From lrobertson@netcom.com Sun Jun 05 18:35:54 1994 Path: ashpoly!bigsur!pacrim!jpgate!kamome!netcom!lrobertson From: lrobertson@netcom.com (Liam Robertson) Newsgroups: comp.dsp Subject: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1994Jun05.183554.3743@netcom.com> Date: Sun, 05 Jun 1994 18:35:54 GMT Organization: NETCOM On-Line Communication Services NNTP-Posting-Host: netcom.com X-Newsreader: WinVN 0.99.5 Lines: 10 Porting stardust.c to the NeXTcube and Atari Falcon DSP subsystems. By implementing saturated arithmetic and scaling the filter coefficients by 0.5, the loop runs in 18 instruction cycles per sample. However, when downsampling from 44.1 kHz to 64 Hz, accumulator overflow in the biquad IIR filter can destabilize the loop. Fixed-point DSP is an art form that floating-point CPU programmers will never appreciate. -- Liam Robertson Tape trading list upon request. SASE required for international mail. From bcampbell@pacbell.net Mon Jun 06 08:01:54 1994 Path: ashpoly!cascade!transat!sunet!nordrelay!pacbell!bcampbell From: bcampbell@pacbell.net (Brian Campbell) Newsgroups: comp.dsp Subject: Re: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1994Jun06.080154.2371@pacbell.net> Date: Mon, 06 Jun 1994 08:01:54 GMT References: <1994Jun05.183554.3743@netcom.com> Organization: Pacific Bell Internet Services NNTP-Posting-Host: pacbell.net X-Newsreader: TIN [version 1.2 PL2] Lines: 10 In article <1994Jun05.183554.3743@netcom.com>, lrobertson@netcom.com (Liam Robertson) writes: > Porting stardust.c to the NeXTcube and Atari Falcon DSP subsystems. > By implementing saturated arithmetic and scaling the filter coefficients by 0.5, the loop runs in 18 instruction cycles per sample. Real programmers write in C, not arcane DSP microcode. With Pentium chips hitting 100 MHz, dedicated DSPs are obsolete. -- === Brian Campbell === Dept of Electrical Engineering, Pacific Bell Internet Services "Hold for relay." From dandersson@cam.ac.uk Tue Jun 07 13:12:54 1994 Path: ashpoly!cascade!transat!sunet!nordrelay!cam!dandersson From: dandersson@cam.ac.uk (Diego Andersson) Newsgroups: comp.dsp Subject: Re: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1994Jun07.131254.7104@cam.ac.uk> Date: Tue, 07 Jun 1994 13:12:54 GMT References: <1994Jun05.183554.3743@netcom.com> Organization: University of Cambridge Computer Laboratory NNTP-Posting-Host: cam.ac.uk X-Newsreader: Netscape 1.1N Lines: 10 In article <1994Jun05.183554.3743@netcom.com>, lrobertson@netcom.com (Liam Robertson) writes: > Porting stardust.c to the NeXTcube and Atari Falcon DSP subsystems. > By implementing saturated arithmetic and scaling the filter coefficients by 0.5, the loop runs in 18 instruction cycles per sample. Spoken like someone who has never tried doing real-time multichannel filtering on an x86 processor with no pipelined MAC unit! -- Diego Andersson *PLONK* filter active for trolls. PC-9821Ap2 (NEC) - NetBSD 1.0 From cstewart@purdue.edu Wed Jun 08 06:40:54 1994 Path: ashpoly!cascade!mxgate!coyoacan!purdue!cstewart From: cstewart@purdue.edu (Camila Stewart) Newsgroups: comp.dsp Subject: Re: Fixed-point Q15 arithmetic on Motorola DSP56001 vs IEEE floating point Message-ID: <1994Jun08.064054.8677@purdue.edu> Date: Wed, 08 Jun 1994 06:40:54 GMT References: <1994Jun05.183554.3743@netcom.com> <1994Jun07.131254.7104@cam.ac.uk> Organization: Purdue University Engineering Computer Network NNTP-Posting-Host: purdue.edu X-Newsreader: Netscape 1.1N Lines: 9 In article <1994Jun07.131254.7104@cam.ac.uk>, dandersson@cam.ac.uk (Diego Andersson) writes: > Spoken like someone who has never tried doing real-time multichannel filtering on an x86 processor with no pipelined MAC unit! Beautiful assembly code! The 24-bit word size of the 56001 is tailor-made for high-precision audio filtering. -- === Camila Stewart === Dept of Electrical Engineering, Purdue University Engineering Computer Network "5 sigma is a discovery. 7 is a question."