[Libclc-dev] [PATCH 2/2] Add rhadd builtin

[Aaron Watry]

rhadd = (x+y+1)>>1

Implemented as:
(x>>1) + (y>>1) + ((x&1)|(y&1))

This prevents us having to do assembly addition and overflow detection
---
 generic/include/clc/clc.h             | 1 +
 generic/include/clc/integer/rhadd.h   | 2 ++
 generic/include/clc/integer/rhadd.inc | 1 +
 generic/lib/SOURCES                   | 1 +
 generic/lib/integer/rhadd.cl          | 4 ++++
 generic/lib/integer/rhadd.inc         | 6 ++++++
 6 files changed, 15 insertions(+)
 create mode 100644 generic/include/clc/integer/rhadd.h
 create mode 100644 generic/include/clc/integer/rhadd.inc
 create mode 100644 generic/lib/integer/rhadd.cl
 create mode 100644 generic/lib/integer/rhadd.inc

diff --git a/generic/include/clc/clc.h b/generic/include/clc/clc.h
index f9b2c38..305f058 100644
--- a/generic/include/clc/clc.h
+++ b/generic/include/clc/clc.h
@@ -69,6 +69,7 @@
 #include <clc/integer/mad24.h>
 #include <clc/integer/mul24.h>
 #include <clc/integer/mul_hi.h>
+#include <clc/integer/rhadd.h>
 #include <clc/integer/rotate.h>
 #include <clc/integer/sub_sat.h>
 #include <clc/integer/upsample.h>
diff --git a/generic/include/clc/integer/rhadd.h b/generic/include/clc/integer/rhadd.h
new file mode 100644
index 0000000..69b43fa
--- /dev/null
+++ b/generic/include/clc/integer/rhadd.h
@@ -0,0 +1,2 @@
+#define __CLC_BODY <clc/integer/rhadd.inc>
+#include <clc/integer/gentype.inc>
diff --git a/generic/include/clc/integer/rhadd.inc b/generic/include/clc/integer/rhadd.inc
new file mode 100644
index 0000000..88ccaf0
--- /dev/null
+++ b/generic/include/clc/integer/rhadd.inc
@@ -0,0 +1 @@
+_CLC_OVERLOAD _CLC_DECL __CLC_GENTYPE rhadd(__CLC_GENTYPE x, __CLC_GENTYPE y);
diff --git a/generic/lib/SOURCES b/generic/lib/SOURCES
index 0e66d71..d6e2d8c 100644
--- a/generic/lib/SOURCES
+++ b/generic/lib/SOURCES
@@ -15,6 +15,7 @@ integer/hadd.cl
 integer/mad24.cl
 integer/mul24.cl
 integer/mul_hi.cl
+integer/rhadd.cl
 integer/rotate.cl
 integer/sub_sat.cl
 integer/sub_sat_if.ll
diff --git a/generic/lib/integer/rhadd.cl b/generic/lib/integer/rhadd.cl
new file mode 100644
index 0000000..c985870
--- /dev/null
+++ b/generic/lib/integer/rhadd.cl
@@ -0,0 +1,4 @@
+#include <clc/clc.h>
+
+#define __CLC_BODY <rhadd.inc>
+#include <clc/integer/gentype.inc>
diff --git a/generic/lib/integer/rhadd.inc b/generic/lib/integer/rhadd.inc
new file mode 100644
index 0000000..3d60768
--- /dev/null
+++ b/generic/lib/integer/rhadd.inc
@@ -0,0 +1,6 @@
+//rhadd = (x+y+1)>>1
+//This can be simplified to x>>1 + y>>1 + (1 if either x or y have the 1s bit set)
+//This saves us having to do any checks for overflow in the addition sums
+_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE rhadd(__CLC_GENTYPE x, __CLC_GENTYPE y) {
+    return (x>>(__CLC_GENTYPE)1)+(y>>(__CLC_GENTYPE)1)+((x&(__CLC_GENTYPE)1)|(y&(__CLC_GENTYPE)1));
+}
-- 
1.8.1.2