[llvm] [TLI] Add missing ArmPL mappings (PR #78474)
Maciej Gabka via llvm-commits
llvm-commits at lists.llvm.org
Thu Jan 18 07:53:33 PST 2024
================
@@ -943,6 +953,16 @@ TLI_DEFINE_VECFUNC("hypotf", "armpl_vhypotq_f32", FIXED(4), NOMASK, "_ZGV_LLVM_N
TLI_DEFINE_VECFUNC("hypot", "armpl_svhypot_f64_x", SCALABLE(2), MASKED, "_ZGVsMxvv")
TLI_DEFINE_VECFUNC("hypotf", "armpl_svhypot_f32_x", SCALABLE(4), MASKED, "_ZGVsMxvv")
+TLI_DEFINE_VECFUNC("ilogb", "armpl_vilogbq_f64", FIXED(2), NOMASK, "_ZGV_LLVM_N2v")
+TLI_DEFINE_VECFUNC("ilogbf", "armpl_vilogbq_f32", FIXED(4), NOMASK, "_ZGV_LLVM_N4v")
+TLI_DEFINE_VECFUNC("ilogb", "armpl_svilogb_f64_x", SCALABLE(2), MASKED, "_ZGVsMxv")
+TLI_DEFINE_VECFUNC("ilogbf", "armpl_svilogb_f32_x", SCALABLE(4), MASKED, "_ZGVsMxv")
+
+TLI_DEFINE_VECFUNC("ldexp", "armpl_vldexpq_f64", FIXED(2), NOMASK, "_ZGV_LLVM_N2vv")
----------------
mgabka wrote:
why do you think so?
the function is defined as:
``
double ldexp( double arg, int exp)
``
both arguments are passed by value, so from the vector function I am expecting to return a vector of results where each element is multiplied by a different power of 2 (i.e different element of the input exp vector) .
https://github.com/llvm/llvm-project/pull/78474
More information about the llvm-commits
mailing list