Searched refs:mantissa (Results 1 - 82 of 82) sorted by relevance

/linux-4.4.14/arch/m68k/fpsp040/
H A Dsgetem.S10 | The entry point sGETMAN extracts the mantissa of the
11 | input argument. The mantissa is converted to an
73 | For normalized numbers, leave the mantissa alone, simply load
89 | For denormalized numbers, shift the mantissa until the j-bit = 1,
96 bsr shft |shift mantissa bits till msbit is set
104 | Shifts the mantissa bits until msbit is set.
106 | ms mantissa part in d0
107 | ls mantissa part in d1
H A Ddecbin.S26 | for the mantissa which is to be interpreted as 17 integer
30 | A2. Convert the bcd mantissa to binary by successive
32 | The mantissa digits will be converted with the decimal point
41 | mantissa the equivalent of forcing in the bcd value:
56 | A5. Form the final binary number by scaling the mantissa by
58 | mantissa in FP0 by the factor in FP1 if the adjusted
171 | Calculate mantissa:
172 | 1. Calculate absolute value of mantissa in fp0 by mul and add.
173 | 2. Correct for mantissa sign.
186 | (*) fp0: mantissa accumulator
197 | mantissa. We will unroll the loop once.
203 | Get the rest of the mantissa.
206 movel (%a0,%d1.L*4),%d4 |load mantissa longword into d4
223 addql #1,%d1 |inc lw pointer in mantissa
231 btst #31,(%a0) |test sign of the mantissa
239 | this routine calculates the amount needed to normalize the mantissa
252 | 6. Multiply the mantissa by 10**count.
258 | 6. Divide the mantissa by 10**count.
321 | Calculate the mantissa multiplier to compensate for the striping of
322 | zeros from the mantissa.
337 fmulx %fp1,%fp0 |mul mantissa by 10**(no_bits_shifted)
363 bgts ap_n_fm |if still pos, go fix mantissa
369 | Calculate the mantissa multiplier to compensate for the appending of
370 | zeros to the mantissa.
385 fdivx %fp1,%fp0 |div mantissa by 10**(no_bits_shifted)
478 | (*) fp0: mantissa accumulator
H A Dx_store.S140 beqs get_mant |if positive, go process mantissa
143 bras get_mant |go process mantissa
154 movel LOCAL_HI(%a1),%d1 |get ms mantissa
158 movel LOCAL_HI(%a1),%d1 |get ms mantissa
162 movel LOCAL_LO(%a1),%d1 |get ls mantissa
213 bras get_sman |get mantissa
222 movel LOCAL_HI(%a1),%d1 |get ms mantissa
H A Dround.S154 tstl LOCAL_LO(%a0) |test lower mantissa
163 movel LOCAL_LO(%a0),%d2 |get lower mantissa for s-bit test
195 bccs scc_clr |no mantissa overflow
272 | is done by shifting the mantissa left while decrementing the
276 | bit of the mantissa (msb in d1).
279 | bit of the mantissa (msb in d1) unless this would mean the exponent
281 | exponent (d0) is set to 0 and the mantissa (d1 & d2) is not
289 | Distance to first 1 bit in mantissa = X
295 | shift mantissa by Y
299 | FP_SCR1 = exponent, ms mantissa part, ls mantissa part
350 movew #0,LOCAL_EX(%a0) |no mantissa bits set. Set exp = 0.
476 movel #0,LOCAL_HI(%a0) |set d1 = 0 (ms mantissa)
477 movel #0,LOCAL_LO(%a0) |set d2 = 0 (ms mantissa)
485 | dnrm_lp --- normalize exponent/mantissa to specified threshold
497 | so that bfext can be used to extract the new low part of the mantissa.
H A Dfpsp.h132 .set WBTEMP_HI,WBTEMP+4 | wbtemp mantissa [63:32] (4 bytes)
133 .set WBTEMP_LO,WBTEMP+8 | wbtemp mantissa [31:00] (4 bytes)
165 .set wbtemp66_bit,2 | wbtemp mantissa bit #66
166 .set wbtemp1_bit,1 | wbtemp mantissa bit #1
167 .set wbtemp0_bit,0 | wbtemp mantissa bit #0
201 .set FPTEMP_HI,FPTEMP+4 | fptemp mantissa [63:32] (4 bytes)
202 .set FPTEMP_LO,FPTEMP+8 | fptemp mantissa [31:00] (4 bytes)
208 .set ETEMP_HI,ETEMP+4 | etemp mantissa [63:32] (4 bytes)
209 .set ETEMP_LO,ETEMP+8 | etemp mantissa [31:00] (4 bytes)
302 .set signan_bit,6 | signalling nan bit in mantissa
305 .set rnd_stky_bit,29 | round/sticky bit of mantissa
H A Dx_snan.S14 | of the mantissa are sent to the integer unit).
161 | Get the 32 most significant bits of etemp mantissa
183 | Get the 16 most significant bits of etemp mantissa
205 | Get the 8 most significant bits of etemp mantissa
H A Dbindec.S68 | The mantissa is scaled to the desired number of
90 | the mantissa by 10.
92 | A14. Convert the mantissa to bcd.
94 | mantissa to bcd in memory. The input to binstr is
95 | to be a fraction; i.e. (mantissa)/10^LEN and adjusted
114 | d2: upper 32-bits of mantissa for binstr
115 | d3: scratch;lower 32-bits of mantissa for binstr
451 | The mantissa is scaled to the desired number of significant
545 oril #1,8(%a2) |or in 1 to lsb of mantissa
611 | the mantissa by 10. The calculation of 10^LEN cannot
716 | A14. Convert the mantissa to bcd.
718 | mantissa to bcd in memory. The input to binstr is
719 | to be a fraction; i.e. (mantissa)/10^LEN and adjusted
736 | /ptr to first mantissa byte in result string
768 tstl %d2 |check for mantissa of zero
902 tstl L_SCR2(%a6) |check sign of original mantissa
H A Dscale.S164 | mantissa right until a zero exponent exists.
177 roxrl #1,%d2 |mantissa to the right
182 blts fix_loop |d0 is zero or the mantissa
297 | by first shifting the bits in the mantissa until it is normalized,
H A Dx_unfl.S10 | shifting the mantissa right while incrementing the exponent until
248 tstl LOCAL_HI(%a0) |check upper mantissa
250 tstl LOCAL_LO(%a0) |check lower mantissa
H A Dget_op.S286 btstb #sign_bit,ETEMP_EX(%a6) |grab sign bit of mantissa
503 tstl ETEMP_HI(%a6) |check ms mantissa
505 tstl ETEMP_LO(%a6) |check ls mantissa
555 tstl ETEMP_HI(%a6) |check ms mantissa
557 tstl ETEMP_LO(%a6) |check ls mantissa
H A Dx_operr.S12 | the upper 32 bits of the mantissa are sent to the integer unit). If
283 | mantissa for $ffffffff. If both are true, return d0 clr
H A Dsmovecr.S152 bsr round |go round the mantissa
H A Dres_func.S300 | The mantissa is zero from the denorm loop. Check sign and rmode
963 | and aovfl, and clr the mantissa (incorrectly set by the
1144 | and aovfl, and clr the mantissa (incorrectly set by the
/linux-4.4.14/drivers/hwmon/pmbus/
H A Dzl6100.c69 s32 mantissa; zl6100_l2d() local
73 mantissa = ((s16)((l & 0x7ff) << 5)) >> 5; zl6100_l2d()
75 val = mantissa; zl6100_l2d()
93 s16 exponent = 0, mantissa; zl6100_d2l() local
105 /* Reduce large mantissa until it fits into 10 bit */ zl6100_d2l()
110 /* Increase small mantissa to improve precision */ zl6100_d2l()
116 /* Convert mantissa from milli-units to units */ zl6100_d2l()
117 mantissa = DIV_ROUND_CLOSEST(val, 1000); zl6100_d2l()
120 if (mantissa > 0x3ff) zl6100_d2l()
121 mantissa = 0x3ff; zl6100_d2l()
125 mantissa = -mantissa; zl6100_d2l()
127 /* Convert to 5 bit exponent, 11 bit mantissa */ zl6100_d2l()
128 return (mantissa & 0x7ff) | ((exponent << 11) & 0xf800); zl6100_d2l()
H A Dpmbus_core.c443 s32 mantissa; pmbus_reg2data_linear() local
448 mantissa = (u16) sensor->data; pmbus_reg2data_linear()
451 mantissa = ((s16)((sensor->data & 0x7ff) << 5)) >> 5; pmbus_reg2data_linear()
454 val = mantissa; pmbus_reg2data_linear()
563 s16 exponent = 0, mantissa; pmbus_data2reg_linear() local
603 /* Reduce large mantissa until it fits into 10 bit */ pmbus_data2reg_linear()
608 /* Increase small mantissa to improve precision */ pmbus_data2reg_linear()
614 /* Convert mantissa from milli-units to units */ pmbus_data2reg_linear()
615 mantissa = DIV_ROUND_CLOSEST(val, 1000); pmbus_data2reg_linear()
618 if (mantissa > 0x3ff) pmbus_data2reg_linear()
619 mantissa = 0x3ff; pmbus_data2reg_linear()
623 mantissa = -mantissa; pmbus_data2reg_linear()
625 /* Convert to 5 bit exponent, 11 bit mantissa */ pmbus_data2reg_linear()
626 return (mantissa & 0x7ff) | ((exponent << 11) & 0xf800); pmbus_data2reg_linear()
H A Dpmbus.h354 int m[PSC_NUM_CLASSES]; /* mantissa for direct data format */
H A Dltc2978.c200 * mantissa is 10 bit + sign, exponent adds up to 15 bit. lin11_to_val()
/linux-4.4.14/drivers/iio/light/
H A Dopt3001.c170 u16 mantissa, int *val, int *val2) opt3001_to_iio_ret()
174 lux = 10 * (mantissa << exponent); opt3001_to_iio_ret()
226 u16 mantissa; opt3001_get_lux() local
298 mantissa = OPT3001_REG_MANTISSA(opt->result); opt3001_get_lux()
300 opt3001_to_iio_ret(opt, exponent, mantissa, val, val2); opt3001_get_lux()
437 u16 mantissa; opt3001_write_event_value() local
454 mantissa = (((val * 1000) + (val2 / 1000)) / 10) >> exponent; opt3001_write_event_value()
455 value = (exponent << 12) | mantissa; opt3001_write_event_value()
460 opt->high_thresh_mantissa = mantissa; opt3001_write_event_value()
465 opt->low_thresh_mantissa = mantissa; opt3001_write_event_value()
169 opt3001_to_iio_ret(struct opt3001 *opt, u8 exponent, u16 mantissa, int *val, int *val2) opt3001_to_iio_ret() argument
/linux-4.4.14/drivers/usb/serial/
H A Dpl2303.c365 unsigned int baseline, mantissa, exponent; pl2303_encode_baud_rate_divisor() local
369 * baudrate = 12M * 32 / (mantissa * 4^exponent) pl2303_encode_baud_rate_divisor()
371 * mantissa = buf[8:0] pl2303_encode_baud_rate_divisor()
375 mantissa = baseline / baud; pl2303_encode_baud_rate_divisor()
376 if (mantissa == 0) pl2303_encode_baud_rate_divisor()
377 mantissa = 1; /* Avoid dividing by zero if baud > 32*12M. */ pl2303_encode_baud_rate_divisor()
379 while (mantissa >= 512) { pl2303_encode_baud_rate_divisor()
381 mantissa >>= 2; /* divide by 4 */ pl2303_encode_baud_rate_divisor()
384 /* Exponent is maxed. Trim mantissa and leave. */ pl2303_encode_baud_rate_divisor()
385 mantissa = 511; pl2303_encode_baud_rate_divisor()
392 buf[1] = exponent << 1 | mantissa >> 8; pl2303_encode_baud_rate_divisor()
393 buf[0] = mantissa & 0xff; pl2303_encode_baud_rate_divisor()
396 baud = (baseline / mantissa) >> (exponent << 1); pl2303_encode_baud_rate_divisor()
/linux-4.4.14/arch/parisc/math-emu/
H A Dfloat.h58 * |s| exp | mantissa |
111 * |s| exponent | mantissa part 1 |
115 * | mantissa part 2 |
188 * |s| exponent | mantissa part 1 |
192 * | mantissa part 2 |
196 * | mantissa part 3 |
200 * | mantissa part 4 |
272 * but one additional word of mantissa is needed.
