I am building sleef using the IntelLLVM (oneAPI) 2024.0.0 compiler. This error comes up immediately:
$ make
[ 0%] Building C object src/libm/CMakeFiles/mkrename_gnuabi.dir/mkrename_gnuabi.c.o
error: unknown argument '--extended_float_type'; did you mean '--extended_float_types'?
make[2]: *** [src/libm/CMakeFiles/mkrename_gnuabi.dir/build.make:76: src/libm/CMakeFiles/mkrename_gnuabi.dir/mkrename_gnuabi.c.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:838: src/libm/CMakeFiles/mkrename_gnuabi.dir/all] Error 2
make: *** [Makefile:146: all] Error 2
If I add the "s" to the offending flag, I see this later in the build:
icx: warning: overriding '-ffp-model=fast=2' option with '-ffp-model=fast' [-Woverriding-t-option]
This page in Intel's documentation does not indicate any numeric suboption to -fp-model=fast.
This PR adds a conditional to Configure.cmake that uses the modified values for the IntelLLVM variant. With this change, I get a clean build.
Note: Intel online community posts imply that the Intel Classic compiler may accept the --extended_float_types (plural) option in addition to the singular form, but I do not have a way of verifying this.
I am building sleef using the IntelLLVM (oneAPI) 2024.0.0 compiler. This error comes up immediately:
If I add the "s" to the offending flag, I see this later in the build:
This page in Intel's documentation does not indicate any numeric suboption to
-fp-model=fast
.This PR adds a conditional to
Configure.cmake
that uses the modified values for the IntelLLVM variant. With this change, I get a clean build.Note: Intel online community posts imply that the Intel Classic compiler may accept the
--extended_float_types
(plural) option in addition to the singular form, but I do not have a way of verifying this.