avx512 gather: add MSVC native fallbacks

fix casts for GCC<9

x86/avx2.h

@@ -2269,11 +2269,11 @@ simde_mm256_i32gather_ps(const simde_float32* base_addr, simde__m256i vindex, co
   return simde__m256_from_private(r_);
 }
 #if defined(SIMDE_X86_AVX2_NATIVE)
-  #define simde_mm256_i32gather_ps(base_addr, vindex, scale) _mm256_i32gather_ps(SIMDE_CHECKED_REINTERPRET_CAST(float const*, simde_float32 const*, base_addr), vindex, scale)
+  #define simde_mm256_i32gather_ps(base_addr, vindex, scale) _mm256_i32gather_ps(SIMDE_CHECKED_REINTERPRET_CAST(float const*, simde_float32 const*, (base_addr)), (vindex), (scale))
 #endif
 #if defined(SIMDE_X86_AVX2_ENABLE_NATIVE_ALIASES)
   #undef _mm256_i32gather_ps
-  #define _mm256_i32gather_ps(base_addr, vindex, scale) simde_mm256_i32gather_ps(SIMDE_CHECKED_REINTERPRET_CAST(simde_float32 const*, float const*, base_addr), vindex, scale)
+  #define _mm256_i32gather_ps(base_addr, vindex, scale) simde_mm256_i32gather_ps(SIMDE_CHECKED_REINTERPRET_CAST(simde_float32 const*, float const*, (base_addr)), (vindex), (scale))
 #endif
 
