Perform sse float comparisons with the floating-point intrinsics (#5115)

Signed-off-by: Chris Burel <burelc@amazon.com>

Code/Framework/AzCore/AzCore/Math/Internal/SimdMathCommon_sse.inl

@@ -383,36 +383,36 @@ namespace AZ
 
             AZ_MATH_INLINE bool CmpAllEq(__m128 arg1, __m128 arg2, int32_t mask)
             {
-                const __m128i compare = CastToInt(CmpNeq(arg1, arg2));
-                return (_mm_movemask_epi8(compare) & mask) == 0;
+                const __m128 compare = CmpEq(arg1, arg2);
+                return (_mm_movemask_ps(compare) & mask) == mask;
             }
 
 
             AZ_MATH_INLINE bool CmpAllLt(__m128 arg1, __m128 arg2, int32_t mask)
             {
-                const __m128i compare = CastToInt(CmpGtEq(arg1, arg2));
-                return (_mm_movemask_epi8(compare) & mask) == 0;
+                const __m128 compare = CmpLt(arg1, arg2);
+                return (_mm_movemask_ps(compare) & mask) == mask;
             }
 
 
             AZ_MATH_INLINE bool CmpAllLtEq(__m128 arg1, __m128 arg2, int32_t mask)
             {
-                const __m128i compare = CastToInt(CmpGt(arg1, arg2));
-                return (_mm_movemask_epi8(compare) & mask) == 0;
+                const __m128 compare = CmpLtEq(arg1, arg2);
+                return (_mm_movemask_ps(compare) & mask) == mask;
             }
 
 
             AZ_MATH_INLINE bool CmpAllGt(__m128 arg1, __m128 arg2, int32_t mask)
             {
-                const __m128i compare = CastToInt(CmpLtEq(arg1, arg2));
-                return (_mm_movemask_epi8(compare) & mask) == 0;
+                const __m128 compare = CmpGt(arg1, arg2);
+                return (_mm_movemask_ps(compare) & mask) == mask;
             }
 
 
             AZ_MATH_INLINE bool CmpAllGtEq(__m128 arg1, __m128 arg2, int32_t mask)
             {
-                const __m128i compare = CastToInt(CmpLt(arg1, arg2));
-                return (_mm_movemask_epi8(compare) & mask) == 0;
+                const __m128 compare = CmpGtEq(arg1, arg2);
+                return (_mm_movemask_ps(compare) & mask) == mask;
             }
 
 

Code/Framework/AzCore/AzCore/Math/Internal/SimdMathVec1_sse.inl

@@ -331,31 +331,32 @@ namespace AZ
 
         AZ_MATH_INLINE bool Vec1::CmpAllEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllEq(arg1, arg2, 0x000F);
+            // Only check the first bit for Vector1
+            return Sse::CmpAllEq(arg1, arg2, 0b0001);
         }
 
 
         AZ_MATH_INLINE bool Vec1::CmpAllLt(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllLt(arg1, arg2, 0x000F);
+            return Sse::CmpAllLt(arg1, arg2, 0b0001);
         }
 
 
         AZ_MATH_INLINE bool Vec1::CmpAllLtEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllLtEq(arg1, arg2, 0x000F);
+            return Sse::CmpAllLtEq(arg1, arg2, 0b0001);
         }
 
 
         AZ_MATH_INLINE bool Vec1::CmpAllGt(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllGt(arg1, arg2, 0x000F);
+            return Sse::CmpAllGt(arg1, arg2, 0b0001);
         }
 
 
         AZ_MATH_INLINE bool Vec1::CmpAllGtEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllGtEq(arg1, arg2, 0x000F);
+            return Sse::CmpAllGtEq(arg1, arg2, 0b0001);
         }
 
 
@@ -397,7 +398,7 @@ namespace AZ
 
         AZ_MATH_INLINE bool Vec1::CmpAllEq(Int32ArgType arg1, Int32ArgType arg2)
         {
-            return Sse::CmpAllEq(arg1, arg2, 0x000F);
+            return Sse::CmpAllEq(arg1, arg2, 0b0001);
         }
 
 

Code/Framework/AzCore/AzCore/Math/Internal/SimdMathVec2_sse.inl

@@ -383,31 +383,32 @@ namespace AZ
 
         AZ_MATH_INLINE bool Vec2::CmpAllEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllEq(arg1, arg2, 0x00FF);
+            // Only check the first two bits for Vector2
+            return Sse::CmpAllEq(arg1, arg2, 0b0011);
         }
 