284 * but two additional words of mantissa are needed.
H A Dfcnvuf.c77 * Generate exponent and normalized mantissa sgl_to_sgl_fcnvuf()
136 * Generate exponent and normalized mantissa sgl_to_dbl_fcnvuf()
177 * Generate exponent and normalized mantissa dbl_to_sgl_fcnvuf()
259 * Generate exponent and normalized mantissa dbl_to_dbl_fcnvuf()
H A Dfcnvxf.c85 * Generate exponent and normalized mantissa sgl_to_sgl_fcnvxf()
156 * Generate exponent and normalized mantissa sgl_to_dbl_fcnvxf()
207 * Generate exponent and normalized mantissa dbl_to_sgl_fcnvxf()
312 * Generate exponent and normalized mantissa dbl_to_dbl_fcnvxf()
H A Dfmpyfadd.c290 * Generate multiply mantissa dbl_fmpyfadd()
314 /* need to normalize results mantissa */ dbl_fmpyfadd()
356 /* need to normalize results mantissa */ dbl_fmpyfadd()
381 * which needs enough room for 106 bits of mantissa, dbl_fmpyfadd()
412 /* result mantissa >= 2 (mantissa overflow) */ dbl_fmpyfadd()
628 * is necessary, then the mantissa is all zeros so no shift is needed. dbl_fmpyfadd()
950 * Generate multiply mantissa dbl_fmpynfadd()
974 /* need to normalize results mantissa */ dbl_fmpynfadd()
1016 /* need to normalize results mantissa */ dbl_fmpynfadd()
1041 * which needs enough room for 106 bits of mantissa, dbl_fmpynfadd()
1072 /* result mantissa >= 2 (mantissa overflow) */ dbl_fmpynfadd()
1288 * is necessary, then the mantissa is all zeros so no shift is needed. dbl_fmpynfadd()
1606 * Generate multiply mantissa sgl_fmpyfadd()
1630 /* need to normalize results mantissa */ sgl_fmpyfadd()
1672 /* need to normalize results mantissa */ sgl_fmpyfadd()
1697 * which needs enough room for 106 bits of mantissa, sgl_fmpyfadd()
1724 /* result mantissa >= 2 (mantissa overflow) */ sgl_fmpyfadd()
1932 * is necessary, then the mantissa is all zeros so no shift is needed. sgl_fmpyfadd()
2248 * Generate multiply mantissa sgl_fmpynfadd()
2272 /* need to normalize results mantissa */ sgl_fmpynfadd()
2314 /* need to normalize results mantissa */ sgl_fmpynfadd()
2339 * which needs enough room for 106 bits of mantissa, sgl_fmpynfadd()
2366 /* result mantissa >= 2 (mantissa overflow) */ sgl_fmpynfadd()
2574 * is necessary, then the mantissa is all zeros so no shift is needed. sgl_fmpynfadd()
H A Ddfmpy.c172 * Generate mantissa dbl_fmpy()
241 /* result mantissa >= 2. */ dbl_fmpy()
256 /* align result mantissa */ dbl_fmpy()
H A Dsfmpy.c170 * Generate mantissa sgl_fmpy()
227 /* result mantissa >= 2. */ sgl_fmpy()
242 /* re-align mantissa */ sgl_fmpy()
H A Ddfrem.c248 * opnd1. The mantissa also needs some correction. dbl_frem()
261 /* normalize result's mantissa */ dbl_frem()
H A Dsfrem.c243 * opnd1. The mantissa also needs some correction. sgl_frem()
255 /* normalize result's mantissa */ sgl_frem()
H A Dsgl_float.h155 /* A quiet NaN has the high mantissa bit clear and at least on other (in this
164 /* An infinity is represented with the max exponent and a zero mantissa */
319 * with 48 bits of mantissa.
H A Ddfsqrt.c178 /* increment result exponent by 1 if mantissa overflowed */ dbl_fsqrt()
H A Dsfsqrt.c170 /* increment result exponent by 1 if mantissa overflowed */ sgl_fsqrt()
H A Ddfadd.c203 /* need to normalize results mantissa */ dbl_fadd()
456 * possible. If a postnormalization is necessary, then the mantissa is dbl_fadd()
H A Ddfsub.c206 /* need to normalize results mantissa */ dbl_fsub()
459 * possible. If a postnormalization is necessary, then the mantissa is dbl_fsub()
H A Dsfadd.c203 /* need to normalize results mantissa */ sgl_fadd()
451 * possible. If a postnormalization is necessary, then the mantissa is sgl_fadd()
H A Dsfsub.c204 /* need to normalize results mantissa */ sgl_fsub()
454 * possible. If a postnormalization is necessary, then the mantissa is sgl_fsub()
H A Ddfdiv.c178 * Generate mantissa dbl_fdiv()
H A Dfcnvff.c239 * check for mantissa overflow after rounding dbl_to_sgl_fcnvff()
H A Dsfdiv.c176 * Generate mantissa sgl_fdiv()
H A Ddbl_float.h243 /* A quiet NaN has the high mantissa bit clear and at least on other (in this
269 /* An infinity is represented with the max exponent and a zero mantissa */
535 * with 106 bits of mantissa.
