AMAX/AMIN integer testing and bug fixes (#457)

* Fixed a bug in XAMAX/XMIN routines that caused the increment and offset to be included in the result

* Perform proper integer-output testing in XAMAX tests

* A few changes towards getting it ready for a PR

* Also fix compilation for clBLAS and cuBLAS references

* Fix a bug that would only use the real part of complex numbers in the amax/amin routines

* A few small fixes related to the AMAX tests

2 files changed, 24 insertions, 11 deletions

diff --git a/scripts/generator/generator/cpp.py b/scripts/generator/generator/cpp.py
index 6dc3fc93..e32738ee 100644
--- a/scripts/generator/generator/cpp.py
+++ b/scripts/generator/generator/cpp.py
@@ -226,7 +226,10 @@ def wrapper_clblas(routine):
 
                 # Convert to float (note: also integer buffers are stored as half/float)
                 for buf in routine.inputs + routine.outputs:
-                    result += "  auto " + buf + "_buffer_bis = HalfToFloatBuffer(" + buf + "_buffer, queues[0]);" + NL
+                    if buf not in routine.index_buffers():
+                        result += "  auto " + buf + "_buffer_bis = HalfToFloatBuffer(" + buf + "_buffer, queues[0]);" + NL
+                    else:
+                        result += "  auto " + buf + "_buffer_bis = " + buf + "_buffer;" + NL
 
                 # Call the float routine
                 result += "  auto status = clblasX" + routine.name + "("
@@ -236,7 +239,8 @@ def wrapper_clblas(routine):
 
                 # Convert back to half
                 for buf in routine.outputs:
-                    result += "  FloatToHalfBuffer(" + buf + "_buffer, " + buf + "_buffer_bis, queues[0]);" + NL
+                    if buf not in routine.index_buffers():
+                        result += "  FloatToHalfBuffer(" + buf + "_buffer, " + buf + "_buffer_bis, queues[0]);" + NL
                 result += "  return status;"
 
             # Complete
@@ -276,10 +280,6 @@ def wrapper_cblas(routine):
                             extra_argument += "," + NL + indent
                             extra_argument += "reinterpret_cast<return_pointer_" + flavour.buffer_type[:-1] + ">"
                             extra_argument += "(&" + output_buffer + "_buffer[" + output_buffer + "_offset])"
-                        elif output_buffer in routine.index_buffers():
-                            assignment = "reinterpret_cast<int*>(&" + output_buffer + "_buffer[0])[" + output_buffer + "_offset] = static_cast<int>("
-                            postpostfix = ")"
-                            indent += " " * (len(assignment) + 1)
                         else:
                             assignment = output_buffer + "_buffer[" + output_buffer + "_offset]"
                             if flavour.name in ["Sc", "Dz"]:
@@ -299,7 +299,10 @@ def wrapper_cblas(routine):
 
                 # Convert to float (note: also integer buffers are stored as half/float)
                 for buf in routine.inputs + routine.outputs:
-                    result += "  auto " + buf + "_buffer_bis = HalfToFloatBuffer(" + buf + "_buffer);" + NL
+                    if buf not in routine.index_buffers():
+                        result += "  auto " + buf + "_buffer_bis = HalfToFloatBuffer(" + buf + "_buffer);" + NL
+                    else:
+                        result += "  auto " + buf + "_buffer_bis = " + buf + "_buffer;" + NL
 
                 # Call the float routine
                 result += "  cblasX" + routine.name + "("
@@ -308,7 +311,8 @@ def wrapper_cblas(routine):
 
                 # Convert back to half
                 for buf in routine.outputs:
-                    result += "  FloatToHalfBuffer(" + buf + "_buffer, " + buf + "_buffer_bis);" + NL
+                    if buf not in routine.index_buffers():
+                        result += "  FloatToHalfBuffer(" + buf + "_buffer, " + buf + "_buffer_bis);" + NL
 
             # Complete
             result += "}" + NL
diff --git a/scripts/generator/generator/routine.py b/scripts/generator/generator/routine.py
index 8b6ab57f..c2201c0d 100644
--- a/scripts/generator/generator/routine.py
+++ b/scripts/generator/generator/routine.py
@@ -282,7 +282,10 @@ class Routine:
         """As above but for OpenCL"""
         prefix = "const " if name in self.inputs else ""
         if name in self.inputs or name in self.outputs:
-            a = [prefix + "Buffer<" + flavour.buffer_type + ">& " + name + "_buffer"]
+            if name == "imax":
+                a = [prefix + "Buffer<unsigned int>& " + name + "_buffer"]
+            else:
+                a = [prefix + "Buffer<" + flavour.buffer_type + ">& " + name + "_buffer"]
             b = ["const size_t " + name + "_offset"]
             c = ["const size_t " + name + "_" + self.postfix(name)] if name not in self.buffers_without_ld_inc() else []
             return [", ".join(a + b + c)]
@@ -292,7 +295,10 @@ class Routine:
         """As above but for CUDA"""
         prefix = "const " if name in self.inputs else ""
         if name in self.inputs or name in self.outputs:
-            a = [prefix + flavour.buffer_type + "* " + name + "_buffer"]
+            if name == "imax":
+                a = [prefix + "unsigned int * " + name + "_buffer"]
+            else:
+                a = [prefix + flavour.buffer_type + "* " + name + "_buffer"]
             b = ["const size_t " + name + "_offset"]
             c = ["const size_t " + name + "_" + self.postfix(name)] if name not in self.buffers_without_ld_inc() else []
             return [", ".join(a + b + c)]
@@ -302,7 +308,10 @@ class Routine:
         """As above but as vectors"""
         prefix = "const " if name in self.inputs else ""
         if name in self.inputs or name in self.outputs:
-            a = [prefix + "std::vector<" + flavour.buffer_type + ">& " + name + "_buffer"]
+            if name == "imax":
+                a = [prefix + "std::vector<unsigned int>& " + name + "_buffer"]
+            else:
+                a = [prefix + "std::vector<" + flavour.buffer_type + ">& " + name + "_buffer"]
             b = ["const size_t " + name + "_offset"]
             c = ["const size_t " + name + "_" + self.postfix(name)] if name not in self.buffers_without_ld_inc() else []
             return [", ".join(a + b + c)]