Added NEON test any/all support.

include/bx/float4_neon.h

@@ -32,6 +32,28 @@ namespace bx
 #undef ELEMy
 #undef ELEMx
 
+#define IMPLEMENT_TEST(_xyzw, _swizzle) \
+			BX_FLOAT4_INLINE bool float4_test_any_##_xyzw(float4_t _test); \
+			BX_FLOAT4_INLINE bool float4_test_all_##_xyzw(float4_t _test);
+
+IMPLEMENT_TEST(x    , xxxx);
+IMPLEMENT_TEST(y    , yyyy);
+IMPLEMENT_TEST(xy   , xyyy);
+IMPLEMENT_TEST(z    , zzzz);
+IMPLEMENT_TEST(xz   , xzzz);
+IMPLEMENT_TEST(yz   , yzzz);
+IMPLEMENT_TEST(xyz  , xyzz);
+IMPLEMENT_TEST(w    , wwww);
+IMPLEMENT_TEST(xw   , xwww);
+IMPLEMENT_TEST(yw   , ywww);
+IMPLEMENT_TEST(xyw  , xyww);
+IMPLEMENT_TEST(zw   , zwww);
+IMPLEMENT_TEST(xzw  , xzww);
+IMPLEMENT_TEST(yzw  , yzww);
+IMPLEMENT_TEST(xyzw , xyzw);
+
+#undef IMPLEMENT_TEST
+
 	BX_FLOAT4_INLINE float4_t float4_shuf_xyAB(float4_t _a, float4_t _b)
 	{
 		return __builtin_shuffle(_a, _b, (_u32x4_t){ 0, 1, 4, 5 });
@@ -457,4 +479,47 @@ namespace bx
 
 #include "float4_ni.h"
 
+namespace bx
+{
+#define IMPLEMENT_TEST(_xyzw, _swizzle) \
+			BX_FLOAT4_INLINE bool float4_test_any_##_xyzw(float4_t _test) \
+			{ \
+				const float4_t tmp0 = float4_swiz_##_swizzle(_test); \
+				return float4_test_any_ni(tmp0); \
+			} \
+			\
+			BX_FLOAT4_INLINE bool float4_test_all_##_xyzw(float4_t _test) \
+			{ \
+				const float4_t tmp0 = float4_swiz_##_swizzle(_test); \
+				return float4_test_all_ni(tmp0); \
+			}
+
+IMPLEMENT_TEST(x    , xxxx);
+IMPLEMENT_TEST(y    , yyyy);
+IMPLEMENT_TEST(xy   , xyyy);
+IMPLEMENT_TEST(z    , zzzz);
+IMPLEMENT_TEST(xz   , xzzz);
+IMPLEMENT_TEST(yz   , yzzz);
+IMPLEMENT_TEST(xyz  , xyzz);
+IMPLEMENT_TEST(w    , wwww);
+IMPLEMENT_TEST(xw   , xwww);
+IMPLEMENT_TEST(yw   , ywww);
+IMPLEMENT_TEST(xyw  , xyww);
+IMPLEMENT_TEST(zw   , zwww);
+IMPLEMENT_TEST(xzw  , xzww);
+IMPLEMENT_TEST(yzw  , yzww);
+
+	BX_FLOAT4_INLINE bool float4_test_any_xyzw(float4_t _test)
+	{
+		return float4_test_any_ni(_test);
+	}
+
+	BX_FLOAT4_INLINE bool float4_test_all_xyzw(float4_t _test)
+	{
+		return float4_test_all_ni(_test);
+	}
+
+#undef IMPLEMENT_TEST
+} // namespace bx
+
 #endif // BX_FLOAT4_NEON_H_HEADER_GUARD

include/bx/float4_ni.h

@@ -437,6 +437,32 @@ namespace bx
 		return result;
 	}
 
+	BX_FLOAT4_INLINE bool float4_test_any_ni(float4_t _a)
+	{
+		const float4_t mask   = float4_sra(_a, 31);
+		const float4_t zwxy   = float4_swiz_zwxy(mask);
+		const float4_t tmp0   = float4_or(mask, zwxy);
+		const float4_t tmp1   = float4_swiz_yyyy(tmp0);
+		const float4_t tmp2   = float4_or(tmp0, tmp1);
+		int res;
+		float4_stx(&res, tmp2);
+		return 0 != res;
+	}
+
+	BX_FLOAT4_INLINE bool float4_test_all_ni(float4_t _a)
+	{
+		const float4_t bits   = float4_sra(_a, 31);
+		const float4_t m1248  = float4_ild(1, 2, 4, 8);
+		const float4_t mask   = float4_and(bits, m1248);
+		const float4_t zwxy   = float4_swiz_zwxy(mask);
+		const float4_t tmp0   = float4_or(mask, zwxy);
+		const float4_t tmp1   = float4_swiz_yyyy(tmp0);
+		const float4_t tmp2   = float4_or(tmp0, tmp1);
+		int res;
+		float4_stx(&res, tmp2);
+		return 0xf == res;
+	}
+
 } // namespace bx
 
 #endif // BX_FLOAT4_NI_H_HEADER_GUARD

tests/float4_t.cpp

@@ -18,6 +18,17 @@ union float4_cast
 	char c[16];
 };
 
+void float4_check_bool(const char* _str, bool _a, bool _0)
+{
+	DBG("%s %d == %d"
+		, _str
+		, _a
+		, _0
+		);
+
+	CHECK_EQUAL(_a, _0);
+}
+
 void float4_check_int32(const char* _str, bx::float4_t _a, int32_t _0, int32_t _1, int32_t _2, int32_t _3)
 {
 	float4_cast c; c.f4 = _a;
@@ -151,6 +162,21 @@ TEST(float4_compare)
 		, float4_icmpgt(float4_ild(0, 1, 2, 3), float4_ild(0, -2, 1, 3) )
 		, 0, -1, -1, 0
 		);
+
+	float4_check_bool("test_any_xyzw"
+		, float4_test_any_xyzw(float4_ild(-1, 0, 0, 0) )
+		, true
+		);
+
+	float4_check_bool("test_all_xw"
+		, float4_test_all_xw(float4_ild(-1, 0, 0, -1) )
+		, true
+		);
+
+	float4_check_bool("test_all_xzw"
+		, float4_test_all_xzw(float4_ild(-1, 0, 0, -1) )
+		, false
+		);
 }
 
 TEST(float4_load)