diff --git a/examples/portable/inplaceTraspose/Makefile_cpu b/examples/portable/inplaceTraspose/Makefile_cpu
new file mode 100644
index 00000000..f50b2d45
--- /dev/null
+++ b/examples/portable/inplaceTraspose/Makefile_cpu
@@ -0,0 +1,9 @@
+
+EXAMPLE=inplaceTranspose
+CPP_SRC=inplaceTranspose.cpp 
+ISPC_SRC=inplaceTranspose.ispc
+ISPC_IA_TARGETS=avx1-i32x8
+ISPC_ARM_TARGETS=neon
+#ISPC_FLAGS=-DDEBUG -g
+
+include ../common_cpu.mk
diff --git a/examples/portable/inplaceTraspose/Makefile_knc b/examples/portable/inplaceTraspose/Makefile_knc
new file mode 100644
index 00000000..1204364f
--- /dev/null
+++ b/examples/portable/inplaceTraspose/Makefile_knc
@@ -0,0 +1,7 @@
+EXAMPLE=radixSort
+CXX_SRC=radixSort.cpp 
+ISPC_SRC=radixSort.ispc
+ISPC_INTRINSICS=../../intrinsics/knc-i1x16.h
+ISPC_TARGET=generic-16
+
+include ../common_knc.mk
diff --git a/examples/portable/inplaceTraspose/Makefile_ptx b/examples/portable/inplaceTraspose/Makefile_ptx
new file mode 100644
index 00000000..da7494e4
--- /dev/null
+++ b/examples/portable/inplaceTraspose/Makefile_ptx
@@ -0,0 +1,15 @@
+PROG=radixSort
+ISPC_SRC=radixSort.ispc
+
+CU_SRC=radixSort.cu
+# NVCC_FLAGS=-Xptxas=-O1
+CXX_SRC=radixSort.cpp  radixSort.cpp
+PTXCC_REGMAX=64
+
+LLVM_GPU=1
+NVVM_GPU=1
+
+include ../common_ptx.mk
+
+
+
diff --git a/examples/portable/inplaceTraspose/inplaceTranspose.cpp b/examples/portable/inplaceTraspose/inplaceTranspose.cpp
new file mode 100644
index 00000000..08e0c806
--- /dev/null
+++ b/examples/portable/inplaceTraspose/inplaceTranspose.cpp
@@ -0,0 +1,114 @@
+#include <cstdio>
+#include <cstdlib>
+#include <algorithm>
+#include <iostream>
+#include <cassert>
+#include <iomanip>
+#include <vector>
+#include "timing.h"
+#include "ispc_malloc.h"
+#include "inplaceTranspose_ispc.h"
+#include "typeT.h"
+
+/* progress bar by Ross Hemsley;
+ * http://www.rosshemsley.co.uk/2011/02/creating-a-progress-bar-in-c-or-any-other-console-app/ */
+static inline void progressbar (unsigned int x, unsigned int n, unsigned int w = 50)
+{
+  if (n < 100)
+  {
+    x *= 100/n;
+    n = 100;
+  }
+
+  if ((x != n) && (x % (n/100) != 0)) return;
+
+  using namespace std;
+  float ratio  =  x/(float)n;
+  int c =  ratio * w;
+
+  cerr << setw(3) << (int)(ratio*100) << "% [";
+  for (int x=0; x<c; x++) cerr << "=";
+  for (int x=c; x<w; x++) cerr << " ";
+  cerr << "]\r" << flush;
+}
+
+int main(int argc, char * argv[])
+{
+  int m = argc > 1 ? atoi(argv[1]) : 8;
+  int n = argc > 2 ? atoi(argv[2]) : 12;
+  bool verbose = argc > 3;
+
+
+  fprintf(stderr, " m= %d  n= %d :: storage= %g MB\n", m, n,
+      m*n*sizeof(int)*2/1e6);
+
+
+  std::vector< std::pair<int,int> > A(m*n);
+
+  int nrep = 10;
+  double dt = 1e10;
+  for (int r = 0; r < nrep; r++)
+  {
+    for (int j = 0; j < n; j++)
+      for (int i = 0; i < m; i++)
+        A[j*m+i] = std::make_pair(i,j);
+
+    if (r == 0 && verbose)
+    {
+      fprintf(stderr, "Original: \n");
+      for (int j = 0; j < n; j++)
+      {
+        for (int i = 0; i < m; i++)
+        {
+          fprintf(stderr, "(%2d,%2d) ", A[j*m+i].first, A[j*m+i].second);
+        }
+        fprintf(stderr, "\n");
+      }
+      fprintf(stderr, "\n");
+      for (int i = 0; i < m*n; i++)
+        fprintf(stderr, "(%2d,%2d) ", A[i].first, A[i].second);
+      fprintf(stderr, "\n");
+      fprintf(stderr, "\n");
+    }
+
+    for (int j = 0; j < n; j++)
+      for (int i = 0; i < m; i++)
+        assert(A[j*m+i].first == i && A[j*m+i].second == j);
+
+    reset_and_start_timer();
+    ispc::transpose((T*)&A[0], n, m);
+    const double t1 = rtc();
