reorganize intrinsic includes to make clang happy

build.bat

@@ -75,7 +75,7 @@ set cl_link=       /link /MANIFEST:EMBED /INCREMENTAL:NO /pdbaltpath:%%%%_PDB%%%
 set cl_out=        /out:
 set cl_obj_out=    /Fo:
 set cl_linker=     
-set clang_common=  -mcx16 -msha -I..\src\ -I..\local\ -fdiagnostics-absolute-paths -Wall -Wno-unknown-warning-option -Wno-missing-braces -Wno-unused-function -Wno-unused-parameter -Wno-writable-strings -Wno-missing-field-initializers -Wno-unused-value -Wno-unused-variable -Wno-unused-local-typedef -Wno-deprecated-register -Wno-deprecated-declarations -Wno-unused-but-set-variable -Wno-single-bit-bitfield-constant-conversion -Wno-compare-distinct-pointer-types -Wno-initializer-overrides -Wno-incompatible-pointer-types-discards-qualifiers -Xclang -flto-visibility-public-std -D_USE_MATH_DEFINES -Dstrdup=_strdup -Dgnu_printf=printf -ferror-limit=10000 -msse2 -mavx512f -mavx512bw -mavx512dq -mavx512vbmi -mbmi2
+set clang_common=  -mcx16 -I..\src\ -I..\local\ -fdiagnostics-absolute-paths -Wall -Wno-unknown-warning-option -Wno-missing-braces -Wno-unused-function -Wno-unused-parameter -Wno-writable-strings -Wno-missing-field-initializers -Wno-unused-value -Wno-unused-variable -Wno-unused-local-typedef -Wno-deprecated-register -Wno-deprecated-declarations -Wno-unused-but-set-variable -Wno-single-bit-bitfield-constant-conversion -Wno-compare-distinct-pointer-types -Wno-initializer-overrides -Wno-incompatible-pointer-types-discards-qualifiers -Xclang -flto-visibility-public-std -D_USE_MATH_DEFINES -Dstrdup=_strdup -Dgnu_printf=printf -ferror-limit=10000
 set clang_debug=   call clang -g -O0 -DBUILD_DEBUG=1 -D_DEBUG %clang_common% %auto_compile_flags%
 set clang_release= call clang -g -O2 -DBUILD_DEBUG=0 -DNDEBUG %clang_common% %auto_compile_flags%
 set clang_link=    -fuse-ld=lld -Xlinker /MANIFEST:EMBED -Xlinker /pdbaltpath:%%%%_PDB%%%% -Xlinker /NATVIS:"%~dp0\src\natvis\base.natvis" -Xlinker /opt:ref -Xlinker /opt:noicf

src/base/base_context_cracking.h

@@ -248,4 +248,48 @@
 # error You tried to build with an unsupported architecture. Currently, only building in x64 mode is supported.
 #endif
 
+////////////////////////////////
+// extra intrinisc includes, so older clang versions are happy
+// when optiona intrinsics like avx2 & avx512 are used later
+// these must come first before any other intrinsic includes
+// otherwise compiler will define macros that prevent declarations
+// of intrinsics when included next time
+
+#if ARCH_X64 && COMPILER_CLANG
+#  if defined(__IMMINTRIN_H)
+#    error "include this header before immintrin.h / x86intrin.h / intrin.h"
+#  endif
+#  pragma clang diagnostic push
+#  pragma clang diagnostic ignored "-Wreserved-macro-identifier"
+#  pragma push_macro("__SHA__")
+#  pragma push_macro("__AVX__")
+#  pragma push_macro("__AVX2__")
+#  pragma push_macro("__BMI2__")
+#  pragma push_macro("__SSE4_1__")
+#  pragma push_macro("__AVX512F__")
+#  pragma push_macro("__AVX512VL__")
+#  pragma push_macro("__AVX512BW__")
+#  pragma push_macro("__AVX512VBMI__")
+#  define __SHA__ 1
+#  define __AVX__ 1
+#  define __AVX2__ 1
+#  define __BMI2__ 1
+#  define __SSE4_1__ 1
+#  define __AVX512F__ 1
+#  define __AVX512VL__ 1
+#  define __AVX512BW__ 1
+#  define __AVX512VBMI__ 1
+#  include <immintrin.h>
+#  pragma pop_macro("__AVX512VBMI__")
+#  pragma pop_macro("__AVX512BW__")
+#  pragma pop_macro("__AVX512VL__")
+#  pragma pop_macro("__AVX512F__")
+#  pragma pop_macro("__SSE4_1__")
+#  pragma pop_macro("__BMI2__")
+#  pragma pop_macro("__AVX2__")
+#  pragma pop_macro("__AVX__")
+#  pragma pop_macro("__SHA__")
+#  pragma clang diagnostic pop
+#endif
+
 #endif // BASE_CONTEXT_CRACKING_H

