[libc-commits] [libc] [libc][mathvec] Initial commit for LIBC vector math component (PR #173058)

[Joseph Huber via libc-commits]

================
@@ -9,28 +9,28 @@
 #ifndef LLVM_LIBC_SRC___SUPPORT_MATHVEC_EXPF_H
 #define LLVM_LIBC_SRC___SUPPORT_MATHVEC_EXPF_H
 
+#include "expf_utils.h"
 #include "src/__support/CPP/simd.h"
 #include "src/__support/FPUtil/FPBits.h"
 #include "src/__support/common.h"
-#include "src/__support/mathvec/expf_utils.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
 namespace mathvec {
 
 template <size_t N>
-LIBC_INLINE static cpp::simd<double, N> inline_exp(cpp::simd<double, N> x) {
-  static constexpr cpp::simd<double, N> shift = 0x1.800000000ffc0p+46;
+LIBC_INLINE cpp::simd<double, N> inline_exp(cpp::simd<double, N> x) {
----------------
jhuber6 wrote:

Static on the function needs to stay, here and exp_lookup. Consider the case where we are computing a float, which turns into a double vector of size 4 for this lookup. A double vector of size 4 is an AVX register ABI on avx2 but two SSE registers in sse4_2. They both have the same mangled name, what happens when we start compiling this multiple times?

https://github.com/llvm/llvm-project/pull/173058