/linux-4.4.14/sound/isa/gus/
H A Dgus_volume.c165 unsigned int mantissa, f1, f2;
179 mantissa = sensitivity % 8192;
186 bend = (int) ((((f2 - f1) * mantissa) >> 13) + f1);
/linux-4.4.14/arch/mips/math-emu/
H A Dsp_flong.c26 u64 xm; /* <--- need 64-bit mantissa temp */ ieee754sp_flong()
H A Ddp_fmax.c115 /* Compare mantissa */ ieee754dp_fmax()
209 /* Compare mantissa */ ieee754dp_fmaxa()
H A Ddp_fmin.c115 /* Compare mantissa */ ieee754dp_fmin()
209 /* Compare mantissa */ ieee754dp_fmina()
H A Dsp_fmax.c115 /* Compare mantissa */ ieee754sp_fmax()
209 /* Compare mantissa */ ieee754sp_fmaxa()
H A Dsp_fmin.c115 /* Compare mantissa */ ieee754sp_fmin()
209 /* Compare mantissa */ ieee754sp_fmina()
H A Dsp_tlong.c32 COMPXDP; /* <-- need 64-bit mantissa tmp */ ieee754sp_tlong()
H A Dieee754dp.c152 /* add causes mantissa overflow */ ieee754dp_format()
H A Dieee754sp.c150 /* add causes mantissa overflow */ ieee754sp_format()
/linux-4.4.14/drivers/isdn/mISDN/
H A Ddsp_audio.c124 int sign, exponent, mantissa; linear2ulaw() local
135 mantissa = (sample >> (exponent + 3)) & 0x0F; linear2ulaw()
136 ulawbyte = ~(sign | (exponent << 4) | mantissa); linear2ulaw()
/linux-4.4.14/arch/arm/vfp/
H A Dvfp.h162 * VFP_SINGLE_MANTISSA_BITS - number of bits in the mantissa
188 * of the single-precision float mantissa with the 1. if necessary,
297 * of the double-precision float mantissa with the 1. if necessary,
/linux-4.4.14/arch/m68k/math-emu/
H A Dfp_util.S108 move.l %d0,(%a0)+ | set mantissa
131 lsl.l #8,%d0 | shift mantissa
168 lsl.l #8,%d0 | shift high mantissa
673 move.l (%a0),%d0 | low lword of mantissa
707 | round (the mantissa, that is) towards infinity
712 | Yow! we have managed to overflow the mantissa. Since this
866 move.l (%a0)+,%d0 | get high lword of mantissa
898 | round (the mantissa, that is) towards infinity
902 | is now zero. We will set the mantissa to reflect this, and
1026 move.l (%a0)+,%d0 | get high lword of mantissa
1058 | round (the mantissa, that is) towards infinity
1062 | is now zero. We will set the mantissa to reflect this, and
H A Dfp_log.c56 * Since only the fractional part of the mantissa is stored and fp_fsqrt()
H A Dfp_arith.c192 /* shift up the mantissa for denormalized numbers, fp_fmul()
279 /* shift up the mantissa for denormalized numbers, fp_fdiv()
H A Dfp_scan.S373 | read this as "1.0 * 2^0" - note the high bit in the mantissa
/linux-4.4.14/arch/arm/nwfpe/
H A Dfpsr.h90 #define BIT_MO 0x08000000 /* mantissa overflow bit */
/linux-4.4.14/drivers/media/i2c/smiapp/
H A Dsmiapp-regs.c57 /* Extract mantissa, add missing '1' bit and it's in MHz */ float_to_u32_mul_1000000()
65 man >>= 23; /* Remove mantissa bias */ float_to_u32_mul_1000000()
/linux-4.4.14/arch/powerpc/kernel/
H A Dvecemu.c69 /* table lookup on top 3 bits of fraction to get mantissa */ eexp2()
234 /* mantissa overflows into exponent - that's OK, rfii()
/linux-4.4.14/kernel/
H A Dacct.c308 #define MANTSIZE 13 /* 13 bit mantissa. */
335 exp += value; /* and add on the mantissa. */ encode_comp_t()
343 * Format: 5 bit base 2 exponent, 20 bits mantissa.
344 * The leading bit of the mantissa is not stored, but implied for
349 #define MANTSIZE2 20 /* 20 bit mantissa. */
/linux-4.4.14/arch/m68k/ifpsp060/src/
H A Dpfpsp.S527 set mantissalen, 64 # length of mantissa in bits
1452 bsr.l norm # normalize mantissa
1484 bsr.l norm # normalize mantissa
2533 # The packed operand is a zero if the mantissa is all zero, else it's
3431 mov.l FP_SRC_HI(%a6),%d1 # load mantissa
3432 lsr.l &0x8,%d1 # shift mantissa for sgl
3446 mov.l FP_SRC_HI(%a6),%d1 # load mantissa
3447 lsr.l &0x8,%d1 # shift mantissa for sgl
3458 mov.l FP_SRC_HI(%a6),%d1 # load hi mantissa
3463 mov.l FP_SRC_HI(%a6),%d1 # load hi mantissa
3467 mov.l FP_SRC_LO(%a6),%d1 # load lo mantissa
3619 # longword integer directly into the upper longword of the mantissa along
5390 # norm() - normalize mantissa after adjusting exponent #
5406 # If the two exponents differ by > the number of mantissa bits #
5448 cmp.w %d0,L_SCR1(%a6) # is difference >= len(mantissa)+2?
5484 cmp.w %d0,2+L_SCR1(%a6) # is difference >= len(mantissa)+2?