+    dt = std::min(dt, get_elapsed_msec());
+    progressbar (r, nrep);
+  }
+  progressbar (nrep, nrep);
+  fprintf(stderr, "\n");
+
+  if (verbose)
+  {
+    fprintf(stderr, "Transposed: \n");
+    for (int j = 0; j < m; j++)
+    {
+      for (int i = 0; i < n; i++)
+      {
+        fprintf(stderr, "(%2d,%2d) ", A[j*n+i].first, A[j*n+i].second);
+      }
+      fprintf(stderr, "\n");
+    }
+    fprintf(stderr, "\n");
+    for (int i = 0; i < m*n; i++)
+      fprintf(stderr, "(%2d,%2d) ", A[i].first, A[i].second);
+    fprintf(stderr, "\n");
+    fprintf(stderr, "\n");
+  }
+
+  for (int j = 0; j < m; j++)
+    for (int i = 0; i < n; i++)
+      assert(A[j*n+i].first == j && A[j*n+i].second == i);
+
+  fprintf(stderr, " tranpose done in %g msec :: BW= % GB/s\n", 
+      dt , 2*m*n*sizeof(int)*2/dt*1e3/1e9);
+
+
+  return 0;
+}
diff --git a/examples/portable/inplaceTraspose/inplaceTranspose.ispc b/examples/portable/inplaceTraspose/inplaceTranspose.ispc
new file mode 100644
index 00000000..08f1d017
--- /dev/null
+++ b/examples/portable/inplaceTraspose/inplaceTranspose.ispc
@@ -0,0 +1,91 @@
+#include "typeT.h"
+
+static inline
+uniform int gcd(uniform int a, uniform int b)
+{
+  while ( a != 0 )
+  {
+    uniform int c = a; 
+    a = b%a;  
+    b = c;
+  }
+  return b;
+}
+
+  static inline
+int __rj(const int i, const uniform int joverb, const uniform int m, const uniform int b)
+{
+  return (i + joverb) % m;
+}
+
+  static inline
+int __di(const int i, const uniform int j, const uniform int m, const uniform int n, const uniform int joverb)
+{
+  return (((i+joverb) % m) + j*m) % n;
+}
+
+  static inline
+int __sj(const int i, const uniform int j, const uniform int m, const uniform int n, const int iovera)
+{
+  return (j + i*n - iovera) % m;
+}
+
+static inline 
+void transpose_serial(uniform T A[], const uniform int m, const uniform int n)
+{
+  const uniform int tmpSize = max(m,n) * programCount;
+  uniform T   * uniform tmp    = uniform new uniform T  [tmpSize];
+  uniform int * uniform joverb = uniform new uniform int[n];
+  uniform int * uniform iovera = uniform new uniform int[m];
+
+  uniform T (*uniform tmp2D)[programCount] = (uniform T (*uniform)[programCount])tmp;
+
+  const uniform int c = gcd(m,n);
+  const uniform int a = m/c;
+  const uniform int b = n/c;
+  foreach (j = 0 ... n)
+    joverb[j] = j/b;
+  foreach (i = 0 ... m)
+    iovera[i] = i/a;
+
+  if (c > 1)
+  {
+    for (uniform int j = 0; j < n; j++)
+    {
+      const uniform int base = j*m;
+      const uniform int __joverb = joverb[j];
+      foreach (i = 0 ... m)
+        tmp[i] = A[base + __rj(i,__joverb,m,b)];
+      foreach (i = 0 ... m)
+        A[base + i] = tmp[i];
+    }
+  }
+
+  foreach (i = 0 ... m)
+  {
+    for (uniform int j = 0; j < n; j++)
+      tmp2D[__di(i,j,m,n,joverb[j])][programIndex] = A[j*m + i];
+    for (uniform int j = 0; j < n; j++)
+      A[j*m + i] = tmp2D[j][programIndex];
+  }
+
+  for (uniform int j = 0; j < n; j++)
+  {
+    const uniform int base = j*m;
+    foreach (i = 0 ... m)
+      tmp[i] = A[base + __sj(i,j,m,n,iovera[i])];
+    foreach (i = 0 ... m)
+      A[base + i] = tmp[i];
+  }
+
+  delete iovera;
+  delete joverb;
+  delete tmp;
+}
+
+export 
+void transpose(uniform T A[], const uniform int m, const uniform int n)
+{
+  transpose_serial(A, m, n);
+}
+
diff --git a/examples/portable/inplaceTraspose/typeT.h b/examples/portable/inplaceTraspose/typeT.h
new file mode 100644
index 00000000..a0aa630a
--- /dev/null
+++ b/examples/portable/inplaceTraspose/typeT.h
@@ -0,0 +1,2 @@
+#pragma once
+typedef double T;