Lines Matching refs:this
24 so long as this entire notice is retained without alteration in any modified and/or
34 # Also, subroutine stubs exist in this file (_fpsp_done for
619 # If the exception is enabled, then this handler must create the #
622 # exceptions enabled, this handler must exit through the "callout" #
624 # can handle this case. #
626 # but the inexact exception was enabled, this handler must exit #
630 # overflow was disabled and the trace exception was enabled, this #
672 # will never take this exception.
749 frestore FP_SRC(%a6) # do this after fmovm,other f<op>s!
768 frestore FP_SRC(%a6) # do this after fmovm,other f<op>s!
859 # If the exception is enabled, then this handler must create the #
862 # exceptions enabled, this handler must exit through the "callout" #
864 # can handle this case. #
867 # this handler must exit through the "callout" _real_inex(). #
870 # underflow was disabled and the trace exception was enabled, this #
910 # will never take this exception.
963 # underflow exception. Since this is incorrect, we need to check
966 # funfl_{unfl,inex}_on() because w/ both exceptions disabled, this
996 # underflow exception. Since this is incorrect, we check here to see
1011 frestore FP_SRC(%a6) # do this after fmovm,other f<op>s!
1042 frestore FP_SRC(%a6) # do this after fmovm,other f<op>s!
1152 # this exception into the FPU in the fsave state frame and then exit #
1221 # stack an ea. however, we don't need an a7' for this case anyways.
1230 # this point.
1252 # I'm not sure at this point what FPSR bits are valid for this instruction.
1280 # at this point
1375 # * this is the case where we must call _real_inex() now or else
1425 # If the input operand to this operation was opclass two and a single
1506 # I'm not sure at this point what FPSR bits are valid for this instruction.
1560 # it here. if it was used from supervisor mode, then we have to handle this
1808 # I'm not sure at this point what FPSR bits are valid for this instruction.
1825 # at this point
1953 # * this is the case where we must call _real_inex() now or else
1977 # this is INCORRECT for enabled SNAN which would give to the user the skewed SNAN!!!
2065 # I'm not sure at this point what FPSR bits are valid for this instruction.
2422 # as a result of this operation emulation. A Trace exception can be #
2426 # was executed from supervisor mode, this handler must store the FP #
2428 # fmovm_dynamic() can't handle this. A normal exit is made through #
2447 # status-longword. These are created within this handler. #
2545 fmovm.x &0x80,FP_SRC(%a6) # make this the srcop
2625 tst.b STORE_FLG(%a6) # does this op store a result?
2928 # this ain't a pretty solution, but it works:
3090 # for opclass zero and two instruction taking this exception, the #
3093 # do just this and then exits through _real_operr(). #
3099 # In this special case, the handler must exit through facc_out() #
3131 # this would be the case for opclass two operations with a source infinity or
3151 # operand error exceptions. we do this here before passing control to
3280 # for opclass zero and two instructions taking this exception, the #
3283 # do just this and then exits through _real_snan(). #
3289 # In this special case, the handler must exit through facc_out() #
3327 # this would be the case for opclass two operations with a source infinity or
3347 # snan exceptions. we do this here before passing control to
3581 # for opclass zero and two instruction taking this exception, the #
3584 # instruction in order to get this value and then store it to the #
3620 # w/ an exponent value of 0x401e. we convert this to extended precision here.
3654 # will never take this exception, but fsincos will.
3712 # inexact exceptions. we do this here before passing control to
3761 # be incorrect for some cases and need to be adjusted. So, this package #
3790 # this would be the case for opclass two operations with a source zero
3900 # an imem_read to fetch this later.
3918 # mode so we don't even worry about this tricky case here : )
4071 # math operation relative to the address of this table. Included are
4073 # this table is for the version if the 060FPSP without transcendentals.
4191 # Add this here so non-fp modules can compile.
