diff --git a/tests/cpu/ppu_fpu/ppu_fpu.cpp b/tests/cpu/ppu_fpu/ppu_fpu.cpp
new file mode 100644
index 0000000..433b73c
--- /dev/null
+++ b/tests/cpu/ppu_fpu/ppu_fpu.cpp
@@ -0,0 +1,121 @@
+#include <stdio.h>
+#include <math.h>
+#include <float.h>
+#include <ppu_intrinsics.h>
+
+#define PRINT_FPR1(name,i,FPR) printf(name "[%02d] -> %016llx\n", i, *(unsigned long long*)&FPR);
+#define PRINT_FPR2(name,i,j,FPR) printf(name "[%02d],[%02d]) -> %016llx\n", i, j, *(unsigned long long*)&FPR);
+#define PRINT_FPR3(name,i,j,k,FPR) printf(name "[%02d],[%02d],[%02d]) -> %016llx\n", i, j, k, *(unsigned long long*)&FPR);
+
+// Loops with floats
+#define ITERATE1f(x) \
+	for (unsigned int i=0; i<sizeof(testFloats)/sizeof(float); i++) \
+		{float f0, f1=testFloats[i]; x;}
+
+#define ITERATE2f(x) \
+	for (unsigned int i=0; i<sizeof(testFloats)/sizeof(float); i++) \
+		for (unsigned int j=0; j<sizeof(testFloats)/sizeof(float); j++) \
+			{float f0, f1=testFloats[i], f2=testFloats[j]; x;}
+
+#define ITERATE3f(x) \
+	for (unsigned int i=0; i<sizeof(testFloats)/sizeof(float); i++) \
+		for (unsigned int j=0; j<sizeof(testFloats)/sizeof(float); j++) \
+			for (unsigned int k=0; k<sizeof(testFloats)/sizeof(float); k++) \
+				{float f0, f1=testFloats[i], f2=testFloats[j], f3=testFloats[k]; x;}
+
+// Loops with doubles
+#define ITERATE1d(x) \
+	for (unsigned int i=0; i<sizeof(testDoubles)/sizeof(double); i++) \
+		{double d0, d1=testDoubles[i]; x;}
+
+#define ITERATE2d(x) \
+	for (unsigned int i=0; i<sizeof(testDoubles)/sizeof(double); i++) \
+		for (unsigned int j=0; j<sizeof(testDoubles)/sizeof(double); j++) \
+			{double d0, d1=testDoubles[i], d2=testDoubles[j]; x;}
+
+#define ITERATE3d(x) \
+	for (unsigned int i=0; i<sizeof(testDoubles)/sizeof(double); i++) \
+		for (unsigned int j=0; j<sizeof(testDoubles)/sizeof(double); j++) \
+			for (unsigned int k=0; k<sizeof(testDoubles)/sizeof(double); k++) \
+				{double d0, d1=testDoubles[i], d2=testDoubles[j], d3=testDoubles[k]; x;}
+
+
+float testFloats[] = {
+	0.0, -0.0,  1.0, -1.0,  1.5, -1.5,
+	1.6, -1.6,  1.4, -1.4,  2.0, -2.0,
+	4.0, -10000000.4, 20000000.0, -20000.5, 20000.6,
+	FLT_MIN, FLT_MAX, INFINITY, -INFINITY, NAN,
+};
+
+double testDoubles[] = {
+	0.0, -0.0,  1.0, -1.0,  1.5, -1.5,
+	1.6, -1.6,  1.4, -1.4,  2.0, -2.0,
+	4.0, -10000000.4, 20000000.0, -20000.5, 20000.6,
+	DBL_MIN, DBL_MAX, INFINITY, -INFINITY, NAN,
+};
+
+
+int main(void)
+{
+	/////////////////////////////////////
+	// PPU FLOATING-POINT INSTRUCTIONS //
+	/////////////////////////////////////
+
+	// Floating-Point Arithmetic Instructions
+	ITERATE2d(__asm__ ("fadd    %0,%1,%2" : "=f"(d0) : "f"(d1), "f"(d2));  PRINT_FPR2("fadd  ",i,j,d0));
+	ITERATE2f(__asm__ ("fadds   %0,%1,%2" : "=f"(f0) : "f"(f1), "f"(f2));  PRINT_FPR2("fadds ",i,j,f0));
+	ITERATE2d(__asm__ ("fsub    %0,%1,%2" : "=f"(d0) : "f"(d1), "f"(d2));  PRINT_FPR2("fsub  ",i,j,d0));
+	ITERATE2f(__asm__ ("fsubs   %0,%1,%2" : "=f"(f0) : "f"(f1), "f"(f2));  PRINT_FPR2("fsubs ",i,j,f0));
+	ITERATE2d(__asm__ ("fmul    %0,%1,%2" : "=f"(d0) : "f"(d1), "f"(d2));  PRINT_FPR2("fmul  ",i,j,d0));
+	ITERATE2f(__asm__ ("fmuls   %0,%1,%2" : "=f"(f0) : "f"(f1), "f"(f2));  PRINT_FPR2("fmuls ",i,j,f0));
+	ITERATE2d(__asm__ ("fdiv    %0,%1,%2" : "=f"(d0) : "f"(d1), "f"(d2));  PRINT_FPR2("fdiv  ",i,j,d0));
+	ITERATE2f(__asm__ ("fdivs   %0,%1,%2" : "=f"(f0) : "f"(f1), "f"(f2));  PRINT_FPR2("fdivs ",i,j,f0));
+	ITERATE1d(__asm__ ("fsqrt   %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("fsqrt   ",i,d0));
+	ITERATE1f(__asm__ ("fsqrts  %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fsqrts  ",i,f0));
+	ITERATE1f(__asm__ ("fres    %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fres    ",i,f0));
+	ITERATE1d(__asm__ ("frsqrte %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("frsqrte ",i,d0));
