| Message ID | 20200918231259.GA16472@aloka.lostca.se |
|---|---|
| State | New |
| Headers | show |
| Series | [v2] benchtests: Run _Float128 tests only on architectures that support it | expand |
Dear Arjun, this patch applies cleanly on master, and I've tested it on x86_64 Linux: make bench works and produces results for expf128, powf128 and sinf128, as expected. Thus ok on my side. I let Florian and Joseph confirm. Paul
* Arjun Shankar: > From: Arjun Shankar <arjun@redhat.com> > > __float128 is a non-standard name and is not available on some architectures > (like aarch64 or s390x) even though they may support the standard _Float128 > type. Other architectures (like armv7) don't support quad-precision > floating-point operations at all. > > This commit replaces benchtests references to __float128 with _Float128 and > runs the corresponding tests only on architectures that support it. This looks good to me. Thanks. Florian
diff --git a/benchtests/Makefile b/benchtests/Makefile index 3095076055..922e2a94b1 100644 --- a/benchtests/Makefile +++ b/benchtests/Makefile @@ -25,7 +25,11 @@ bench-math := acos acosh asin asinh atan atanh cos cosh exp exp2 log log2 \ modf pow rint sin sincos sinh sqrt tan tanh fmin fmax fminf \ fmaxf powf trunc truncf roundeven roundevenf expf exp2f logf \ log2f sincosf sinf cosf isnan isinf isfinite hypot logb logbf \ - exp10f expf128 powf128 sinf128 + exp10f + +ifneq (,$(filter yes,$(float128-fcts) $(float128-alias-fcts))) +bench-math += expf128 powf128 sinf128 +endif bench-pthread := pthread_once thread_create diff --git a/benchtests/expf128-inputs b/benchtests/expf128-inputs index 5b36f8672a..5dc0f4e49b 100644 --- a/benchtests/expf128-inputs +++ b/benchtests/expf128-inputs @@ -1,6 +1,6 @@ ## includes: math.h -## args: __float128 -## ret: __float128 +## args: _Float128 +## ret: _Float128 # Random inputs in [-10,10] ## name: workload-random.wrf 0x4.e6f9d6da10d9a422942a89cdfa1p+0L diff --git a/benchtests/powf128-inputs b/benchtests/powf128-inputs index 7cbabafff6..6826266ad5 100644 --- a/benchtests/powf128-inputs +++ b/benchtests/powf128-inputs @@ -1,6 +1,6 @@ ## includes: math.h -## args: __float128:__float128 -## ret: __float128 +## args: _Float128:_Float128 +## ret: _Float128 # Random inputs in [-10,10] such that x and y are not both negative ## name: workload-random.wrf 0x8.130b31ed5288656428a29cead83p+0L, -0x6.e7ead09b7877a118813b50cfb3c8p+0L diff --git a/benchtests/sinf128-inputs b/benchtests/sinf128-inputs index 9aaf312413..01b7533680 100644 --- a/benchtests/sinf128-inputs +++ b/benchtests/sinf128-inputs @@ -1,6 +1,6 @@ ## includes: math.h -## args: __float128 -## ret: __float128 +## args: _Float128 +## ret: _Float128 # Random inputs in [-10,10] ## name: workload-random.wrf 0x4.e6f9d6da10d9a422942a89cdfa1p+0
From: Arjun Shankar <arjun@redhat.com> __float128 is a non-standard name and is not available on some architectures (like aarch64 or s390x) even though they may support the standard _Float128 type. Other architectures (like armv7) don't support quad-precision floating-point operations at all. This commit replaces benchtests references to __float128 with _Float128 and runs the corresponding tests only on architectures that support it. --- Thanks for looking at v1, Florian, Joseph and Paul. v2 changes: dropped configure[.ac] changes; used the float128[-alias]-fcts variables instead. Tested on armv7hl, i686, x86_64, aarch64, ppc64le, and s390x. benchtests/Makefile | 6 +++++- benchtests/expf128-inputs | 4 ++-- benchtests/powf128-inputs | 4 ++-- benchtests/sinf128-inputs | 4 ++-- 4 files changed, 11 insertions(+), 7 deletions(-)