[Libclc-dev] [PATCH 3/3] Implement async_work_group_strided_copy builtin

[Tom Stellard]

This is a simple implementation which just copies data synchronously.
---
 .../clc/async/async_work_group_strided_copy.h      | 15 ++++++++++
 .../clc/async/async_work_group_strided_copy.inc    |  6 ++++
 generic/include/clc/clc.h                          |  1 +
 generic/lib/SOURCES                                |  1 +
 generic/lib/async/async_work_group_strided_copy.cl |  9 ++++++
 .../lib/async/async_work_group_strided_copy.inc    | 34 ++++++++++++++++++++++
 6 files changed, 66 insertions(+)
 create mode 100644 generic/include/clc/async/async_work_group_strided_copy.h
 create mode 100644 generic/include/clc/async/async_work_group_strided_copy.inc
 create mode 100644 generic/lib/async/async_work_group_strided_copy.cl
 create mode 100644 generic/lib/async/async_work_group_strided_copy.inc

diff --git a/generic/include/clc/async/async_work_group_strided_copy.h b/generic/include/clc/async/async_work_group_strided_copy.h
new file mode 100644
index 0000000..bfa6f31
--- /dev/null
+++ b/generic/include/clc/async/async_work_group_strided_copy.h
@@ -0,0 +1,15 @@
+#define __CLC_DST_ADDR_SPACE local
+#define __CLC_SRC_ADDR_SPACE global
+#define __CLC_BODY <clc/async/async_work_group_strided_copy.inc>
+#include <clc/async/gentype.inc>
+#undef __CLC_DST_ADDR_SPACE
+#undef __CLC_SRC_ADDR_SPACE
+#undef __CLC_BODY
+
+#define __CLC_DST_ADDR_SPACE global
+#define __CLC_SRC_ADDR_SPACE local
+#define __CLC_BODY <clc/async/async_work_group_strided_copy.inc>
+#include <clc/async/gentype.inc>
+#undef __CLC_DST_ADDR_SPACE
+#undef __CLC_SRC_ADDR_SPACE
+#undef __CLC_BODY
diff --git a/generic/include/clc/async/async_work_group_strided_copy.inc b/generic/include/clc/async/async_work_group_strided_copy.inc
new file mode 100644
index 0000000..bdbea3a
--- /dev/null
+++ b/generic/include/clc/async/async_work_group_strided_copy.inc
@@ -0,0 +1,6 @@
+_CLC_OVERLOAD _CLC_DECL event_t async_work_group_strided_copy(
+  __CLC_DST_ADDR_SPACE __CLC_GENTYPE *dst,
+  const __CLC_SRC_ADDR_SPACE __CLC_GENTYPE *src,
+  size_t num_gentypes,
+  size_t stride,
+  event_t event);
diff --git a/generic/include/clc/clc.h b/generic/include/clc/clc.h
index ed741b1..ecb761a 100644
--- a/generic/include/clc/clc.h
+++ b/generic/include/clc/clc.h
@@ -126,6 +126,7 @@
 
 /* 6.11.10 Async Copy and Prefetch Functions */
 #include <clc/async/async_work_group_copy.h>
+#include <clc/async/async_work_group_strided_copy.h>
 #include <clc/async/prefetch.h>
 #include <clc/async/wait_group_events.h>
 
diff --git a/generic/lib/SOURCES b/generic/lib/SOURCES
index e7dbca5..6cfcffe 100644
--- a/generic/lib/SOURCES
+++ b/generic/lib/SOURCES
@@ -1,4 +1,5 @@
 async/async_work_group_copy.cl
+async/async_work_group_strided_copy.cl
 async/prefetch.cl
 async/wait_group_events.cl
 atomic/atomic_impl.ll
diff --git a/generic/lib/async/async_work_group_strided_copy.cl b/generic/lib/async/async_work_group_strided_copy.cl
new file mode 100644
index 0000000..61b8898
--- /dev/null
+++ b/generic/lib/async/async_work_group_strided_copy.cl
@@ -0,0 +1,9 @@
+#include <clc/clc.h>
+
+#ifdef cl_khr_fp64
+#pragma OPENCL EXTENSION cl_khr_fp64 : enable
+#endif
+
+#define __CLC_BODY <async_work_group_strided_copy.inc>
+#include <clc/async/gentype.inc>
+#undef __CLC_BODY
diff --git a/generic/lib/async/async_work_group_strided_copy.inc b/generic/lib/async/async_work_group_strided_copy.inc
new file mode 100644
index 0000000..30efa3c
--- /dev/null
+++ b/generic/lib/async/async_work_group_strided_copy.inc
@@ -0,0 +1,34 @@
+
+#define STRIDED_COPY(dst, src, num_gentypes, dst_stride, src_stride)       \
+  size_t size = get_local_size(0) * get_local_size(1) * get_local_size(2); \
+  size_t id = (get_local_size(1) * get_local_size(2) * get_local_id(0)) +  \
+              (get_local_size(2) * get_local_id(1)) +                      \
+              get_local_id(2);                                             \
+  unsigned i;                                                              \
+                                                                           \
+  for (i = id; i < num_gentypes; i += size) {                              \
+    dst[i * dst_stride] = src[i * src_stride];                             \
+  }
+
+
+_CLC_OVERLOAD _CLC_DEF event_t async_work_group_strided_copy(
+    local __CLC_GENTYPE *dst,
+    const global __CLC_GENTYPE *src,
+    size_t num_gentypes,
+    size_t src_stride,
+    event_t event) {
+
+  STRIDED_COPY(dst, src, num_gentypes, 1, src_stride);
+  return event;
+}
+
+_CLC_OVERLOAD _CLC_DEF event_t async_work_group_strided_copy(
+    global __CLC_GENTYPE *dst,
+    const local __CLC_GENTYPE *src,
+    size_t num_gentypes,
+    size_t dst_stride,
+    event_t event) {
+
+  STRIDED_COPY(dst, src, num_gentypes, dst_stride, 1);
+  return event;
+}
-- 
1.8.1.5