
+ * Gets float64 from int64_t.
+ *
+ * https://gcc.gnu.org/onlinedocs/gccint/the-gcc-low-level-runtime-library/routines-for-floating-point-emulation.html#c.__floatsitf
+ */
+float128 __floatsitf(int32_t i) {
+    Float128Union ret;
+    int32_t sign = 0;
+    int32_t exponent;
+    uint32_t value;
+
+    ret.hex.upper = 0;
+    ret.hex.lower = 0;
+
+    if (i == 0) {
+        return 0;
+    }
+
+    if (i < 0) {
+        sign = 1;
+        i = (~i) + 1;
+    }
+    value = i;
+
+    exponent = -1;
+    while (value > 0) {
+        exponent++;
+        value = value >> 1;
+    }
+
+    // set sign
+    ret.hex.upper |= ((uint64_t)sign) << 63;
+
+    // hide the leading bit
+    i &= ~(1 << exponent);
+
+    // set the biased exponent
+    ret.hex.upper |= ((uint64_t)(exponent + 0x3FFF)) << 48;
+
+    // set the mantissa
+    ret.hex.upper |= ((uint64_t)i) << (48 - exponent);
+
+    return ret.ld;
+}
+#endif