| Add Proposal | Add Analysis | Edit Class, Environment, or Release |
| Number | 111
|
| Category | errata
|
| Synopsis | 17.2.4.4: How should $fread deal with partial data to a memory?
|
| State | open
|
| Class | errata-discuss
|
| Arrival-Date | Sep 09 2002
|
| Originator | Charles Dawson
|
| Release | 2001b: 17.2.4.4
|
| Environment |
|
| Description |
Section 17.2.4.4 Reading binary Data This section does not adequately describe what should happen if there is partial data. Should the partial data be read and placed into the memory using the most or least significant words? Should it not be read at all since it doesn't fit (I think this is what C does)? For example, suppose you have: reg [15:0] mem [1:0]; And you have 3 bytes of data. Should the $fread result in: A) mem[0] = {byte1,byte2} mem[1] = {byte3,0} B) mem[0] = {byte1,byte2} mem[1] = {0,byte3} C) mem[0] = {byte2,byte1} mem[1] = {0,0} In these cases I am assuming that the bytes not containing data from the file are set to zero. Should they keep whatever value was there previously? This seems like a bad idea to me (especially for cases A and B above), but the spec doesn't say. |
| Fix |
I think option B makes the most sense, since that is the way a string would be read. |
| Audit-Trail |
From: Gordon Vreugdenhil <gvreugde@synopsys.com> To: etf-bugs@boyd.com Cc: Subject: Re: errata/111: How should $fread deal with partial data to a memory? Date: Tue, 10 Sep 2002 10:40:50 -0700 I neglected to send this to etf-bugs (thanks Stephen). Gord. -------- Original Message -------- Subject: Re: errata/111: How should $fread deal with partial data to a memory? Date: Mon, 09 Sep 2002 14:20:22 -0700 From: Gordon Vreugdenhil <Gordon.Vreugdenhil@synopsys.COM> Reply-To: Gordon.Vreugdenhil@synopsys.COM To: chas@cadence.com CC: etf@boyd.com References: <200209091941.g89JfQK01158@max.boyd.com> Precedence: bulk chas@cadence.com wrote: > > Precedence: bulk > > >Number: 111 > >Category: errata > >Originator: Charles Dawson > >Environment: > > >Description: > > Section 17.2.4.4 Reading binary Data > > This section does not adequately describe what should > happen if there is partial data. Should the partial > data be read and placed into the memory using the most > or least significant words? Should it not be read at > all since it doesn't fit (I think this is what C does)? > > For example, suppose you have: > > reg [15:0] mem [1:0]; > > And you have 3 bytes of data. Should the $fread result > in: > > A) mem[0] = {byte1,byte2} > mem[1] = {byte3,0} > > B) mem[0] = {byte1,byte2} > mem[1] = {0,byte3} > > C) mem[0] = {byte2,byte1} > mem[1] = {0,0} > > In these cases I am assuming that the bytes not containing > data from the file are set to zero. Should they keep > whatever value was there previously? This seems like a > bad idea to me (especially for cases A and B above), but > the spec doesn't say. I think that the "strict" LRM answer is (A) but without zeroing the low bits of the word. It certainly is not clear what to do with a partial read however. What I would prefer (and what VCS currently does coincidentally :-) is to ignore any trailing partial read. The problem that does show up there is what to do with the "extra" bytes of data. One could unget them and return a count that would always be a multiple of the byte size of the word or one could just throw them away and return a count that includes "thrown away" bytes. Again, VCS currently does the latter (throws away the bytes and returns the byte count including the thrown away bytes). Note that we *never* clear the remainder of the memory in any case nor do we write a 0 for a partially read trailing word. Gord. -- ---------------------------------------------------------------------- Gord Vreugdenhil gvreugde@synopsys.com Staff Engineer, VCS (Verification Tech. Group) (503) 547-6054 Synopsys Inc., Beaverton OR |
| Unformatted |
|
Hosted by Boyd Technology