 SIMDE_FUNCTION_ATTRIBUTES

x86/avx512/gather.h

@@ -66,6 +66,13 @@ simde_mm512_i32gather_ps(simde__m512i vindex, const void* base_addr, const int32
       HEDLEY_STATIC_CAST(float const*, (base_addr)), simde_mm512_i32gather_ps_vindex_.m256i[1], (scale)); \
     simde__m512_from_private(simde_mm512_i32gather_ps_r_); \
   }))
+#elif defined(SIMDE_X86_AVX2_NATIVE) && !defined(SIMDE_STATEMENT_EXPR_)
+  #define simde_mm512_i32gather_ps(vindex, base_addr, scale) \
+    simde_x_mm512_set_m256( \
+      _mm256_i32gather_ps(HEDLEY_STATIC_CAST(float const*, (base_addr)), \
+        simde_mm512_extracti32x8_epi32((vindex), 1), (scale)), \
+      _mm256_i32gather_ps(HEDLEY_STATIC_CAST(float const*, (base_addr)), \
+        simde_mm512_extracti32x8_epi32((vindex), 0), (scale)) )
 #endif
 #if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
   #undef _mm512_i32gather_ps
@@ -105,6 +112,15 @@ simde_mm512_i64gather_epi32(simde__m512i vindex, const void* base_addr, const in
       HEDLEY_STATIC_CAST(int const*, (base_addr)), simde_mm512_i64gather_epi32_vindex_.m256i[1], (scale)); \
     simde__m256i_from_private(simde_mm512_i64gather_epi32_r_); \
   }))
+#elif defined(SIMDE_X86_AVX2_NATIVE) && !defined(SIMDE_STATEMENT_EXPR_)
+  #define simde_mm512_i64gather_epi32(vindex, base_addr, scale) \
+    _mm256_insertf128_si256( \
+      _mm256_castsi128_si256( \
+        _mm256_i64gather_epi32(HEDLEY_STATIC_CAST(int const*, (base_addr)), \
+          simde_mm512_extracti64x4_epi64((vindex), 0), (scale))), \
+      _mm256_i64gather_epi32(HEDLEY_STATIC_CAST(int const*, (base_addr)), \
+        simde_mm512_extracti64x4_epi64((vindex), 1), (scale)), \
+      1)
 #endif
 #if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
   #undef _mm512_i64gather_epi32
@@ -148,11 +164,18 @@ simde_mm512_i64gather_epi64(simde__m512i vindex, const void* base_addr, const in
     simde__m512i_private simde_mm512_i64gather_epi64_r_, \
       simde_mm512_i64gather_epi64_vindex_ = simde__m512i_to_private((vindex)); \
     simde_mm512_i64gather_epi64_r_.m256i[0] = _mm256_i64gather_epi64( \
-      HEDLEY_STATIC_CAST(int64_t const*, (base_addr)), simde_mm512_i64gather_epi64_vindex_.m256i[0], (scale)); \
+      HEDLEY_STATIC_CAST(long long const*, (base_addr)), simde_mm512_i64gather_epi64_vindex_.m256i[0], (scale)); \
     simde_mm512_i64gather_epi64_r_.m256i[1] = _mm256_i64gather_epi64( \
-      HEDLEY_STATIC_CAST(int64_t const*, (base_addr)), simde_mm512_i64gather_epi64_vindex_.m256i[1], (scale)); \
+      HEDLEY_STATIC_CAST(long long const*, (base_addr)), simde_mm512_i64gather_epi64_vindex_.m256i[1], (scale)); \
     simde__m512i_from_private(simde_mm512_i64gather_epi64_r_); \
   }))
+#elif defined(SIMDE_X86_AVX2_NATIVE) && !defined(SIMDE_STATEMENT_EXPR_)
+  #define simde_mm512_i64gather_epi64(vindex, base_addr, scale) \
+    simde_x_mm512_set_m256i( \
+      _mm256_i64gather_epi64(HEDLEY_STATIC_CAST(long long const*, (base_addr)), \
+        simde_mm512_extracti32x8_epi32((vindex), 1), (scale)), \
+      _mm256_i64gather_epi64(HEDLEY_STATIC_CAST(long long const*, (base_addr)), \
+        simde_mm512_extracti32x8_epi32((vindex), 0), (scale)) )
 #endif
 #if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
   #undef _mm512_i64gather_epi64
@@ -202,6 +225,13 @@ simde_mm512_i64gather_pd(simde__m512i vindex, const void* base_addr, const int32
       HEDLEY_STATIC_CAST(double const*, (base_addr)), simde_mm512_i64gather_pd_vindex_.m256i[1], (scale)); \
     simde__m512d_from_private(simde_mm512_i64gather_pd_r_); \
   }))
+#elif defined(SIMDE_X86_AVX2_NATIVE) && !defined(SIMDE_STATEMENT_EXPR_)
+  #define simde_mm512_i64gather_pd(vindex, base_addr, scale) \
+    simde_x_mm512_set_m256d( \
+      _mm256_i64gather_pd(HEDLEY_STATIC_CAST(double const*, (base_addr)), \
+        simde_mm512_extracti64x4_epi64((vindex), 1), (scale)), \
+      _mm256_i64gather_pd(HEDLEY_STATIC_CAST(double const*, (base_addr)), \
+        simde_mm512_extracti64x4_epi64((vindex), 0), (scale)) )
 #endif
 #if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
   #undef _mm512_i64gather_pd
@@ -251,6 +281,15 @@ simde_mm512_i64gather_ps(simde__m512i vindex, const void* base_addr, const int32
       HEDLEY_STATIC_CAST(float const*, (base_addr)), simde_mm512_i64gather_ps_vindex_.m256i[1], (scale)); \
     simde__m256_from_private(simde_mm512_i64gather_ps_r_); \
   }))
+#elif defined(SIMDE_X86_AVX2_NATIVE) && !defined(SIMDE_STATEMENT_EXPR_)
+  #define simde_mm512_i64gather_ps(vindex, base_addr, scale) \
+    _mm256_insertf128_ps( \
+      _mm256_castps128_ps256( \
+        _mm256_i64gather_ps(HEDLEY_STATIC_CAST(float const*, (base_addr)), \
+          simde_mm512_extracti64x4_epi64((vindex), 0), (scale))), \
+      _mm256_i64gather_ps(HEDLEY_STATIC_CAST(float const*, (base_addr)), \
+        simde_mm512_extracti64x4_epi64((vindex), 1), (scale)), \
+      1)
 #endif
 #if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
   #undef _mm512_i64gather_ps