summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorJan Vesely <jan.vesely@rutgers.edu>2016-08-17 20:02:11 +0000
committerJan Vesely <jan.vesely@rutgers.edu>2016-08-17 20:02:11 +0000
commit1b4b5532e5d0911ed6581cb8b93a8c286931773a (patch)
tree164ab8b6dd7e49cfaa289ed9b8f69360762ba6b3
parent693817b954658d02ccaea50a9e5afc1f4504d5d0 (diff)
Implement vstore_half{,n}
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu> git-svn-id: https://llvm.org/svn/llvm-project/libclc/trunk@278962 91177308-0d34-0410-b5e6-96231b3b80d8
-rw-r--r--generic/include/clc/shared/vstore.h45
-rw-r--r--generic/lib/shared/vstore.cl32
-rw-r--r--generic/lib/shared/vstore_half.inc10
3 files changed, 68 insertions, 19 deletions
diff --git a/generic/include/clc/shared/vstore.h b/generic/include/clc/shared/vstore.h
index 1f784f8..ea8d4cf 100644
--- a/generic/include/clc/shared/vstore.h
+++ b/generic/include/clc/shared/vstore.h
@@ -1,17 +1,20 @@
-#define _CLC_VSTORE_DECL(PRIM_TYPE, VEC_TYPE, WIDTH, ADDR_SPACE) \
- _CLC_OVERLOAD _CLC_DECL void vstore##WIDTH(VEC_TYPE vec, size_t offset, ADDR_SPACE PRIM_TYPE *out);
+#define _CLC_VSTORE_DECL(SUFFIX, PRIM_TYPE, VEC_TYPE, WIDTH, ADDR_SPACE) \
+ _CLC_OVERLOAD _CLC_DECL void vstore##SUFFIX##WIDTH(VEC_TYPE vec, size_t offset, ADDR_SPACE PRIM_TYPE *out);
-#define _CLC_VECTOR_VSTORE_DECL(PRIM_TYPE, ADDR_SPACE) \
- _CLC_VSTORE_DECL(PRIM_TYPE, PRIM_TYPE##2, 2, ADDR_SPACE) \
- _CLC_VSTORE_DECL(PRIM_TYPE, PRIM_TYPE##3, 3, ADDR_SPACE) \
- _CLC_VSTORE_DECL(PRIM_TYPE, PRIM_TYPE##4, 4, ADDR_SPACE) \
- _CLC_VSTORE_DECL(PRIM_TYPE, PRIM_TYPE##8, 8, ADDR_SPACE) \
- _CLC_VSTORE_DECL(PRIM_TYPE, PRIM_TYPE##16, 16, ADDR_SPACE)
+#define _CLC_VECTOR_VSTORE_DECL(SUFFIX, MEM_TYPE, PRIM_TYPE, ADDR_SPACE) \
+ _CLC_VSTORE_DECL(SUFFIX, MEM_TYPE, PRIM_TYPE##2, 2, ADDR_SPACE) \
+ _CLC_VSTORE_DECL(SUFFIX, MEM_TYPE, PRIM_TYPE##3, 3, ADDR_SPACE) \
+ _CLC_VSTORE_DECL(SUFFIX, MEM_TYPE, PRIM_TYPE##4, 4, ADDR_SPACE) \
+ _CLC_VSTORE_DECL(SUFFIX, MEM_TYPE, PRIM_TYPE##8, 8, ADDR_SPACE) \
+ _CLC_VSTORE_DECL(SUFFIX, MEM_TYPE, PRIM_TYPE##16, 16, ADDR_SPACE)
+
+#define _CLC_VECTOR_VSTORE_PRIM3(SUFFIX, MEM_TYPE, PRIM_TYPE) \
+ _CLC_VECTOR_VSTORE_DECL(SUFFIX, MEM_TYPE, PRIM_TYPE, __private) \
+ _CLC_VECTOR_VSTORE_DECL(SUFFIX, MEM_TYPE, PRIM_TYPE, __local) \
+ _CLC_VECTOR_VSTORE_DECL(SUFFIX, MEM_TYPE, PRIM_TYPE, __global) \
#define _CLC_VECTOR_VSTORE_PRIM1(PRIM_TYPE) \
- _CLC_VECTOR_VSTORE_DECL(PRIM_TYPE, __private) \
- _CLC_VECTOR_VSTORE_DECL(PRIM_TYPE, __local) \
- _CLC_VECTOR_VSTORE_DECL(PRIM_TYPE, __global) \
+ _CLC_VECTOR_VSTORE_PRIM3(,PRIM_TYPE, PRIM_TYPE) \