+	ITERATE3d(__asm__ ("fsel    %0,%1,%2,%3" : "=f"(d0) : "f"(d1), "f"(d2), "f"(d3));  PRINT_FPR3("fsel    ",i,j,k,d0));
+
+	// Floating-Point Multiply-Add Instructions
+	ITERATE3d(__asm__ ("fmadd   %0,%1,%2,%3" : "=f"(d0) : "f"(d1), "f"(d2), "f"(d3));  PRINT_FPR3("fmadd   ",i,j,k,d0));
+	ITERATE3f(__asm__ ("fmadds  %0,%1,%2,%3" : "=f"(f0) : "f"(f1), "f"(f2), "f"(f3));  PRINT_FPR3("fmadds  ",i,j,k,f0));
+	ITERATE3d(__asm__ ("fmsub   %0,%1,%2,%3" : "=f"(d0) : "f"(d1), "f"(d2), "f"(d3));  PRINT_FPR3("fmsub   ",i,j,k,d0));
+	ITERATE3f(__asm__ ("fmsubs  %0,%1,%2,%3" : "=f"(f0) : "f"(f1), "f"(f2), "f"(f3));  PRINT_FPR3("fmsubs  ",i,j,k,f0));
+	ITERATE3d(__asm__ ("fnmadd  %0,%1,%2,%3" : "=f"(d0) : "f"(d1), "f"(d2), "f"(d3));  PRINT_FPR3("fnmadd  ",i,j,k,d0));
+	ITERATE3f(__asm__ ("fnmadds %0,%1,%2,%3" : "=f"(f0) : "f"(f1), "f"(f2), "f"(f3));  PRINT_FPR3("fnmadds ",i,j,k,f0));
+	ITERATE3d(__asm__ ("fnmsub  %0,%1,%2,%3" : "=f"(d0) : "f"(d1), "f"(d2), "f"(d3));  PRINT_FPR3("fnmsub  ",i,j,k,d0));
+	ITERATE3f(__asm__ ("fnmsubs %0,%1,%2,%3" : "=f"(f0) : "f"(f1), "f"(f2), "f"(f3));  PRINT_FPR3("fnmsubs ",i,j,k,f0));
+
+	// Floating-Point Rounding and Conversion Instructions
+	ITERATE1d(__asm__ ("frsp    %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("frsp   ",i,d0));
+	ITERATE1f(__asm__ ("frsp    %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("frsp   ",i,f0));
+	ITERATE1d(__asm__ ("fcfid   %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("fcfid  ",i,d0));
+	ITERATE1f(__asm__ ("fcfid   %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fcfid  ",i,f0));
+	ITERATE1d(__asm__ ("fctidz  %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("fctidz ",i,d0));
+	ITERATE1f(__asm__ ("fctidz  %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fctidz ",i,f0));
+	ITERATE1d(__asm__ ("fctiw   %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("fctiw  ",i,d0));
+	ITERATE1f(__asm__ ("fctiw   %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fctiw  ",i,f0));
+	ITERATE1d(__asm__ ("fctiwz  %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("fctiwz ",i,d0));
+	ITERATE1f(__asm__ ("fctiwz  %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fctiwz ",i,f0));
+
+	// Floating-Point Compare Instructions
+	ITERATE2d(__asm__ ("fcmpu   %0,%1,%2" : "=f"(d0) : "f"(d1), "f"(d2));  PRINT_FPR2("fcmpu ",i,j,d0));
+	ITERATE2f(__asm__ ("fcmpu   %0,%1,%2" : "=f"(f0) : "f"(f1), "f"(f2));  PRINT_FPR2("fcmpu ",i,j,f0));
+	ITERATE2d(__asm__ ("fcmpo   %0,%1,%2" : "=f"(d0) : "f"(d1), "f"(d2));  PRINT_FPR2("fcmpo ",i,j,d0));
+	ITERATE2f(__asm__ ("fcmpo   %0,%1,%2" : "=f"(f0) : "f"(f1), "f"(f2));  PRINT_FPR2("fcmpo ",i,j,f0));
+
+	// Floating-Point Status and Control Register Instructions
+	// TODO ?
+
+	// Floating-Point Move Instructions
+	ITERATE1d(__asm__ ("fmr     %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("fmr   ",i,d0));
+	ITERATE1f(__asm__ ("fmr     %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fmr   ",i,f0));
+	ITERATE1d(__asm__ ("fneg    %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("fneg  ",i,d0));
+	ITERATE1f(__asm__ ("fneg    %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fneg  ",i,f0));
+	ITERATE1d(__asm__ ("fabs    %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("fabs  ",i,d0));
+	ITERATE1f(__asm__ ("fabs    %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fabs  ",i,f0));
+	ITERATE1d(__asm__ ("fnabs   %0,%1" : "=f"(d0) : "f"(d1));  PRINT_FPR1("fnabs ",i,d0));
+	ITERATE1f(__asm__ ("fnabs   %0,%1" : "=f"(f0) : "f"(f1));  PRINT_FPR1("fnabs ",i,f0));
+
+	return 0;
+}
diff --git a/tests/cpu/ppu_fpu/ppu_fpu.ppu.self b/tests/cpu/ppu_fpu/ppu_fpu.ppu.self
new file mode 100644
index 0000000..0fa7c48
Binary files /dev/null and b/tests/cpu/ppu_fpu/ppu_fpu.ppu.self differ