4225 # The effective address must be calculated since this is entered #
4241 # so that the move can occur outside of this routine. This special #
4575 # currently, MODE and REG are taken from the EXC_OPWORD. this could be
5267 # Fetch them using _imem_read_long(). If this fetch fails, exit through #
5405 # do the opposite. Return this scale factor in d0. #
5524 # normalize the operand if it was a DENORM. Add this normalization #
5651 # normalize the operand if it was a DENORM. Add this normalization #
6231 # Note that this will round to even in case of a tie. #
6337 # prior to usage, and needs to restore d1 to original. this
6349 # this function. so, as long as we don't disturb it, we are "returning" it.
6577 # whole mantissa is zero so this UNNORM is actually a zero
6790 # scaled extended precision number; this is used by #
7082 # the rounded result. Also, since exceptions are disabled, this also #
7266 # 16-bit field gets zeroed. we do this since we promise not to disturb
7306 # so _mem_write2() handles this for us.
7356 # Unnormalized inputs can come through this point.
7594 # Unnormalized inputs can come through this point.
7884 # we'll have to change this, but for now, tough luck!!!
7929 # so _mem_write2() handles this for us.
8050 # normalized then we really don't need to go through this scaling. but for now,
8051 # this will do.
8087 # extended precision. if the original operation was extended, then we have this
8090 # of this operation then has its exponent scaled by -0x6000 to create the
8106 # save setting this until now because this is where fmul_may_ovfl may jump in
8198 # extended precision. if the original operation was extended, then we have this
8201 # of this operation then has its exponent scaled by -0x6000 to create the
8280 # -use the correct rounding mode and precision. this code favors operations
8305 # this case should be relatively rare.
8521 # the accrued bit is NOT set in this instance(no inexactness!)
8759 # do the fmove in; at this point, only possible ops are ZERO and INF.
8761 # prec:mode should be zero at this point but it won't affect answer anyways.
9073 # result is. this case should be relatively rare.
9298 # the accrued bit is NOT set in this instance(no inexactness!)
9539 # do the fneg; at this point, only possible ops are ZERO and INF.
9541 # prec:mode should be zero at this point but it won't affect answer anyways.
9918 # the accrued bit is NOT set in this instance(no inexactness!)
10466 # save setting this until now because this is where fsglmul_may_ovfl may jump in
10600 # this case should be relatively rare.
10936 # result is. this case should be relatively rare.
11271 # if the precision is extended, this result could not have come from an
11292 # 0x8000000000000000 and this mantissa is the result of rounding non-zero
11296 # so, we do this be re-executing the add using RZ as the rounding mode and
11724 # if the precision is extended, this result could not have come from an
11745 # 0x8000000000000000 and this mantissa is the result of rounding non-zero
11749 # so, we do this be re-executing the add using RZ as the rounding mode and
12288 # this routine leaves d1 intact for subsequent store_dreg calls.
12560 # this routine adds the increment value in d0 to the address register #
12565 # an access error exception occurs later in emulation, this address #
12624 # this routine adds the decrement value in d0 to the address register #
12629 # an access error exception occurs later in emulation, this address #
12950 fmovm.x &0x80,FP_SRC(%a6) # make this the srcop
12965 # and NaN operands are dispatched without entering this routine) #
12972 # Note: this operation can never overflow. #
12978 # Note: this operation can never overflow. #
12990 # this will insure that any value, regardless of its #
13132 # else continue with this long word.
13135 dbf.w %d2,md2b # check for last digit in this lw
13151 # this routine calculates the amount needed to normalize the mantissa
13154 # of this is to reduce the value of the exponent and the possibility
13175 # of 27 or less are exact, there is no need to use this routine to
13192 # First check the absolute value of the exponent to see if this
13411 beq.b no_exc # skip this if no exc
13786 bgt.b k_pos # if pos and > 0, skip this
13788 blt.b k_pos # if ILOG >= k, skip this
13860 # to overflow. Only a negative iscale can cause this, so
13903 # in this way, we avoid underflow on intermediate stages of the
14104 # 10^LEN-1 is computed for this test and A14. If the input was
14138 # and shouldn't have another; this is to check if abs(YINT) = 10^LEN
14156 fbneq.w A14_st # if not, skip this