#define _CLC_VECTOR_VSTORE_PRIM() \
_CLC_VECTOR_VSTORE_PRIM1(char) \
@@ -23,14 +26,18 @@
_CLC_VECTOR_VSTORE_PRIM1(long) \
_CLC_VECTOR_VSTORE_PRIM1(ulong) \
_CLC_VECTOR_VSTORE_PRIM1(float) \
-
+ _CLC_VECTOR_VSTORE_PRIM3(_half, half, float)
+
#ifdef cl_khr_fp64
-#define _CLC_VECTOR_VSTORE() \
- _CLC_VECTOR_VSTORE_PRIM1(double) \
- _CLC_VECTOR_VSTORE_PRIM()
-#else
-#define _CLC_VECTOR_VSTORE() \
- _CLC_VECTOR_VSTORE_PRIM()
+#pragma cl_khr_fp64: enable
+ _CLC_VECTOR_VSTORE_PRIM1(double)
+ _CLC_VECTOR_VSTORE_PRIM3(_half, half, double)
+ _CLC_VSTORE_DECL(_half, half, double, , __private)
+ _CLC_VSTORE_DECL(_half, half, double, , __local)
+ _CLC_VSTORE_DECL(_half, half, double, , __global)
#endif
-_CLC_VECTOR_VSTORE()
+_CLC_VECTOR_VSTORE_PRIM()
+_CLC_VSTORE_DECL(_half, half, float, , __private)
+_CLC_VSTORE_DECL(_half, half, float, , __local)
+_CLC_VSTORE_DECL(_half, half, float, , __global)
diff --git a/generic/lib/shared/vstore.cl b/generic/lib/shared/vstore.cl
index 4777b7e..ebc9446 100644
--- a/generic/lib/shared/vstore.cl
+++ b/generic/lib/shared/vstore.cl
@@ -50,3 +50,35 @@ VSTORE_TYPES()
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
VSTORE_ADDR_SPACES(double)
#endif
+
+/* vstore_half are legal even without cl_khr_fp16 */
+
+#define VEC_STORE1(val) mem[offset++] = val;
+#define VEC_STORE2(val) \
+ VEC_STORE1(val.lo) \
+ VEC_STORE1(val.hi)
+#define VEC_STORE3(val) \
+ VEC_STORE1(val.s0) \
+ VEC_STORE1(val.s1) \
+ VEC_STORE1(val.s2)
+#define VEC_STORE4(val) \
+ VEC_STORE2(val.lo) \
+ VEC_STORE2(val.hi)
+#define VEC_STORE8(val) \
+ VEC_STORE4(val.lo) \
+ VEC_STORE4(val.hi)
+#define VEC_STORE16(val) \
+ VEC_STORE8(val.lo) \
+ VEC_STORE8(val.hi)
+
+#define __FUNC(SUFFIX, VEC_SIZE, TYPE, AS) \
+ _CLC_OVERLOAD _CLC_DEF void vstore_half##SUFFIX(TYPE vec, size_t offset, AS half *mem) { \
+ offset *= VEC_SIZE; \
+ VEC_STORE##VEC_SIZE(vec) \
+ }
+
+#define FUNC(SUFFIX, VEC_SIZE, TYPE, AS) __FUNC(SUFFIX, VEC_SIZE, TYPE, AS)
+
+#define __CLC_BODY "vstore_half.inc"
+#include <clc/math/gentype.inc>
+
diff --git a/generic/lib/shared/vstore_half.inc b/generic/lib/shared/vstore_half.inc
new file mode 100644
index 0000000..8ed03a0
--- /dev/null
+++ b/generic/lib/shared/vstore_half.inc
@@ -0,0 +1,10 @@
+
+#ifdef __CLC_VECSIZE
+ FUNC(__CLC_VECSIZE, __CLC_VECSIZE, __CLC_GENTYPE, __private);
+ FUNC(__CLC_VECSIZE, __CLC_VECSIZE, __CLC_GENTYPE, __local);
+ FUNC(__CLC_VECSIZE, __CLC_VECSIZE, __CLC_GENTYPE, __global);
+#else
+ FUNC(, 1, __CLC_GENTYPE, __private);
+ FUNC(, 1, __CLC_GENTYPE, __local);
+ FUNC(, 1, __CLC_GENTYPE, __global);
+#endif