diff mbox series

[2/2] benchtests: Add sqrtf benchmark

Message ID 20260731053259.86382-2-jz531210@gmail.com (mailing list archive)
State Superseded
Headers
Series [1/2] benchtests: Add fmaximum and fminimum family benchmarks |

Checks

Context Check Description
redhat-pt-bot/TryBot-apply_patch success Patch applied to master at the time it was sent
redhat-pt-bot/TryBot-32bit fail Patch caused testsuite regressions
linaro-tcwg-bot/tcwg_glibc_build--master-aarch64 success Build passed
linaro-tcwg-bot/tcwg_glibc_build--master-arm success Build passed
linaro-tcwg-bot/tcwg_glibc_check--master-arm success Test passed
linaro-tcwg-bot/tcwg_glibc_check--master-aarch64 success Test passed

Commit Message

Julian Zhu July 31, 2026, 5:32 a.m. UTC 
  Add representative sqrtf inputs and disable compiler builtin expansion so the benchmark measures the libc implementation.

Signed-off-by: Julian Zhu <jz531210@gmail.com>
---
 benchtests/Makefile     |  2 ++
 benchtests/sqrtf-inputs | 10 ++++++++++
 2 files changed, 12 insertions(+)
 create mode 100644 benchtests/sqrtf-inputs

Comments

Paul Zimmermann July 31, 2026, 5:50 a.m. UTC | #1 
Hi Julian,

your input file is quite short. Many input files have at least 1000 entries,
to avoid that branch prediction perturbs results.

Paul
diff mbox series

Patch

diff --git a/benchtests/Makefile b/benchtests/Makefile
index a288c6ca74..6c21cddcbf 100644
--- a/benchtests/Makefile
+++ b/benchtests/Makefile
@@ -150,6 +150,7 @@  bench-math := \
   sinpi \
   sinpif \
   sqrt \
+  sqrtf \
   tan \
   tanf \
   tanh \
@@ -356,6 +357,7 @@  endif
 CFLAGS-bench-ffs.c += -fno-builtin
 CFLAGS-bench-ffsll.c += -fno-builtin
 CFLAGS-bench-sqrt.c += -fno-builtin
+CFLAGS-bench-sqrtf.c += -fno-builtin
 CFLAGS-bench-fmin.c += -fno-builtin
 CFLAGS-bench-fminf.c += -fno-builtin
 CFLAGS-bench-fma.c += -fno-builtin
diff --git a/benchtests/sqrtf-inputs b/benchtests/sqrtf-inputs
new file mode 100644
index 0000000000..675c462740
--- /dev/null
+++ b/benchtests/sqrtf-inputs
@@ -0,0 +1,10 @@ 
+## args: float
+## ret: float
+## includes: math.h
+0.25
+0.75
+2.0
+4.0
+2209.0
+6642.25
+15190.5625