Commit 5e4ef64c authored by ths's avatar ths
Browse files

Simplify round/ceil/floor implementation, spotted by Fabrice Bellard.


git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@3031 c046a42c-6fe2-441c-8c8c-71466251a162
parent ea6e4fe2
......@@ -782,8 +782,7 @@ FLOAT_OP(cvtw, d)
FLOAT_OP(roundl, d)
{
set_float_rounding_mode(float_round_nearest_even, &env->fp_status);
DT2 = float64_round_to_int(FDT0, &env->fp_status);
DT2 = float64_to_int64(DT2, &env->fp_status);
DT2 = float64_to_int64(FDT0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -792,8 +791,7 @@ FLOAT_OP(roundl, d)
FLOAT_OP(roundl, s)
{
set_float_rounding_mode(float_round_nearest_even, &env->fp_status);
WT2 = float32_round_to_int(FST0, &env->fp_status);
DT2 = float32_to_int64(WT2, &env->fp_status);
DT2 = float32_to_int64(FST0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -802,10 +800,7 @@ FLOAT_OP(roundl, s)
FLOAT_OP(roundw, d)
{
set_float_rounding_mode(float_round_nearest_even, &env->fp_status);
DT2 = float64_round_to_int(FDT0, &env->fp_status);
// ???
env->fp_status.float_exception_flags &= ~float_flag_inexact;
WT2 = float64_to_int32(DT2, &env->fp_status);
WT2 = float64_to_int32(FDT0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -814,8 +809,7 @@ FLOAT_OP(roundw, d)
FLOAT_OP(roundw, s)
{
set_float_rounding_mode(float_round_nearest_even, &env->fp_status);
WT2 = float32_round_to_int(FST0, &env->fp_status);
WT2 = float32_to_int32(WT2, &env->fp_status);
WT2 = float32_to_int32(FST0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -854,8 +848,7 @@ FLOAT_OP(truncw, s)
FLOAT_OP(ceill, d)
{
set_float_rounding_mode(float_round_up, &env->fp_status);
DT2 = float64_round_to_int(FDT0, &env->fp_status);
DT2 = float64_to_int64(DT2, &env->fp_status);
DT2 = float64_to_int64(FDT0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -864,8 +857,7 @@ FLOAT_OP(ceill, d)
FLOAT_OP(ceill, s)
{
set_float_rounding_mode(float_round_up, &env->fp_status);
WT2 = float32_round_to_int(FST0, &env->fp_status);
DT2 = float32_to_int64(WT2, &env->fp_status);
DT2 = float32_to_int64(FST0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -874,10 +866,7 @@ FLOAT_OP(ceill, s)
FLOAT_OP(ceilw, d)
{
set_float_rounding_mode(float_round_up, &env->fp_status);
DT2 = float64_round_to_int(FDT0, &env->fp_status);
// ???
env->fp_status.float_exception_flags &= ~float_flag_inexact;
WT2 = float64_to_int32(DT2, &env->fp_status);
WT2 = float64_to_int32(FDT0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -886,8 +875,7 @@ FLOAT_OP(ceilw, d)
FLOAT_OP(ceilw, s)
{
set_float_rounding_mode(float_round_up, &env->fp_status);
WT2 = float32_round_to_int(FST0, &env->fp_status);
WT2 = float32_to_int32(WT2, &env->fp_status);
WT2 = float32_to_int32(FST0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -897,8 +885,7 @@ FLOAT_OP(ceilw, s)
FLOAT_OP(floorl, d)
{
set_float_rounding_mode(float_round_down, &env->fp_status);
DT2 = float64_round_to_int(FDT0, &env->fp_status);
DT2 = float64_to_int64(DT2, &env->fp_status);
DT2 = float64_to_int64(FDT0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -907,8 +894,7 @@ FLOAT_OP(floorl, d)
FLOAT_OP(floorl, s)
{
set_float_rounding_mode(float_round_down, &env->fp_status);
WT2 = float32_round_to_int(FST0, &env->fp_status);
DT2 = float32_to_int64(WT2, &env->fp_status);
DT2 = float32_to_int64(FST0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -917,10 +903,7 @@ FLOAT_OP(floorl, s)
FLOAT_OP(floorw, d)
{
set_float_rounding_mode(float_round_down, &env->fp_status);
DT2 = float64_round_to_int(FDT0, &env->fp_status);
// ???
env->fp_status.float_exception_flags &= ~float_flag_inexact;
WT2 = float64_to_int32(DT2, &env->fp_status);
WT2 = float64_to_int32(FDT0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......@@ -929,8 +912,7 @@ FLOAT_OP(floorw, d)
FLOAT_OP(floorw, s)
{
set_float_rounding_mode(float_round_down, &env->fp_status);
WT2 = float32_round_to_int(FST0, &env->fp_status);
WT2 = float32_to_int32(WT2, &env->fp_status);
WT2 = float32_to_int32(FST0, &env->fp_status);
RESTORE_ROUNDING_MODE;
update_fcr31();
if (GET_FP_CAUSE(env->fcr31) & (FP_OVERFLOW | FP_INVALID))
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment