TCG: 0.958
graphic: 0.896
files: 0.867
device: 0.800
user-level: 0.720
performance: 0.604
semantic: 0.540
socket: 0.512
boot: 0.509
network: 0.445
vnc: 0.434
ppc: 0.398
risc-v: 0.379
PID: 0.370
debug: 0.334
register: 0.334
kernel: 0.306
mistranslation: 0.286
VMM: 0.266
arm: 0.254
architecture: 0.142
permissions: 0.108
virtual: 0.096
x86: 0.080
KVM: 0.075
hypervisor: 0.060
i386: 0.052
peripherals: 0.049
assembly: 0.030
--------------------
TCG: 0.906
files: 0.744
kernel: 0.416
debug: 0.194
user-level: 0.144
VMM: 0.036
semantic: 0.023
register: 0.018
virtual: 0.014
device: 0.013
performance: 0.010
peripherals: 0.010
assembly: 0.009
PID: 0.006
architecture: 0.006
KVM: 0.005
hypervisor: 0.004
graphic: 0.004
risc-v: 0.003
network: 0.002
boot: 0.001
vnc: 0.001
socket: 0.001
permissions: 0.001
mistranslation: 0.001
x86: 0.000
ppc: 0.000
arm: 0.000
i386: 0.000

s390x floating point conversion functions broken
Description of problem:
While collecting additional reference files for float_convs (and float_convd) I noticed that the s390x handling of some cases is broken. See diff for details:

```
 diff -y tests/tcg/s390x-linux-user/float_convs.out ../../tests/tcg/s390x/float_convs.ref
#