5512 # norm() - normalize the mantissa if the operand was a DENORM #
5565 # norm() - normalize the mantissa if the operand was a DENORM #
5639 # norm() - normalize the mantissa if the operand was a DENORM #
5814 # precision, shift the mantissa bits to the right in order raise the #
5816 # mantissa bits right, maintain the value of the guard, round, and #
5837 # simply calculate the sticky bit and zero the mantissa. otherwise
5858 # calculate if the sticky should be set and clear the entire mantissa.
5863 clr.l FTEMP_HI(%a0) # set d1 = 0 (ms mantissa)
5864 clr.l FTEMP_LO(%a0) # set d2 = 0 (ms mantissa)
5868 # dnrm_lp(): normalize exponent/mantissa to specified threshold #
6042 # Return a zero mantissa with the sticky bit set
6044 clr.l FTEMP_HI(%a0) # clear hi(mantissa)
6045 clr.l FTEMP_LO(%a0) # clear lo(mantissa)
6070 mov.l FTEMP_HI(%a0), %d0 # fetch hi(mantissa)
6098 mov.l FTEMP_HI(%a0), %d0 # fetch hi(mantissa)
6115 # the entire mantissa is zero.
6117 clr.l FTEMP_HI(%a0) # clear hi(mantissa)
6118 clr.l FTEMP_LO(%a0) # clear lo(mantissa)
6123 # the entire mantissa is zero.
6127 clr.l FTEMP_HI(%a0) # clear hi(mantissa)
6128 clr.l FTEMP_LO(%a0) # clear lo(mantissa)
6253 bcc.b scc_clr # no mantissa overflow
6380 tst.l FTEMP_LO(%a0) # test lower mantissa
6402 mov.l FTEMP_LO(%a0), %d2 # get lower mantissa for s-bit test
6420 # norm(): normalize the mantissa of an extended precision input. the #
6433 # d0 = number of bit positions the mantissa was shifted #
6434 # a0 = the input operand's mantissa is normalized; the exponent #
6443 mov.l FTEMP_HI(%a0), %d0 # load hi(mantissa)
6444 mov.l FTEMP_LO(%a0), %d1 # load lo(mantissa)
6489 # norm() - normalize the mantissa #
6497 # zero; both the exponent and mantissa are changed. #
6560 # only mantissa bits set are in lo(man)
6577 # whole mantissa is zero so this UNNORM is actually a zero
6600 # Simply test the exponent, j-bit, and mantissa values to #
6680 # Simply test the exponent, j-bit, and mantissa values to #
6743 # Simply test the exponent, j-bit, and mantissa values to #
7319 bsr.l norm # normalize the mantissa
7780 bpl.b dst_get_dman # if positive, go process mantissa
7783 mov.l FTEMP_HI(%a0),%d1 # get ms mantissa
7787 mov.l FTEMP_HI(%a0),%d1 # get ms mantissa
7791 mov.l FTEMP_LO(%a0),%d1 # get ls mantissa
7848 mov.l FTEMP_HI(%a0),%d1 # get ms mantissa
7903 # "mantissa" is all zero which means that the answer is zero. but, the '040
7905 # if the mantissa is zero, I will zero the exponent, too.
8455 # norm() - normalize mantissa for EXOP on denorm #
8539 # normalize the mantissa and add the bias of 0x6000 to the resulting negative
9322 # normalize the mantissa and add the bias of 0x6000 to the resulting negative
9849 # norm() - normalize denorm mantissa to provide EXOP #
9940 # normalize the mantissa and add the bias of 0x6000 to the resulting negative
11291 # exponent for the selected precision. also, the mantissa is equal to
11292 # 0x8000000000000000 and this mantissa is the result of rounding non-zero
11744 # exponent for the selected precision. also, the mantissa is equal to
11745 # 0x8000000000000000 and this mantissa is the result of rounding non-zero
12936 # The packed operand is a zero if the mantissa is all zero, else it's
12970 # for the mantissa which is to be interpreted as 17 integer #
12974 # A2. Convert the bcd mantissa to binary by successive #
12976 # The mantissa digits will be converted with the decimal point #
12985 # mantissa the equivalent of forcing in the bcd value: #
13000 # A5. Form the final binary number by scaling the mantissa by #
13002 # mantissa in FP0 by the factor in FP1 if the adjusted #
13085 # Calculate mantissa:
13086 # 1. Calculate absolute value of mantissa in fp0 by mul and add.
13087 # 2. Correct for mantissa sign.
13100 # (*) fp0: mantissa accumulator
13111 # mantissa. We will unroll the loop once.
13117 # Get the rest of the mantissa.
13120 mov.l (%a0,%d1.L*4),%d4 # load mantissa lonqword into d4
13137 addq.l &1,%d1 # inc lw pointer in mantissa
13144 btst &31,(%a0) # test sign of the mantissa
13151 # this routine calculates the amount needed to normalize the mantissa
13164 # 6. Multiply the mantissa by 10**count.
13170 # 6. Divide the mantissa by 10**count.
13233 # Calculate the mantissa multiplier to compensate for the striping of
13234 # zeros from the mantissa.
13249 fmul.x %fp1,%fp0 # mul mantissa by 10**(no_bits_shifted)
13275 bgt.b ap_n_fm # if still pos, go fix mantissa
13281 # Calculate the mantissa multiplier to compensate for the appending of
13282 # zeros to the mantissa.
13297 fdiv.x %fp1,%fp0 # div mantissa by 10**(no_bits_shifted)
13390 # (*) fp0: mantissa accumulator
13481 # The mantissa is scaled to the desired number of #
13503 # the mantissa by 10. #
13505 # A14. Convert the mantissa to bcd. #
13507 # mantissa to bcd in memory. The input to binstr is #
13508 # to be a fraction; i.e. (mantissa)/10^LEN and adjusted #
13553 # d2: upper 32-bits of mantissa for binstr
13554 # d3: scratch;lower 32-bits of mantissa for binstr
13856 # The mantissa is scaled to the desired number of significant
13922 mov.l 0x8(%a0),-(%sp) # put input op mantissa on stk
13930 mov.l 36+8(%a1),-(%sp) # get 10^8 mantissa
13933 mov.l 48+8(%a1),-(%sp) # get 10^16 mantissa
13977 or.l &1,8(%a2) # or in 1 to lsb of mantissa
14058 # the mantissa by 10. The calculation of 10^LEN cannot
14162 # A14. Convert the mantissa to bcd.
