diff options
| author | Tue Ly <lntue@google.com> | 2022-04-07 16:02:10 -0400 |
|---|---|---|
| committer | Tue Ly <lntue@google.com> | 2022-04-08 14:12:24 -0400 |
| commit | c5f8a0a1e929ea41bd3e0d781c6b394a3f619427 (patch) | |
| tree | 9edde00f445c547d31f78d29be562c11d9834cc7 /libc/src/math/generic/logf.cpp | |
| parent | d49c0ba5abdeb8f6ebe081d96e0fdf44de46e7f8 (diff) | |
[libc] Add support for x86-64 targets that do not have FMA instructions.
Make FMA flag checks more accurate for x86-64 targets, and refactor
polyeval to use multiply and add instead when FMA instructions are not
available.
Reviewed By: michaelrj, sivachandra
Differential Revision: https://reviews.llvm.org/D123335
Diffstat (limited to 'libc/src/math/generic/logf.cpp')
| -rw-r--r-- | libc/src/math/generic/logf.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/libc/src/math/generic/logf.cpp b/libc/src/math/generic/logf.cpp index 3e712378b64c..747f8c73c27c 100644 --- a/libc/src/math/generic/logf.cpp +++ b/libc/src/math/generic/logf.cpp @@ -120,7 +120,7 @@ LLVM_LIBC_FUNCTION(float, logf, (float x)) { d *= ONE_OVER_F[f_index]; double extra_factor = - fputil::fma(static_cast<double>(m), LOG_2, LOG_F[f_index]); + fputil::multiply_add(static_cast<double>(m), LOG_2, LOG_F[f_index]); double r = __llvm_libc::fputil::polyeval( d, extra_factor, 0x1.fffffffffffacp-1, -0x1.fffffffef9cb2p-2, |