src/linker/base_ext/base_blake3.h

@@ -6,31 +6,6 @@
 #define BLAKE3_API static
 #define BLAKE3_PRIVATE static
 
-#if defined(__clang__) && defined(__x86_64__)
-#  if defined(__IMMINTRIN_H)
-#    error "include this header before immintrin.h / x86intrin.h / intrin.h"
-#  endif
-#  pragma clang diagnostic push
-#  pragma clang diagnostic ignored "-Wreserved-macro-identifier"
-#  pragma push_macro("__AVX__")
-#  pragma push_macro("__AVX2__")
-#  pragma push_macro("__SSE4_1__")
-#  pragma push_macro("__AVX512F__")
-#  pragma push_macro("__AVX512VL__")
-#  define __AVX__ 1
-#  define __AVX2__ 1
-#  define __SSE4_1__ 1
-#  define __AVX512F__ 1
-#  define __AVX512VL__ 1
-#  include <immintrin.h>
-#  pragma pop_macro("__AVX512VL__")
-#  pragma pop_macro("__AVX512F__")
-#  pragma pop_macro("__SSE4_1__")
-#  pragma pop_macro("__AVX2__")
-#  pragma pop_macro("__AVX__")
-#  pragma clang diagnostic pop
-#endif
-
 #include "third_party/blake3/c/blake3.h"
 
 static void

src/linker/lnk.c

@@ -9,11 +9,6 @@
 #define ARENA_FREE_LIST 1
 #define NO_ASYNC 1
 
-// --- Third Party -------------------------------------------------------------
-
-#include "base_ext/base_blake3.h"
-#include "base_ext/base_blake3.c"
-
 // --- Code Base ---------------------------------------------------------------
 
 #include "base/base_inc.h"
@@ -58,6 +53,11 @@
 #include "llvm/llvm.c"
 #include "dwarf/x64/dwarf_x64.c"
 
+// --- Third Party -------------------------------------------------------------
+
+#include "base_ext/base_blake3.h"
+#include "base_ext/base_blake3.c"
+
 // --- Code Base Extensions ----------------------------------------------------
 
 #include "base_ext/base_inc.h"

src/third_party/blake3/c/blake3_dispatch.c

@@ -55,7 +55,7 @@ static uint64_t xgetbv(void) {
 }
 
 static void cpuid(uint32_t out[4], uint32_t id) {
-#if defined(_MSC_VER)
+#if defined(_MSC_VER) && !defined(__clang__)
   __cpuid((int *)out, id);
 #elif defined(__i386__) || defined(_M_IX86)
   __asm__ __volatile__("movl %%ebx, %1\n"
@@ -71,7 +71,7 @@ static void cpuid(uint32_t out[4], uint32_t id) {
 }
 
 static void cpuidex(uint32_t out[4], uint32_t id, uint32_t sid) {
-#if defined(_MSC_VER)
+#if defined(_MSC_VER) && !defined(__clang__)
   __cpuidex((int *)out, id, sid);
 #elif defined(__i386__) || defined(_M_IX86)
   __asm__ __volatile__("movl %%ebx, %1\n"