Code compiles (under nvidia)

src/pyFAI/resources/openCL/medfilt.cl

@@ -63,9 +63,7 @@ float2 inline sum_float2_reduction(local float* shared)
             here.s1 = shared[pos_here+1];
             there.s0 = shared[pos_there];
             there.s1 = shared[pos_there+1];
-            shared[2*tid+1] = here.s1;
-            here =
-            += shared[tid+stride];
+            here = dw_plus_dw(here, there);
             shared[2*tid] = here.s0;
             shared[2*tid+1] = here.s1;
         }
@@ -113,7 +111,7 @@ csr_medfilt    (  const   global  float4  *data4,
                           global  float   *averint,
                           global  float   *stdevpix,
                           global  float   *stderrmean,
-                          local   int*    shared_int   // size of the workgroup size
+                          local   int*    shared_int,  // size of the workgroup size
                           local   float*  shared_float // size of 2x the workgroup size
                           )
 {
@@ -128,7 +126,7 @@ csr_medfilt    (  const   global  float4  *data4,
     char curr_error_model=error_model;
     float8 result;
     float sum=0.0f, ratio=1.3f;
-    float2 acc_sig, acc_nrm, acc_var;
+    float2 acc_sig, acc_nrm, acc_var, acc_nrm2;
 
     // first populate the work4 array from data4
     for (int i=start+tid; i<stop; i+=wg)
@@ -152,29 +150,29 @@ csr_medfilt    (  const   global  float4  *data4,
     step = first_step(step, size, ratio);
 
     for (step=step; step>0; step=previous_step(step, ratio))
-        cnt = passe_float4(&work4[start], size, step, shared);
+        cnt = passe_float4(&work4[start], size, step, shared_int);
 
     while (cnt)
-        cnt = passe_float4(&work4[start], size, 1, shared);
+        cnt = passe_float4(&work4[start], size, 1, shared_int);
 
     // Then perform the cumsort of the weights to s0
     // In blelloch scan, one workgroup can process 2wg in size.
     niter = (size + 2*wg-1)/(2*wg);
     sum = 0.0f;
-    for (int i=0; i<niter; i+=1)
+    for (int i=0; i<niter; i++)
     {
-        idx += start + tid + 2*wg*i;
+        idx = start + tid + 2*wg*i;
 
         shared_float[tid] = (idx<stop)?work4[idx].s3:0.0f;
         shared_float[tid+wg] = ((idx+wg)<stop)?work4[idx+wg].s3:0.0f;
 
         blelloch_scan_float(shared_float);
 
-        if (idx<size)
-            work4[idx].s0 = sum + shared[lid];
-        if (i+ws<size)
-            work4[idx+wg].s0 = sum + shared[lid+g];
-        sum += shared[2*ws-1];
+        if (idx<stop)
+            work4[idx].s0 = sum + shared_float[tid];
+        if ((i+wg)<stop)
+            work4[idx+wg].s0 = sum + shared_float[tid+wg];
+        sum += shared_float[2*wg-1];
 
     }
     // Perform the sum for accumulator of signal, variance, normalization and count