From 24087ff3cc08c9ef050165566d25257067744e10 Mon Sep 17 00:00:00 2001
From: Jean-Luc Duprat <jld@acm.org>
Date: Tue, 27 Nov 2012 13:38:28 -0800
Subject: [PATCH] Expose none() in the ISPC standard library. On KNC: all(),
 any() and none() do not generate a redundant movmsk instruction.

---
 docs/ispc.rst             | 11 +++++++----
 examples/intrinsics/knc.h |  2 +-
 stdlib.ispc               | 23 +++++++++++++++++------
 3 files changed, 25 insertions(+), 11 deletions(-)

diff --git a/docs/ispc.rst b/docs/ispc.rst
index 609b7838..9412e649 100644
--- a/docs/ispc.rst
+++ b/docs/ispc.rst
@@ -3701,15 +3701,18 @@ where the ``i`` th element of ``x`` has been replaced with the value ``v``
 Reductions
 ----------
 
-A number routines are available to evaluate conditions across the running
-program instances.  For example, ``any()`` returns ``true`` if the given
-value ``v`` is ``true`` for any of the SPMD program instances currently
-running, and ``all()`` returns ``true`` if it true for all of them.
+A number routines are available to evaluate conditions across the
+running program instances.  For example, ``any()`` returns ``true`` if
+the given value ``v`` is ``true`` for any of the SPMD program
+instances currently running, ``all()`` returns ``true`` if it true
+for all of them, and ``none()`` returns ``true`` if ``v`` is always
+``false``.
 
 ::
 
     uniform bool any(bool v)
     uniform bool all(bool v)
+    uniform bool none(bool v)
 
 You can also compute a variety of reductions across the program instances.
 For example, the values of the given value in each of the active program
diff --git a/examples/intrinsics/knc.h b/examples/intrinsics/knc.h
index 08d2e8bf..dba551fe 100644
--- a/examples/intrinsics/knc.h
+++ b/examples/intrinsics/knc.h
@@ -482,7 +482,7 @@ static FORCEINLINE bool __all(__vec16_i1 mask) {
 }
 
 static FORCEINLINE bool __none(__vec16_i1 mask) {
-    return !__any(mask);
+    return _mm512_kortestz(mask, mask);
 }
 
 static FORCEINLINE __vec16_i1 __equal_i1(__vec16_i1 a, __vec16_i1 b) {
diff --git a/stdlib.ispc b/stdlib.ispc
index 80f44e91..81ebac70 100644
--- a/stdlib.ispc
+++ b/stdlib.ispc
@@ -340,9 +340,9 @@ static inline uniform bool any(bool v) {
     // We only care about whether "any" is true for the active program instances,
     // so we have to make v with the current program mask.
 #ifdef ISPC_TARGET_GENERIC
-    return __movmsk(v & __mask) != 0;
+    return __any(v | !__mask);
 #else
-    return __movmsk(__sext_varying_bool(v) & __mask) != 0;
+    return __any(__sext_varying_bool(v) | !__mask);
 #endif
 }
 
@@ -350,13 +350,24 @@ __declspec(safe)
 static inline uniform bool all(bool v) {
     // As with any(), we need to explicitly mask v with the current program mask
     // so we're only looking at the current lanes
+
 #ifdef ISPC_TARGET_GENERIC
-    bool match = ((v & __mask) == __mask);
+    return __all(v | !__mask);
 #else
-    int32 match = __sext_varying_bool((__sext_varying_bool(v) & __mask) == __mask);
+    return __all(__sext_varying_bool(v) | !__mask);
+#endif
+}
+
+__declspec(safe)
+static inline uniform bool none(bool v) {
+    // As with any(), we need to explicitly mask v with the current program mask
+    // so we're only looking at the current lanes
+
+#ifdef ISPC_TARGET_GENERIC
+    return __none(v | !__mask);
+#else
+    return __none(__sext_varying_bool(v) | !__mask);
 #endif
-    return __movmsk(match) == ((programCount == 64) ? ~0ull : 
-                               ((1ull << programCount) - 1));
 }
 
 __declspec(safe)