From webbra.mlist at verizon.net Tue Jan 15 06:45:14 2008 From: webbra.mlist at verizon.net (Rich Webb) Date: Tue Jan 15 07:00:28 2008 Subject: [Icc-arm] ARM doubles In-Reply-To: <200712252026.lBPKQNc8070266@dragonsgate2.imagecraft.com> References: <200712252026.lBPKQNc8070266@dragonsgate2.imagecraft.com> Message-ID: <478CC6FA.4070507@verizon.net> Moving back into the ARM world after several 8051-family projects and find that I may need to use doubles in an upcoming application. Both the flyer at http://www.imagecraft.com/pub/ICCV7_ARMflyer.pdf and the ICC V7 ARM Help file under C Runtime Architecture are pretty clear that the compiler supports standard doubles. I.e., "doubles are in IEEE-standard 64-bit format with 11-bit exponent, 52-bit mantissa, and 1 sign bit." However, I can't seem to coerce the compiler into actually using 64-bit doubles. Is there a command-line flag that I'm missing? From richard-lists at imagecraft.com Tue Jan 15 14:10:11 2008 From: richard-lists at imagecraft.com (Richard Man) Date: Tue Jan 15 14:25:09 2008 Subject: [Icc-arm] ARM doubles In-Reply-To: <478CC6FA.4070507@verizon.net> References: <200712252026.lBPKQNc8070266@dragonsgate2.imagecraft.com> <478CC6FA.4070507@verizon.net> Message-ID: <200801152225.m0FMP7Lr053408@dragonsgate2.imagecraft.com> Hmm... the manual is in error. I will fix it. The PRO version *WILL* support 64 bits double in some near future. We need to finalize the AVR version first. At 06:45 AM 1/15/2008, Rich Webb wrote: >Moving back into the ARM world after several 8051-family projects >and find that I may need to use doubles in an upcoming application. > >Both the flyer at http://www.imagecraft.com/pub/ICCV7_ARMflyer.pdf >and the ICC V7 ARM Help file under C Runtime Architecture are pretty >clear that the compiler supports standard doubles. I.e., "doubles >are in IEEE-standard 64-bit format with 11-bit exponent, 52-bit >mantissa, and 1 sign bit." > >However, I can't seem to coerce the compiler into actually using >64-bit doubles. Is there a command-line flag that I'm missing? > >_ // richard (This email is for mailing lists. To reach me directly, please use richard at imagecraft.com) From webbra.mlist at verizon.net Wed Jan 16 13:12:29 2008 From: webbra.mlist at verizon.net (Rich Webb) Date: Wed Jan 16 13:27:53 2008 Subject: [Icc-arm] ARM doubles In-Reply-To: <200801162027.m0GKRdOu065660@dragonsgate2.imagecraft.com> References: <200801162027.m0GKRdOu065660@dragonsgate2.imagecraft.com> Message-ID: <478E733D.4070705@verizon.net> > Hmm... the manual is in error. I will fix it. The PRO version *WILL* > support 64 bits double in some near future. We need to finalize the > AVR version first. Maybe I don't understand but both the 2006 flyer (the one that is currently linked from the product page) and the current version help files state that the ARM compiler is a "FULL" ANSI compiler with support for IEEE doubles. Not so with, e.g., the AVR compilers. At the time the ARM flyer was distributed, the STD version was the only one available so a natural reading would be that all ARM compiler versions supported doubles and that the other "availability TBD" versions would have that in addition to their other stated capabilities (thumb support, global optimizations). Yes, it sometimes takes longer that we'd like to get to "feature complete" but IEEE doubles are what Imagecraft has been advertising as a capability across all of the ARM compiler family. From richard-lists at imagecraft.com Wed Jan 16 14:51:52 2008 From: richard-lists at imagecraft.com (Richard Man) Date: Wed Jan 16 15:06:38 2008 Subject: [Icc-arm] ARM doubles In-Reply-To: <478E733D.4070705@verizon.net> References: <200801162027.m0GKRdOu065660@dragonsgate2.imagecraft.com> <478E733D.4070705@verizon.net> Message-ID: <200801162306.m0GN6bg3067508@dragonsgate2.imagecraft.com> Well, in the ideal case, we will have added support for it two years ago, but in the ideal case, I wouldn't be working 60 hours week either. So the short answer is, yes, I am sorry the description is in error. We have full intention to add it in the near future. If you think we have misled you, contact me privately and I will issue you a refund. At 01:12 PM 1/16/2008, Rich Webb wrote: >>Hmm... the manual is in error. I will fix it. The PRO version >>*WILL* support 64 bits double in some near future. We need to >>finalize the AVR version first. > >Maybe I don't understand but both the 2006 flyer (the one that is >currently linked from the product page) and the current version help >files state that the ARM compiler is a "FULL" ANSI compiler with >support for IEEE doubles. Not so with, e.g., the AVR compilers. > >At the time the ARM flyer was distributed, the STD version was the >only one available so a natural reading would be that all ARM compiler >versions supported doubles and that the other "availability TBD" >versions would have that in addition to their other stated capabilities >(thumb support, global optimizations). > >Yes, it sometimes takes longer that we'd like to get to "feature >complete" but IEEE doubles are what Imagecraft has been advertising >as a capability across all of the ARM compiler family. > >_______________________________________________ >Icc-arm mailing list >Icc-arm@imagecraft.com >http://dragonsgate.net/mailman/listinfo/icc-arm // richard (This email is for mailing lists. To reach me directly, please use richard at imagecraft.com) From webbra.mlist at verizon.net Thu Jan 17 12:21:16 2008 From: webbra.mlist at verizon.net (Rich Webb) Date: Thu Jan 17 12:36:41 2008 Subject: [Icc-arm] Re: Icc-arm Digest, Vol 33, Issue 3 In-Reply-To: <200801172008.m0HK8Ujb081757@dragonsgate2.imagecraft.com> References: <200801172008.m0HK8Ujb081757@dragonsgate2.imagecraft.com> Message-ID: <478FB8BC.2020209@verizon.net> > Well, in the ideal case, we will have added support for it two years > ago, but in the ideal case, I wouldn't be working 60 hours week either. > > So the short answer is, yes, I am sorry the description is in error. > We have full intention to add it in the near future. If you think we > have misled you, contact me privately and I will issue you a refund. Nah, I'll stick with the didjiman that I know. And, since I'm already running the ADV version, it's not going to be toooo painful to move to the PRO release if/when it becomes necessary. Might be a good idea to scrub those flyers, though. The good news is that testing today on an LPC with simulated data against a Mathcad model seems to show that "regular" floats will be sufficient, which is A Good Thing since they're faster and I'm the glue between a navigation sensor block and a stack of synchro converters.