14164 # mantissa to bcd in memory. The input to binstr is
14165 # to be a fraction; i.e. (mantissa)/10^LEN and adjusted
14182 # /ptr to first mantissa byte in result string
14214 tst.l %d2 # check for mantissa of zero
14348 tst.l L_SCR2(%a6) # check sign of original mantissa
H A Dfpsp.S528 set mantissalen, 64 # length of mantissa in bits
1453 bsr.l norm # normalize mantissa
1485 bsr.l norm # normalize mantissa
2534 # The packed operand is a zero if the mantissa is all zero, else it's
3432 mov.l FP_SRC_HI(%a6),%d1 # load mantissa
3433 lsr.l &0x8,%d1 # shift mantissa for sgl
3447 mov.l FP_SRC_HI(%a6),%d1 # load mantissa
3448 lsr.l &0x8,%d1 # shift mantissa for sgl
3459 mov.l FP_SRC_HI(%a6),%d1 # load hi mantissa
3464 mov.l FP_SRC_HI(%a6),%d1 # load hi mantissa
3468 mov.l FP_SRC_LO(%a6),%d1 # load lo mantissa
3620 # longword integer directly into the upper longword of the mantissa along
7607 # sgetman(): extracts the mantissa of the input argument. The #
7608 # mantissa is converted to an extended precision number w/ #
7617 # fp0 = exponent(X) or mantissa(X) #
7662 # For denormalized numbers, shift the mantissa until the j-bit = 1,
9477 bsr.l _round # round the mantissa
9633 mov.l &0x80000000,%d1 # load normalized mantissa
9639 clr.l -(%sp) # insert zero low mantissa
9640 mov.l %d1,-(%sp) # insert new high mantissa
9645 lsr.l %d0,%d1 # make low mantissa longword
9646 mov.l %d1,-(%sp) # insert new low mantissa
9647 clr.l -(%sp) # insert zero high mantissa
9661 mov.l &0x80000000,-(%sp) # insert new high mantissa
9662 mov.l %d0,-(%sp) # insert new lo mantissa
10209 # so, normalize the mantissa, add 0x6000 to the new exponent,
10217 bsr.l norm # normalize mantissa
10285 # dst op is a DENORM. we have to normalize the mantissa to see if the
10293 bsr.l norm # normalize mantissa
10307 andi.l &0x7ff,%d1 # dbl mantissa set?
12024 # norm() - normalize mantissa for EXOP on denorm #
12108 # normalize the mantissa and add the bias of 0x6000 to the resulting negative
12891 # normalize the mantissa and add the bias of 0x6000 to the resulting negative
13418 # norm() - normalize denorm mantissa to provide EXOP #
13509 # normalize the mantissa and add the bias of 0x6000 to the resulting negative
14860 # exponent for the selected precision. also, the mantissa is equal to
14861 # 0x8000000000000000 and this mantissa is the result of rounding non-zero
15313 # exponent for the selected precision. also, the mantissa is equal to
15314 # 0x8000000000000000 and this mantissa is the result of rounding non-zero
15844 # norm() - normalize mantissa after adjusting exponent #
15860 # If the two exponents differ by > the number of mantissa bits #
15902 cmp.w %d0,L_SCR1(%a6) # is difference >= len(mantissa)+2?
15938 cmp.w %d0,2+L_SCR1(%a6) # is difference >= len(mantissa)+2?
15966 # norm() - normalize the mantissa if the operand was a DENORM #
16019 # norm() - normalize the mantissa if the operand was a DENORM #
16093 # norm() - normalize the mantissa if the operand was a DENORM #
20418 bsr.l norm # normalize the mantissa
20879 bpl.b dst_get_dman # if positive, go process mantissa
20882 mov.l FTEMP_HI(%a0),%d1 # get ms mantissa
20886 mov.l FTEMP_HI(%a0),%d1 # get ms mantissa
20890 mov.l FTEMP_LO(%a0),%d1 # get ls mantissa
20947 mov.l FTEMP_HI(%a0),%d1 # get ms mantissa
21002 # "mantissa" is all zero which means that the answer is zero. but, the '040
21004 # if the mantissa is zero, I will zero the exponent, too.
21692 # precision, shift the mantissa bits to the right in order raise the #
21694 # mantissa bits right, maintain the value of the guard, round, and #
21715 # simply calculate the sticky bit and zero the mantissa. otherwise
21736 # calculate if the sticky should be set and clear the entire mantissa.
21741 clr.l FTEMP_HI(%a0) # set d1 = 0 (ms mantissa)
21742 clr.l FTEMP_LO(%a0) # set d2 = 0 (ms mantissa)
21746 # dnrm_lp(): normalize exponent/mantissa to specified threshold #
21920 # Return a zero mantissa with the sticky bit set
21922 clr.l FTEMP_HI(%a0) # clear hi(mantissa)
21923 clr.l FTEMP_LO(%a0) # clear lo(mantissa)
21948 mov.l FTEMP_HI(%a0), %d0 # fetch hi(mantissa)
21976 mov.l FTEMP_HI(%a0), %d0 # fetch hi(mantissa)
21993 # the entire mantissa is zero.
21995 clr.l FTEMP_HI(%a0) # clear hi(mantissa)
21996 clr.l FTEMP_LO(%a0) # clear lo(mantissa)
22001 # the entire mantissa is zero.
