add compress&expand in avx512vbmi2

Author: junparser

include/xsimd/arch/xsimd_avx512vbmi2.hpp

@@ -17,4 +17,57 @@
 
 #include "../types/xsimd_avx512vbmi2_register.hpp"
 
+namespace xsimd
+{
+
+    namespace kernel
+    {
+        using namespace types;
+
+        // compress
+        template <class A>
+        XSIMD_INLINE batch<int16_t, A> compress(batch<int16_t, A> const& self, batch_bool<int16_t, A> const& mask, requires_arch<avx512vbmi2>) noexcept
+        {
+            return _mm512_maskz_compress_epi16(mask.mask(), self);
+        }
+        template <class A>
+        XSIMD_INLINE batch<uint16_t, A> compress(batch<uint16_t, A> const& self, batch_bool<uint16_t, A> const& mask, requires_arch<avx512vbmi2>) noexcept
+        {
+            return _mm512_maskz_compress_epi16(mask.mask(), self);
+        }
+        template <class A>
+        XSIMD_INLINE batch<int8_t, A> compress(batch<int8_t, A> const& self, batch_bool<int8_t, A> const& mask, requires_arch<avx512vbmi2>) noexcept
+        {
+            return _mm512_maskz_compress_epi8(mask.mask(), self);
+        }
+        template <class A>
+        XSIMD_INLINE batch<uint8_t, A> compress(batch<uint8_t, A> const& self, batch_bool<uint8_t, A> const& mask, requires_arch<avx512vbmi2>) noexcept
+        {
+            return _mm512_maskz_compress_epi8(mask.mask(), self);
+        }
+
+        // expand
+        template <class A>
+        XSIMD_INLINE batch<int16_t, A> expand(batch<int16_t, A> const& self, batch_bool<int16_t, A> const& mask, requires_arch<avx512vbmi2>) noexcept
+        {
+            return _mm512_maskz_expand_epi16(mask.mask(), self);
+        }
+        template <class A>
+        XSIMD_INLINE batch<uint16_t, A> expand(batch<uint16_t, A> const& self, batch_bool<uint16_t, A> const& mask, requires_arch<avx512vbmi2>) noexcept
+        {
+            return _mm512_maskz_expand_epi16(mask.mask(), self);
+        }
+        template <class A>
+        XSIMD_INLINE batch<int8_t, A> expand(batch<int8_t, A> const& self, batch_bool<int8_t, A> const& mask, requires_arch<avx512vbmi2>) noexcept
+        {
+            return _mm512_maskz_expand_epi8(mask.mask(), self);
+        }
+        template <class A>
+        XSIMD_INLINE batch<uint8_t, A> expand(batch<uint8_t, A> const& self, batch_bool<uint8_t, A> const& mask, requires_arch<avx512vbmi2>) noexcept
+        {
+            return _mm512_maskz_expand_epi8(mask.mask(), self);
+        }
+    }
+}
+
 #endif

test/test_shuffle.cpp

@@ -347,7 +347,7 @@ struct compress_test
     }
 };
 
-TEST_CASE_TEMPLATE("[compress]", B, BATCH_FLOAT_TYPES, xsimd::batch<uint32_t>, xsimd::batch<int32_t>, xsimd::batch<uint64_t>, xsimd::batch<int64_t>)
+TEST_CASE_TEMPLATE("[compress]", B, BATCH_FLOAT_TYPES, xsimd::batch<uint8_t>, xsimd::batch<int8_t>, xsimd::batch<uint16_t>, xsimd::batch<int16_t>, xsimd::batch<uint32_t>, xsimd::batch<int32_t>, xsimd::batch<uint64_t>, xsimd::batch<int64_t>)
 {
     compress_test<B> Test;
     SUBCASE("empty")
@@ -443,7 +443,7 @@ struct expand_test
     }
 };
 
-TEST_CASE_TEMPLATE("[expand]", B, BATCH_FLOAT_TYPES, xsimd::batch<uint32_t>, xsimd::batch<int32_t>, xsimd::batch<uint64_t>, xsimd::batch<int64_t>)
+TEST_CASE_TEMPLATE("[expand]", B, BATCH_FLOAT_TYPES, xsimd::batch<uint8_t>, xsimd::batch<int8_t>, xsimd::batch<uint16_t>, xsimd::batch<int16_t>, xsimd::batch<uint32_t>, xsimd::batch<int32_t>, xsimd::batch<uint64_t>, xsimd::batch<int64_t>)
 {
     expand_test<B> Test;
     SUBCASE("empty")