From 2a798152d4bdcd86d1233c5b4ce169916aa76a17 Mon Sep 17 00:00:00 2001
From: Ben Sander <ben.sander@amd.com>
Date: Mon, 8 Aug 2016 12:07:12 -0500
Subject: [PATCH] Move copy kernel templates into hip_memory.cpp

Change-Id: I862529f3fa8232372c6bacaa5d36f035bbdd32a1
---
 include/hcc_detail/hip_hcc.h | 73 ------------------------------------
 src/hip_memory.cpp           | 73 ++++++++++++++++++++++++++++++++++++
 2 files changed, 73 insertions(+), 73 deletions(-)

diff --git a/include/hcc_detail/hip_hcc.h b/include/hcc_detail/hip_hcc.h
index fb8e11aa14..65959fd7d0 100644
--- a/include/hcc_detail/hip_hcc.h
+++ b/include/hcc_detail/hip_hcc.h
@@ -671,82 +671,9 @@ ihipCtx_t * ihipGetPrimaryCtx(unsigned deviceIndex);
 
 extern void ihipSetTs(hipEvent_t e);
 
-template<typename T>
-hc::completion_future ihipMemcpyKernel(hipStream_t, T*, const T*, size_t);
-
-template<typename T>
-hc::completion_future ihipMemsetKernel(hipStream_t, T*, T, size_t);
 
 hipStream_t ihipSyncAndResolveStream(hipStream_t);
-template <typename T>
-
-hc::completion_future
-ihipMemsetKernel(hipStream_t stream, T * ptr, T val, size_t sizeBytes)
-{
-    int wg = std::min((unsigned)8, stream->getDevice()->_compute_units);
-    const int threads_per_wg = 256;
-
-    int threads = wg * threads_per_wg;
-    if (threads > sizeBytes) {
-        threads = ((sizeBytes + threads_per_wg - 1) / threads_per_wg) * threads_per_wg;
-    }
 
 
-    hc::extent<1> ext(threads);
-    auto ext_tile = ext.tile(threads_per_wg);
-
-    hc::completion_future cf =
-    hc::parallel_for_each(
-            stream->_av,
-            ext_tile,
-            [=] (hc::tiled_index<1> idx)
-            __attribute__((hc))
-    {
-        int offset = amp_get_global_id(0);
-        // TODO-HCC - change to hc_get_local_size()
-        int stride = amp_get_local_size(0) * hc_get_num_groups(0) ;
-
-        for (int i=offset; i<sizeBytes; i+=stride) {
-            ptr[i] = val;
-        }
-    });
-
-    return cf;
-}
-
-template <typename T>
-hc::completion_future
-ihipMemcpyKernel(hipStream_t stream, T * c, const T * a, size_t sizeBytes)
-{
-    int wg = std::min((unsigned)8, stream->getDevice()->_compute_units);
-    const int threads_per_wg = 256;
-
-    int threads = wg * threads_per_wg;
-    if (threads > sizeBytes) {
-        threads = ((sizeBytes + threads_per_wg - 1) / threads_per_wg) * threads_per_wg;
-    }
-
-
-    hc::extent<1> ext(threads);
-    auto ext_tile = ext.tile(threads_per_wg);
-
-    hc::completion_future cf =
-    hc::parallel_for_each(
-            stream->_av,
-            ext_tile,
-            [=] (hc::tiled_index<1> idx)
-            __attribute__((hc))
-    {
-        int offset = amp_get_global_id(0);
-        // TODO-HCC - change to hc_get_local_size()
-        int stride = amp_get_local_size(0) * hc_get_num_groups(0) ;
-
-        for (int i=offset; i<sizeBytes; i+=stride) {
-            c[i] = a[i];
-        }
-    });
-
-    return cf;
-}
 
 #endif
diff --git a/src/hip_memory.cpp b/src/hip_memory.cpp
index 7146bf76a3..4ea3a64b18 100644
--- a/src/hip_memory.cpp
+++ b/src/hip_memory.cpp
@@ -587,6 +587,79 @@ hipError_t hipMemcpyToArray(hipArray* dst, size_t wOffset, size_t hOffset,
 }
 
 
+
+template <typename T>
+hc::completion_future
+ihipMemsetKernel(hipStream_t stream, T * ptr, T val, size_t sizeBytes)
+{
+    int wg = std::min((unsigned)8, stream->getDevice()->_compute_units);
+    const int threads_per_wg = 256;
+
+    int threads = wg * threads_per_wg;
+    if (threads > sizeBytes) {
+        threads = ((sizeBytes + threads_per_wg - 1) / threads_per_wg) * threads_per_wg;
+    }
+
+
+    hc::extent<1> ext(threads);
+    auto ext_tile = ext.tile(threads_per_wg);
+
+    hc::completion_future cf =
+    hc::parallel_for_each(
+            stream->_av,
+            ext_tile,
+            [=] (hc::tiled_index<1> idx)
+            __attribute__((hc))
+    {
+        int offset = amp_get_global_id(0);
+        // TODO-HCC - change to hc_get_local_size()
+        int stride = amp_get_local_size(0) * hc_get_num_groups(0) ;
+
+        for (int i=offset; i<sizeBytes; i+=stride) {
+            ptr[i] = val;
+        }
+    });
+
+    return cf;
+}
+
+template <typename T>
+hc::completion_future
+ihipMemcpyKernel(hipStream_t stream, T * c, const T * a, size_t sizeBytes)
+{
+    int wg = std::min((unsigned)8, stream->getDevice()->_compute_units);
+    const int threads_per_wg = 256;
+
+    int threads = wg * threads_per_wg;
+    if (threads > sizeBytes) {
+        threads = ((sizeBytes + threads_per_wg - 1) / threads_per_wg) * threads_per_wg;
+    }
+
+
+    hc::extent<1> ext(threads);
+    auto ext_tile = ext.tile(threads_per_wg);
+
+    hc::completion_future cf =
+    hc::parallel_for_each(
+            stream->_av,
+            ext_tile,
+            [=] (hc::tiled_index<1> idx)
+            __attribute__((hc))
+    {
+        int offset = amp_get_global_id(0);
+        // TODO-HCC - change to hc_get_local_size()
+        int stride = amp_get_local_size(0) * hc_get_num_groups(0) ;
+
+        for (int i=offset; i<sizeBytes; i+=stride) {
+            c[i] = a[i];
+        }
+    });
+
+    return cf;
+}
+
+
+
 // TODO-sync: function is async unless target is pinned host memory - then these are fully sync.
 /** @return #hipErrorInvalidValue
 */