22005 clr.l FTEMP_HI(%a0) # clear hi(mantissa)
22006 clr.l FTEMP_LO(%a0) # clear lo(mantissa)
22131 bcc.b scc_clr # no mantissa overflow
22258 tst.l FTEMP_LO(%a0) # test lower mantissa
22280 mov.l FTEMP_LO(%a0), %d2 # get lower mantissa for s-bit test
22298 # norm(): normalize the mantissa of an extended precision input. the #
22311 # d0 = number of bit positions the mantissa was shifted #
22312 # a0 = the input operand's mantissa is normalized; the exponent #
22321 mov.l FTEMP_HI(%a0), %d0 # load hi(mantissa)
22322 mov.l FTEMP_LO(%a0), %d1 # load lo(mantissa)
22367 # norm() - normalize the mantissa #
22375 # zero; both the exponent and mantissa are changed. #
22438 # only mantissa bits set are in lo(man)
22455 # whole mantissa is zero so this UNNORM is actually a zero
22478 # Simply test the exponent, j-bit, and mantissa values to #
22558 # Simply test the exponent, j-bit, and mantissa values to #
22621 # Simply test the exponent, j-bit, and mantissa values to #
22976 # The packed operand is a zero if the mantissa is all zero, else it's
23010 # for the mantissa which is to be interpreted as 17 integer #
23014 # A2. Convert the bcd mantissa to binary by successive #
23016 # The mantissa digits will be converted with the decimal point #
23025 # mantissa the equivalent of forcing in the bcd value: #
23040 # A5. Form the final binary number by scaling the mantissa by #
23042 # mantissa in FP0 by the factor in FP1 if the adjusted #
23125 # Calculate mantissa:
23126 # 1. Calculate absolute value of mantissa in fp0 by mul and add.
23127 # 2. Correct for mantissa sign.
23140 # (*) fp0: mantissa accumulator
23151 # mantissa. We will unroll the loop once.
23157 # Get the rest of the mantissa.
23160 mov.l (%a0,%d1.L*4),%d4 # load mantissa lonqword into d4
23177 addq.l &1,%d1 # inc lw pointer in mantissa
23184 btst &31,(%a0) # test sign of the mantissa
23191 # this routine calculates the amount needed to normalize the mantissa
23204 # 6. Multiply the mantissa by 10**count.
23210 # 6. Divide the mantissa by 10**count.
23273 # Calculate the mantissa multiplier to compensate for the striping of
23274 # zeros from the mantissa.
23289 fmul.x %fp1,%fp0 # mul mantissa by 10**(no_bits_shifted)
23315 bgt.b ap_n_fm # if still pos, go fix mantissa
23321 # Calculate the mantissa multiplier to compensate for the appending of
23322 # zeros to the mantissa.
23337 fdiv.x %fp1,%fp0 # div mantissa by 10**(no_bits_shifted)
23430 # (*) fp0: mantissa accumulator
23521 # The mantissa is scaled to the desired number of #
23543 # the mantissa by 10. #
23545 # A14. Convert the mantissa to bcd. #
23547 # mantissa to bcd in memory. The input to binstr is #
23548 # to be a fraction; i.e. (mantissa)/10^LEN and adjusted #
23593 # d2: upper 32-bits of mantissa for binstr
23594 # d3: scratch;lower 32-bits of mantissa for binstr
23896 # The mantissa is scaled to the desired number of significant
23962 mov.l 0x8(%a0),-(%sp) # put input op mantissa on stk
23970 mov.l 36+8(%a1),-(%sp) # get 10^8 mantissa
23973 mov.l 48+8(%a1),-(%sp) # get 10^16 mantissa
24017 or.l &1,8(%a2) # or in 1 to lsb of mantissa
24098 # the mantissa by 10. The calculation of 10^LEN cannot
24202 # A14. Convert the mantissa to bcd.
24204 # mantissa to bcd in memory. The input to binstr is
24205 # to be a fraction; i.e. (mantissa)/10^LEN and adjusted
24222 # /ptr to first mantissa byte in result string
24254 tst.l %d2 # check for mantissa of zero
24388 tst.l L_SCR2(%a6) # check sign of original mantissa
H A Dfplsp.S508 set mantissalen, 64 # length of mantissa in bits
7501 # sgetman(): extracts the mantissa of the input argument. The #
7502 # mantissa is converted to an extended precision number w/ #
7511 # fp0 = exponent(X) or mantissa(X) #
7556 # For denormalized numbers, shift the mantissa until the j-bit = 1,
9274 mov.l &0x80000000,%d1 # load normalized mantissa
9280 clr.l -(%sp) # insert zero low mantissa
9281 mov.l %d1,-(%sp) # insert new high mantissa
9286 lsr.l %d0,%d1 # make low mantissa longword
9287 mov.l %d1,-(%sp) # insert new low mantissa
9288 clr.l -(%sp) # insert zero high mantissa
9302 mov.l &0x80000000,-(%sp) # insert new high mantissa
9303 mov.l %d0,-(%sp) # insert new lo mantissa
9764 # Simply test the exponent, j-bit, and mantissa values to #
10037 # dst op is a DENORM. we have to normalize the mantissa to see if the
10045 bsr.l norm # normalize mantissa
10058 andi.l &0x7ff,%d1 # dbl mantissa set?
10817 # norm(): normalize the mantissa of an extended precision input. the #
10830 # d0 = number of bit positions the mantissa was shifted #
10831 # a0 = the input operand's mantissa is normalized; the exponent #
10840 mov.l FTEMP_HI(%a0), %d0 # load hi(mantissa)
10841 mov.l FTEMP_LO(%a0), %d1 # load lo(mantissa)
10886 # norm() - normalize the mantissa #
10894 # zero; both the exponent and mantissa are changed. #
10957 # only mantissa bits set are in lo(man)
10974 # whole mantissa is zero so this UNNORM is actually a zero
/linux-4.4.14/drivers/tty/serial/
H A Dstm32-usart.c350 u32 usartdiv, mantissa, fraction, oversampling; stm32_set_termios() local
403 mantissa = (usartdiv / oversampling) << USART_BRR_DIV_M_SHIFT; stm32_set_termios()
405 writel_relaxed(mantissa | fraction, port->membase + USART_BRR); stm32_set_termios()
/linux-4.4.14/arch/x86/math-emu/
H A Dreg_u_div.S151 /* shift the mantissa right one bit */
427 /* shift the mantissa right one bit */
H A Dreg_ld_str.c442 /* Truncate the mantissa */ FPU_store_double()
628 /* Truncate part of the mantissa */ FPU_store_single()
/linux-4.4.14/drivers/md/bcache/
H A Dbset.c243 unsigned mantissa:BKEY_MANTISSA_BITS; member in struct:bkey_float
499 * bits we're going to store in bkey_float->mantissa. t->prev[j] stores the size
584 f->mantissa = bfloat_mantissa(m, f) - 1; make_bfloat()
908 * n = (f->mantissa > bfloat_mantissa()) bset_search_tree()
912 * We need to subtract 1 from f->mantissa for the sign bit trick bset_search_tree()
917 (f->mantissa - bset_search_tree()
H A Dbset.h127 * point, with an exponent and a mantissa. The exponent needs to be big enough
129 * mantissa is somewhat arbitrary; more bits just gets us fewer failures.
132 * are 8 byte aligned); using 22 bits for the mantissa means a node is 4 bytes.
/linux-4.4.14/drivers/isdn/i4l/
H A Disdn_audio.c265 mantissa; isdn_audio_linear2ulaw() local
278 mantissa = (sample >> (exponent + 3)) & 0x0F; isdn_audio_linear2ulaw()
279 ulawbyte = ~(sign | (exponent << 4) | mantissa); isdn_audio_linear2ulaw()
/linux-4.4.14/include/scsi/
H A Dosd_protocol.h104 * byte offset = mantissa * (2^(exponent+8))
106 * unsigned mantissa: 28;
/linux-4.4.14/drivers/staging/comedi/drivers/
H A Ddt2811.c147 bits 5-3 - Timer frequency control (mantissa)
/linux-4.4.14/include/net/
H A Dcodel.h203 * Here, invsqrt is a fixed point number (< 1.0), 32bit mantissa, aka Q0.32
H A Dred.h106 * max_P is a Q0.32 fixed point number (with 32 bits mantissa)
/linux-4.4.14/drivers/pcmcia/
H A Dcistpl.c35 static const u_char mantissa[] = { variable
46 (mantissa[(((v)>>3)&15)-1] * exponent[(v)&7] / 10)
49 (mantissa[((v)>>3)&15] * exponent[(v)&7] / 10)
/linux-4.4.14/drivers/isdn/hardware/mISDN/
H A Dhfcmulti.c1842 u_int mantissa; hfcmulti_dtmf() local
1881 mantissa = w_float & 0x0fff; hfcmulti_dtmf()
1883 mantissa |= 0xfffff000; hfcmulti_dtmf()
1886 mantissa ^= 0x1000; hfcmulti_dtmf()
1887 mantissa <<= (exponent - 1); hfcmulti_dtmf()
1891 coeff[co << 1] = mantissa; hfcmulti_dtmf()
1900 mantissa = w_float & 0x0fff; hfcmulti_dtmf()
1902 mantissa |= 0xfffff000; hfcmulti_dtmf()
1905 mantissa ^= 0x1000; hfcmulti_dtmf()
1906 mantissa <<= (exponent - 1); hfcmulti_dtmf()
1910 coeff[(co << 1) | 1] = mantissa; hfcmulti_dtmf()
/linux-4.4.14/drivers/gpu/drm/i915/
H A Dintel_tv.c159 * exp.mantissa (ee.mmmmmmmmm)
165 * exp.mantissa (eee.mmmmmmmmm)
177 * exp.mantissa (ee.mmmmmm)
H A Di915_reg.h3786 * mantissa and 2 or 3 bits of exponent. The exponent is represented as 2**-n,
/linux-4.4.14/drivers/atm/
H A Diphase.c299 ** | R | NZ | 5-bit exponent | 9-bit mantissa |
312 #define M_BITS 9 /* Number of bits in mantissa */ cellrate_to_float()
341 u32 exp, mantissa, cps;
345 mantissa = rate & M_MASK;
348 cps = (1 << M_BITS) | mantissa;
H A Dfirestream.c412 // Now the mantissa is in positions bit 16-25. Excepf for the "hidden 1" that's in bit 26.
505 the bits of the mantissa that are to be discarded). make_rate()
H A Dambassador.c943 // the bits of the mantissa that are to be discarded). make_rate()
H A Dhe.c706 * this table maps the upper 5 bits of exponent and mantissa he_init_cs_block_rcm()
/linux-4.4.14/arch/powerpc/perf/
H A Dpower8-pmu.c359 * Check the mantissa upper two bits are not zero, unless the power8_get_constraint()
/linux-4.4.14/drivers/scsi/osd/
H A Dosd_initiator.c2034 * byte offset = mantissa * (2^(exponent+8))
2067 OSD_DEBUG("offset=%llu mantissa=%llu exp=%d encoded=%x pad=%d\n", __osd_encode_offset()
/linux-4.4.14/drivers/net/wireless/ath/ath5k/
H A Dphy.c279 * mantissa and provide these values on hw.
288 /* Get exponent and mantissa and set it */ ath5k_hw_write_ofdm_timings()
327 /* Get mantissa (significant digits) ath5k_hw_write_ofdm_timings()
333 * and mantissa (remove scaling) and set them on hw */ ath5k_hw_write_ofdm_timings()
/linux-4.4.14/include/uapi/linux/usb/
H A Dch9.h894 #define USB_SSP_SUBLINK_SPEED_LSM (0xff << 16) /* Lanespeed mantissa */

Completed in 2428 milliseconds