Implement generic upsample()

Reduces all vector upsamples down to its scalar components, so probably
not the most efficient thing in the world, but it does what the
spec says it needs to do.

Another possible implementation would be to convert/cast everything as
unsigned if necessary, upsample the input vectors, create the upsampled
value, and then cast back to signed if required.

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Tom Stellard <thomas.stellard at amd.com>

git-svn-id: https://llvm.org/svn/llvm-project/libclc/trunk@186691 91177308-0d34-0410-b5e6-96231b3b80d8

4 files changed, 61 insertions, 0 deletions

diff --git a/generic/include/clc/clc.h b/generic/include/clc/clc.h
index dfdf747..9a2f443 100644
--- a/generic/include/clc/clc.h
+++ b/generic/include/clc/clc.h
@@ -68,6 +68,7 @@
 #include <clc/integer/mul24.h>
 #include <clc/integer/rotate.h>
 #include <clc/integer/sub_sat.h>
+#include <clc/integer/upsample.h>
 
 /* 6.11.2 and 6.11.3 Shared Integer/Math Functions */
 #include <clc/shared/clamp.h>
diff --git a/generic/include/clc/integer/upsample.h b/generic/include/clc/integer/upsample.h
new file mode 100644
index 0000000..127debf
--- /dev/null
+++ b/generic/include/clc/integer/upsample.h
@@ -0,0 +1,25 @@
+#define __CLC_UPSAMPLE_DECL(BGENTYPE, GENTYPE, UGENTYPE) \
+    _CLC_OVERLOAD _CLC_DECL BGENTYPE upsample(GENTYPE hi, UGENTYPE lo);
+
+#define __CLC_UPSAMPLE_VEC(BGENTYPE, GENTYPE, UGENTYPE) \
+    __CLC_UPSAMPLE_DECL(BGENTYPE, GENTYPE, UGENTYPE); \
+    __CLC_UPSAMPLE_DECL(BGENTYPE##2, GENTYPE##2, UGENTYPE##2); \
+    __CLC_UPSAMPLE_DECL(BGENTYPE##3, GENTYPE##3, UGENTYPE##3); \
+    __CLC_UPSAMPLE_DECL(BGENTYPE##4, GENTYPE##4, UGENTYPE##4); \
+    __CLC_UPSAMPLE_DECL(BGENTYPE##8, GENTYPE##8, UGENTYPE##8); \
+    __CLC_UPSAMPLE_DECL(BGENTYPE##16, GENTYPE##16, UGENTYPE##16); \
+
+#define __CLC_UPSAMPLE_TYPES() \
+    __CLC_UPSAMPLE_VEC(short, char, uchar) \
+    __CLC_UPSAMPLE_VEC(ushort, uchar, uchar) \
+    __CLC_UPSAMPLE_VEC(int, short, ushort) \
+    __CLC_UPSAMPLE_VEC(uint, ushort, ushort) \
+    __CLC_UPSAMPLE_VEC(long, int, uint) \
+    __CLC_UPSAMPLE_VEC(ulong, uint, uint) \
+
+__CLC_UPSAMPLE_TYPES()
+
+#undef __CLC_UPSAMPLE_TYPES
+#undef __CLC_UPSAMPLE_DECL
+#undef __CLC_UPSAMPLE_VEC
+
diff --git a/generic/lib/SOURCES b/generic/lib/SOURCES
index 21a7eaa..9ac08bd 100644
--- a/generic/lib/SOURCES
+++ b/generic/lib/SOURCES
@@ -17,6 +17,7 @@ integer/rotate.cl
 integer/sub_sat.cl
 integer/sub_sat_if.ll
 integer/sub_sat_impl.ll
+integer/upsample.cl
 math/fmax.cl
 math/fmin.cl
 math/hypot.cl
diff --git a/generic/lib/integer/upsample.cl b/generic/lib/integer/upsample.cl
new file mode 100644
index 0000000..7301cc3
--- /dev/null
+++ b/generic/lib/integer/upsample.cl
@@ -0,0 +1,34 @@
+#include <clc/clc.h>
+
+#define __CLC_UPSAMPLE_IMPL(BGENTYPE, GENTYPE, UGENTYPE, GENSIZE) \
+    _CLC_OVERLOAD _CLC_DECL BGENTYPE upsample(GENTYPE hi, UGENTYPE lo){ \
+        return ((BGENTYPE)hi << GENSIZE) | lo; \
+    } \
+    _CLC_OVERLOAD _CLC_DECL BGENTYPE##2 upsample(GENTYPE##2 hi, UGENTYPE##2 lo){ \
+        return (BGENTYPE##2){upsample(hi.s0, lo.s0), upsample(hi.s1, lo.s1)}; \
+    } \
+    _CLC_OVERLOAD _CLC_DECL BGENTYPE##3 upsample(GENTYPE##3 hi, UGENTYPE##3 lo){ \
+        return (BGENTYPE##3){upsample(hi.s0, lo.s0), upsample(hi.s1, lo.s1), upsample(hi.s2, lo.s2)}; \
+    } \
+    _CLC_OVERLOAD _CLC_DECL BGENTYPE##4 upsample(GENTYPE##4 hi, UGENTYPE##4 lo){ \
+        return (BGENTYPE##4){upsample(hi.lo, lo.lo), upsample(hi.hi, lo.hi)}; \
+    } \
+    _CLC_OVERLOAD _CLC_DECL BGENTYPE##8 upsample(GENTYPE##8 hi, UGENTYPE##8 lo){ \
+        return (BGENTYPE##8){upsample(hi.lo, lo.lo), upsample(hi.hi, lo.hi)}; \
+    } \
+    _CLC_OVERLOAD _CLC_DECL BGENTYPE##16 upsample(GENTYPE##16 hi, UGENTYPE##16 lo){ \
+        return (BGENTYPE##16){upsample(hi.lo, lo.lo), upsample(hi.hi, lo.hi)}; \
+    } \
+
+#define __CLC_UPSAMPLE_TYPES() \
+    __CLC_UPSAMPLE_IMPL(short, char, uchar, 8) \
+    __CLC_UPSAMPLE_IMPL(ushort, uchar, uchar, 8) \
+    __CLC_UPSAMPLE_IMPL(int, short, ushort, 16) \
+    __CLC_UPSAMPLE_IMPL(uint, ushort, ushort, 16) \
+    __CLC_UPSAMPLE_IMPL(long, int, uint, 32) \
+    __CLC_UPSAMPLE_IMPL(ulong, uint, uint, 32) \
+
+__CLC_UPSAMPLE_TYPES()
+
+#undef __CLC_UPSAMPLE_TYPES
+#undef __CLC_UPSAMPLE_IMPL