diff --git a/src/Microsoft.ML.CpuMath/AvxIntrinsics.cs b/src/Microsoft.ML.CpuMath/AvxIntrinsics.cs
index b31a427139..54226b5f35 100644
--- a/src/Microsoft.ML.CpuMath/AvxIntrinsics.cs
+++ b/src/Microsoft.ML.CpuMath/AvxIntrinsics.cs
@@ -417,6 +417,7 @@ public static unsafe void AddScalarU(float scalar, Span<float> dst)
             {
                 float* pDstEnd = pdst + dst.Length;
                 float* pDstCurrent = pdst;
+                int destinationEnd = pDstEnd - 4;
 
                 Vector256<float> scalarVector256 = Avx.SetAllVector256(scalar);
 
@@ -431,7 +432,7 @@ public static unsafe void AddScalarU(float scalar, Span<float> dst)
 
                 Vector128<float> scalarVector128 = Sse.SetAllVector128(scalar);
 
-                if (pDstCurrent + 4 <= pDstEnd)
+                if (pDstCurrent <= destinationEnd)
                 {
                     Vector128<float> dstVector = Sse.LoadVector128(pDstCurrent);
                     dstVector = Sse.Add(dstVector, scalarVector128);
@@ -502,6 +503,7 @@ public static unsafe void ScaleSrcU(float scale, Span<float> src, Span<float> ds
                 float* pDstEnd = pdst + dst.Length;
                 float* pSrcCurrent = psrc;
                 float* pDstCurrent = pdst;
+                int destinationEnd = pDstEnd - 4;
 
                 Vector256<float> scaleVector256 = Avx.SetAllVector256(scale);
 
@@ -517,7 +519,7 @@ public static unsafe void ScaleSrcU(float scale, Span<float> src, Span<float> ds
 
                 Vector128<float> scaleVector128 = Sse.SetAllVector128(scale);
 
-                if (pDstCurrent + 4 <= pDstEnd)
+                if (pDstCurrent <= destinationEnd)
                 {
                     Vector128<float> srcVector = Sse.LoadVector128(pSrcCurrent);
                     srcVector = Sse.Multiply(srcVector, scaleVector128);
@@ -546,6 +548,7 @@ public static unsafe void ScaleAddU(float a, float b, Span<float> dst)
             {
                 float* pDstEnd = pdst + dst.Length;
                 float* pDstCurrent = pdst;
+                int destinationEnd = pDstEnd - 4;
 
                 Vector256<float> a256 = Avx.SetAllVector256(a);
                 Vector256<float> b256 = Avx.SetAllVector256(b);
@@ -563,7 +566,7 @@ public static unsafe void ScaleAddU(float a, float b, Span<float> dst)
                 Vector128<float> a128 = Sse.SetAllVector128(a);
                 Vector128<float> b128 = Sse.SetAllVector128(b);
 
-                if (pDstCurrent + 4 <= pDstEnd)
+                if (pDstCurrent <= destinationEnd)
                 {
                     Vector128<float> dstVector = Sse.LoadVector128(pDstCurrent);
                     dstVector = Sse.Add(dstVector, b128);
diff --git a/src/Microsoft.ML.CpuMath/SseIntrinsics.cs b/src/Microsoft.ML.CpuMath/SseIntrinsics.cs
index 4c36d0094e..4a3dae4d82 100644
--- a/src/Microsoft.ML.CpuMath/SseIntrinsics.cs
+++ b/src/Microsoft.ML.CpuMath/SseIntrinsics.cs
@@ -414,10 +414,11 @@ public static unsafe void AddScalarU(float scalar, Span<float> dst)
             {
                 float* pDstEnd = pdst + dst.Length;
                 float* pDstCurrent = pdst;
+                int destinationEnd = pDstEnd - 4;
 
                 Vector128<float> scalarVector = Sse.SetAllVector128(scalar);
 
-                while (pDstCurrent + 4 <= pDstEnd)
+                while (pDstCurrent <= destinationEnd)
                 {
                     Vector128<float> dstVector = Sse.LoadVector128(pDstCurrent);
                     dstVector = Sse.Add(dstVector, scalarVector);
@@ -476,10 +477,11 @@ public static unsafe void ScaleSrcU(float scale, Span<float> src, Span<float> ds
                 float* pDstEnd = pdst + dst.Length;
                 float* pSrcCurrent = psrc;
                 float* pDstCurrent = pdst;
+                int destinationEnd = pDstEnd - 4;
 
                 Vector128<float> scaleVector = Sse.SetAllVector128(scale);
 
-                while (pDstCurrent + 4 <= pDstEnd)
+                while (pDstCurrent <= destinationEnd)
                 {
                     Vector128<float> srcVector = Sse.LoadVector128(pSrcCurrent);
                     srcVector = Sse.Multiply(srcVector, scaleVector);
@@ -508,11 +510,12 @@ public static unsafe void ScaleAddU(float a, float b, Span<float> dst)
             {
                 float* pDstEnd = pdst + dst.Length;
                 float* pDstCurrent = pdst;
+                int destinationEnd = pDstEnd - 4;
 
                 Vector128<float> aVector = Sse.SetAllVector128(a);
                 Vector128<float> bVector = Sse.SetAllVector128(b);
 
-                while (pDstCurrent + 4 <= pDstEnd)
+                while (pDstCurrent <= destinationEnd)
                 {
                     Vector128<float> dstVector = Sse.LoadVector128(pDstCurrent);
                     dstVector = Sse.Add(dstVector, bVector);