Sign in
Search adsp
Subscribe to adsp
adsp by Keywords
AD1819 | AD7332 | ADSP-2106 | ADSP-21060 | ADSP-21065L | ADSP-2116 | ADSP-21160M | ADSP-2181 | ADSP-218x | ADSP-219 | ADSP-2199 | ADSP219 | BF531 | BF532 | BF533 | BF535 | Blackfin | FFT | JTAG | LDF | SDRAM | SHARC | SPORT | UART | VDSP++ | VisualDSP
Discussion Groups
Discussion Groups | Analog Devices DSPs | floating point multiplication in ADSP 218x
Technical discussions related to Analog Devices DSPs (including Blackfin, TigerSHARC, SHARC and ADSP-21xx DSPs).
floating point multiplication in ADSP 218x - Sonali Agarwal - Mar 15 0:29:00 2005
|
Hello, I have some confusion regarding the multiplication in ADSP 218x. For example, i want to multiply 1.5 * 1.5 which should give 2.25 in assembly language of ADSP 218x. to represent 1.5 , i use the 2.14 format which gives it to be 0x6000. Now when i keep MX0 = 0x6000 and MY0 = 0x6000 and give MR = MX0 * MY0, the result is MR2 = 00 MR1 = 4800 MR0 = 0000 Now the problem is how to shift the data to bring it to the format understandable by the DSP. Can anybody help me in this regard. I hope i am able to make my point clear. Or if you can suggest soem technical material for this. Thanks, SOnali __________________________________ |
(You need to be a member of adsp -- send a blank email to adsp-subscribe@yahoogroups.com )
Re: floating point multiplication in ADSP 218x - Mike Rosing - Mar 15 10:04:00 2005
|
On Tue, 15 Mar 2005, Sonali Agarwal wrote: > I have some confusion regarding the > multiplication in ADSP 218x. > For example, i want to multiply 1.5 * 1.5 which should > give 2.25 in assembly language of ADSP 218x. > to represent 1.5 , i use the 2.14 format which gives > it to be 0x6000. > Now when i keep MX0 = 0x6000 and MY0 = 0x6000 and give > MR = MX0 * MY0, the result is > MR2 = 00 > MR1 = 4800 > MR0 = 0000 > Now the problem is how to shift the data to bring it > to the format understandable by the DSP. > Can anybody help me in this regard. I hope i am able > to make my point clear. Or if you can suggest soem > technical material for this. the multiplier is doing things with fractions, and you want a different representation. 0x6000 = .75 and .75*.75 = 0x4800. Since you want to scale everything by 2 in each number you want to scale things by 4 in the final result. So 0x4800 is now in 3.13 format (1 bit sign and 2 bits before binary point since multiplier takes care of the sign for you). Patience, persistence, truth, Dr. mike |
(You need to be a member of adsp -- send a blank email to adsp-subscribe@yahoogroups.com )
Re: floating point multiplication in ADSP 218x - Jaime Andres Aranguren Cardona - Mar 16 14:54:00 2005
|
Hi, The 218x has a mode (M_MODE, if recall well) on which the DSP makes the shift itself, but it only works for 1.15 x 1.15 = 2.30 -> 1.31. For other formats, you'd heve to do the shift by yourself, using the barrel shifter. What result do you expect (expressed in your 2.14 format?) JaaC --- Sonali Agarwal <> wrote: > Hello, > I have some confusion regarding the > multiplication in ADSP 218x. > For example, i want to multiply 1.5 * 1.5 which > should > give 2.25 in assembly language of ADSP 218x. > to represent 1.5 , i use the 2.14 format which gives > it to be 0x6000. > Now when i keep MX0 = 0x6000 and MY0 = 0x6000 and > give > MR = MX0 * MY0, the result is > MR2 = 00 > MR1 = 4800 > MR0 = 0000 > Now the problem is how to shift the data to bring it > to the format understandable by the DSP. > Can anybody help me in this regard. I hope i am able > to make my point clear. Or if you can suggest soem > technical material for this. > Thanks, > SOnali Jaime Andrés Aranguren Cardona __________________________________ |
(You need to be a member of adsp -- send a blank email to adsp-subscribe@yahoogroups.com )