 
         AZ_MATH_INLINE bool Vec2::CmpAllLt(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllLt(arg1, arg2, 0x00FF);
+            return Sse::CmpAllLt(arg1, arg2, 0b0011);
         }
 
 
         AZ_MATH_INLINE bool Vec2::CmpAllLtEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllLtEq(arg1, arg2, 0x00FF);
+            return Sse::CmpAllLtEq(arg1, arg2, 0b0011);
         }
 
 
         AZ_MATH_INLINE bool Vec2::CmpAllGt(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllGt(arg1, arg2, 0x00FF);
+            return Sse::CmpAllGt(arg1, arg2, 0b0011);
         }
 
 
         AZ_MATH_INLINE bool Vec2::CmpAllGtEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllGtEq(arg1, arg2, 0x00FF);
+            return Sse::CmpAllGtEq(arg1, arg2, 0b0011);
         }
 
 

Code/Framework/AzCore/AzCore/Math/Internal/SimdMathVec3_sse.inl

@@ -419,31 +419,32 @@ namespace AZ
 
         AZ_MATH_INLINE bool Vec3::CmpAllEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllEq(arg1, arg2, 0x0FFF);
+            // Only check the first three bits for Vector3
+            return Sse::CmpAllEq(arg1, arg2, 0b0111);
         }
 
 
         AZ_MATH_INLINE bool Vec3::CmpAllLt(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllLt(arg1, arg2, 0x0FFF);
+            return Sse::CmpAllLt(arg1, arg2, 0b0111);
         }
 
 
         AZ_MATH_INLINE bool Vec3::CmpAllLtEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllLtEq(arg1, arg2, 0x0FFF);
+            return Sse::CmpAllLtEq(arg1, arg2, 0b0111);
         }
 
 
         AZ_MATH_INLINE bool Vec3::CmpAllGt(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllGt(arg1, arg2, 0x0FFF);
+            return Sse::CmpAllGt(arg1, arg2, 0b0111);
         }
 
 
         AZ_MATH_INLINE bool Vec3::CmpAllGtEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllGtEq(arg1, arg2, 0x0FFF);
+            return Sse::CmpAllGtEq(arg1, arg2, 0b0111);
         }
 
 
@@ -485,7 +486,7 @@ namespace AZ
 
         AZ_MATH_INLINE bool Vec3::CmpAllEq(Int32ArgType arg1, Int32ArgType arg2)
         {
-            return Sse::CmpAllEq(arg1, arg2, 0x0FFF);
+            return Sse::CmpAllEq(arg1, arg2, 0b0111);
         }
 
 

Code/Framework/AzCore/AzCore/Math/Internal/SimdMathVec4_sse.inl

@@ -455,31 +455,32 @@ namespace AZ
 
         AZ_MATH_INLINE bool Vec4::CmpAllEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllEq(arg1, arg2, 0xFFFF);
+            // Check the first four bits for Vector4
+            return Sse::CmpAllEq(arg1, arg2, 0b1111);
         }
 
 
         AZ_MATH_INLINE bool Vec4::CmpAllLt(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllLt(arg1, arg2, 0xFFFF);
+            return Sse::CmpAllLt(arg1, arg2, 0b1111);
         }
 
 
         AZ_MATH_INLINE bool Vec4::CmpAllLtEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllLtEq(arg1, arg2, 0xFFFF);
+            return Sse::CmpAllLtEq(arg1, arg2, 0b1111);
         }
 
 
         AZ_MATH_INLINE bool Vec4::CmpAllGt(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllGt(arg1, arg2, 0xFFFF);
+            return Sse::CmpAllGt(arg1, arg2, 0b1111);
         }
 
 
         AZ_MATH_INLINE bool Vec4::CmpAllGtEq(FloatArgType arg1, FloatArgType arg2)
         {
-            return Sse::CmpAllGtEq(arg1, arg2, 0xFFFF);
+            return Sse::CmpAllGtEq(arg1, arg2, 0b1111);
         }
 
 
@@ -521,7 +522,7 @@ namespace AZ
 
         AZ_MATH_INLINE bool Vec4::CmpAllEq(Int32ArgType arg1, Int32ArgType arg2)
         {
-            return Sse::CmpAllEq(arg1, arg2, 0xFFFF);
+            return Sse::CmpAllEq(arg1, arg2, 0b1111);
         }