Incorrect CPUID feature name in the Intrinsics Guide: FP16C instead of F16C #738
Comments
Centril
added a commit
to Centril/rust
that referenced
this issue
Apr 23, 2019
Add f16c target_feature This is requires for Intel 16-bit half-precision float intrinsics: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=fp16&expand=1769 - see companion stdsimd PR: rust-lang/stdarch#737. LLVM, Wikipedia CPUID page, and the Intel Dev Manual all call this CPUID feature "F16C", but the Intel intrinsics guide calls this "FP16C" - this is probably a bug in the intrinsics guide which we are tracking here: rust-lang/stdarch#738 r? @alexcrichton
Centril
added a commit
to Centril/rust
that referenced
this issue
Apr 23, 2019
Add f16c target_feature This is requires for Intel 16-bit half-precision float intrinsics: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=fp16&expand=1769 - see companion stdsimd PR: rust-lang/stdarch#737. LLVM, Wikipedia CPUID page, and the Intel Dev Manual all call this CPUID feature "F16C", but the Intel intrinsics guide calls this "FP16C" - this is probably a bug in the intrinsics guide which we are tracking here: rust-lang/stdarch#738 r? @alexcrichton
Centril
added a commit
to Centril/rust
that referenced
this issue
Apr 23, 2019
Add f16c target_feature This is requires for Intel 16-bit half-precision float intrinsics: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=fp16&expand=1769 - see companion stdsimd PR: rust-lang/stdarch#737. LLVM, Wikipedia CPUID page, and the Intel Dev Manual all call this CPUID feature "F16C", but the Intel intrinsics guide calls this "FP16C" - this is probably a bug in the intrinsics guide which we are tracking here: rust-lang/stdarch#738 r? @alexcrichton
Centril
added a commit
to Centril/rust
that referenced
this issue
Apr 23, 2019
Add f16c target_feature This is requires for Intel 16-bit half-precision float intrinsics: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=fp16&expand=1769 - see companion stdsimd PR: rust-lang/stdarch#737. LLVM, Wikipedia CPUID page, and the Intel Dev Manual all call this CPUID feature "F16C", but the Intel intrinsics guide calls this "FP16C" - this is probably a bug in the intrinsics guide which we are tracking here: rust-lang/stdarch#738 r? @alexcrichton
|
Intel has fixed this bug (this is not in version 3.6.8 anymore) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The Intel Intrinsics Guide for the 16-bit half-precision float intrinsics calls the CPUID feature flag "FP16C", see: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#techs=Other&text=fp16&expand=1711
LLVM (e.g. ), CPUID Wikipedia page (https://en.wikipedia.org/wiki/CPUID), and thIntel® 64 and IA-32 Architectures Software Developer’s Manual Volume 2 (2A, 2B, 2C & 2D): Instruction Set Reference, A-Z (https://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf) all call this feature "F16C" instead.
This is probably a bug in the Intrinsics Guide that we should report and track.
The text was updated successfully, but these errors were encountered: