ShuriZma
/
xqemu
mirror of https://github.com/xqemu/xqemu.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

hardfloat: fix float32/64 fused multiply-add 

Before falling back to softfloat FMA, we do not restore the original
values of inputs A and C. Fix it.

This bug was caught by running gcc's testsuite on RISC-V qemu.

Note that this change gives a small perf increase for fp-bench:

  Host: Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz
  Command: perf stat -r 3 taskset -c 0 ./fp-bench -o mulAdd -p $prec

- $prec = single:
  - before:
    101.71 MFlops
    102.18 MFlops
    100.96 MFlops
  - after:
    103.63 MFlops
    103.05 MFlops
    102.96 MFlops

- $prec = double:
  - before:
    173.10 MFlops
    173.93 MFlops
    172.11 MFlops
  - after:
    178.49 MFlops
    178.88 MFlops
    178.66 MFlops

Signed-off-by: Kito Cheng <kito.cheng@gmail.com>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20190322204320.17777-1-cota@braap.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
This commit is contained in:
Kito Cheng 2019-03-22 16:43:20 -04:00 committed by Alex Bennée
parent 7ca96e1a9c
commit 896f51fbfa
1 changed files with 10 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
fpu/softfloat.c
View File

@ -1596,6 +1596,9 @@ float32_muladd(float32 xa, float32 xb, float32 xc, int flags, float_status *s)
}
ur.h = up.h + uc.h;
} else {
union_float32 ua_orig = ua;
union_float32 uc_orig = uc;
if (flags & float_muladd_negate_product) {
ua.h = -ua.h;
}
@ -1608,6 +1611,8 @@ float32_muladd(float32 xa, float32 xb, float32 xc, int flags, float_status *s)
if (unlikely(f32_is_inf(ur))) {
s->float_exception_flags |= float_flag_overflow;
} else if (unlikely(fabsf(ur.h) <= FLT_MIN)) {
ua = ua_orig;
uc = uc_orig;
goto soft;
}
}
@ -1662,6 +1667,9 @@ float64_muladd(float64 xa, float64 xb, float64 xc, int flags, float_status *s)
}
ur.h = up.h + uc.h;
} else {
union_float64 ua_orig = ua;
union_float64 uc_orig = uc;
if (flags & float_muladd_negate_product) {
ua.h = -ua.h;
}
@ -1674,6 +1682,8 @@ float64_muladd(float64 xa, float64 xb, float64 xc, int flags, float_status *s)
if (unlikely(f64_is_inf(ur))) {
s->float_exception_flags |= float_flag_overflow;
} else if (unlikely(fabs(ur.h) <= FLT_MIN)) {
ua = ua_orig;
uc = uc_orig;
goto soft;
}
}