From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mike Wellington Date: Fri, 09 Jan 2004 08:50:50 -0700 Subject: [U-Boot-Users] BDI2000 vs Vision-ICE In-Reply-To: <8039970AF146314597457D3B51A68B382E0461@cossmgmbx02.email.corp.tld> References: <8039970AF146314597457D3B51A68B382E0461@cossmgmbx02.email.corp.tld> Message-ID: <3FFECDDA.4000105@lucent.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Guys- My replies are in post. -thanks mike wellington wellington at lucent.com VanBaren, Gerald (AGRE) wrote: > I've used both VisionICE and BDI2000 (we bought two VisionICEs initially and only BDI2000s since then :-). Both connect via the JTAG port and neither support traceback. As Charlie points out, traceback requires capturing the address and data bus which is bloody tricky on todays processors. > my project lead thinks visionIce/visionEvent has solved that problem. Since our CPU core is inside an FPGA I guess I could conceivably watch the bus with ChipScope - an FPGA-based logic analyzer which would give me raw bus cycles. > With a JTAG or a software only debugger you could theoretically enable the "trace on branch" PPC exception and run at full speed in (very short :-) bursts, saving each branch location and then rebuild the traceback using the saved the branches, but that would still cause a significant speed degradation because of all the exceptions. I don't know if anyone does this. I don't know if anyone does this either, but I think it is an excellent idea. I have my doubts whether this would be useful due to the speed slowdown. > I think my application can handle the speed slowdown. >>-----Original Message----- >>From: u-boot-users-admin at lists.sourceforge.net >>[mailto:u-boot-users-admin at lists.sourceforge.net]On Behalf Of Wells, >>Charles >>Sent: Thursday, January 08, 2004 5:57 PM >>To: u-boot-users at lists.sourceforge.net >>Cc: 'Mike Wellington' >>Subject: RE: [U-Boot-Users] BDI2000 vs Vision-ICE >> >> >>Mike, >> >> >>>People at my place of work are telling me that the >>>Vision-Ice supports "backtrace" and the BDI2000 does >>>not. >> >>I don't believe visionICE supports what you describe, but visionEVENT >>(another WRS/EST product) does. visionEVENT behaves like a >>classic "bus >>capture" analyzer. It's no longer a "10-bin BDM port >>connected device." >>visionEVENT is housed in a seperate box that attaches to the >>bottom of the >>visionICE case and requires two 80-pin high-density connectors on your >>target for its connection. Further, visionEVENT imposes some nasty >>restrictions on the target (e.g. not being able to run the >>CPU clock at 2x >>bus clock). >> >>We bought the visionICE/visionEVENT stuff a couple of years >>ago. I use >>visionICE regularly and it works adequately for bringing up >>new targets and >>debugging startup code. We haven't use visionEVENT much at >>all. We've just >>never needed its capabilities. I've never used the BDI2000, >>but it sounds >>like its Linux integration is better than either visionICE or >>visionEVENT >>(although WRS may have improved this since we took delivery >>of ours). Some think the visionICE, visionEVENT stuff has improved. I'll find out soon enough since management went ahead and bought it. >> >>BTW, I agree with Wolfgang's earlier point. What really >>matters is the >>capabilities of the debugger software front-end. One of >>these days, I need >>to get someone around here to approve the purchase of a >>BDI2000 and see how >>it compares. >> Our debugger front-end is supposed to be great. I haven't fired it up yet for real. I kinda need a U-Boot bootloader for the Xilinx ML300/PPC405 first. Or I could use Redboot ( part of eCos Real-Time OS ) which I already have sorce for a ML300/PPC405 but I haven't figured out how to build it yet. -EOF-