Second part of patches submitted in the v3.
This series adds a subset of single precision and
double precision instructions using
set_floatx80_rounding_precision() to round the result.
For some other instructions, we introduce a new
function, floatx80_round(), to round them manually.
It also adds instructions fsglmul, fsgldiv,
fmovecr and fscc (since v3, use tcg_gen_setcond())
fmovem manages static and dynamic register list,
all the code has been moved to an helper.
v4:
Update floatx80_round() comment (and add Aurélien's R-b)
Add a comment for undefined fmovecr offsets
update gen_fcc_cond() with more fixes from Richard
v3:
Remove extra space in fmovem patch
truncate operands of fsglmul and fslgdiv
update gen_fcc_cond() as suggested by Richard
fmoverc: use opmode instead of rom_offset,
define and free the tcg constant
v2:
Fix gen_fcc_cond(): g[12] mark global variables,
not variables to free.
add 2 patches to compute trigonometric functions
and fmod
use floatx80_round() to compute fsglmul and fsgldiv
split fmovem helper into 6 helpers
Laurent Vivier (7):
target/m68k: add fscc.
target/m68k: add fmovecr
target/m68k: add explicit single and double precision operations
softfloat: define floatx80_round()
target/m68k: add fsglmul and fsgldiv
target/m68k: add explicit single and double precision operations (part
2)
target/m68k: add fmovem
fpu/softfloat.c | 16 ++
include/fpu/softfloat.h | 1 +
target/m68k/fpu_helper.c | 310 ++++++++++++++++++++++++++++++++++++-
target/m68k/helper.h | 27 +++-
target/m68k/translate.c | 388 ++++++++++++++++++++++++++++++++---------------
5 files changed, 619 insertions(+), 123 deletions(-)
--
2.9.4