Re: DirectX 9.0C - Bug

From: Robin (sanchrp_at_auburn.edu)
Date: 10/13/04 

Date: Wed, 13 Oct 2004 18:09:08 +0200

Lior, why not dump out your managed buffers back to the disk after loading
everything up that way you can see what the faulty data looks like, perhaps
you'll need to tweak the lock offset or size for the managed buffers in
order to get the right input...

Robin

"Lior Canetti" <LiorCanetti@discussions.microsoft.com> a écrit dans le
message de news:CEBA1178-403E-4085-9B1C-D278609DD24D@microsoft.com...
> Hi,
> A piece of code that run fine on DirectX 9.0b generates artifacts with
9.0c .
>
> The code involves streaming of vertices and indices from the disk to
vertex
> and index buffers - The streaming is made by performing a series of
updates,
> and in each one, locking only the relevant part of the relevant buffer .
>
> On DirectX 9.0b the system allways works .
> On DirectX 9.0c ths system works if the buffers are created with the
> D3DPOOL_SYSMEM or D3DPOOL_DEFAULT, and generates artifacts ( Most of the
sece
> is fine, but there is a small amount of artifacts, which can be
geometrically
> very big ) if the D3DPOOL_MANGED is sent .
>
> It might be a lock bug for the D3DPOOL_MANGED memory system .
>
> Using the D3DPOOL_SYSMEM is not an option for Us because our scene is made
> of a very big amount of triangles .
>
> 2 Quastions :
>
> 1. Is this a known issue, and Do You know a way to bypass it ?
> 2. Since My company hase customers using applications with this code,
> We want to provide a working system, So in the meantime, one of our
> solutions is to change the DirectX runtime version to 9.0b - Do You know
any
> standard operation to do this ?
>
>
> Thank's In Advance : Lior Canetti .
>

Relevant Pages

  • Re: Larrabee delayed: anyone know whats happening?
    ... Do you mean that there are modern systems that *can't* easily get 90% of theoretical disk bandwidth without using massive buffers simply by using multi-buffering in system RAM? ... stage is not written to enable streaming, ...
    (comp.arch)
  • Re: Larrabee delayed: anyone know whats happening?
    ... Do you mean that there are modern systems that *can't* easily get 90% of theoretical disk bandwidth without using massive buffers simply by using multi-buffering in system RAM? ... All that's required is a disk which can transfer the data for the previous request to the host and accept a subsequent request for the next data to be read while it's reading the current request's data from its platters - or issue completion notification for the previous write request and accept the data for the subsequent write request while transferring the current write request's data to its platters, capabilities present in all current SATA disks, in SCSI disks for decades, and in some disks even earlier. ...
    (comp.arch)
  • Re: [vfs] buf_daemon() slows down write() severely on low-speed CPU
    ... tracing of related VFS, filesystem, and disk function calls. ... its buffers flushing mission is failing. ... best copy times were achived when bufdaemon was excluded from buffers ... blocks often are written and that evokes more disk sector flushes. ...
    (freebsd-hackers)
  • Re: How does disk caching work?
    ... >>What's a good way to tune amount of space dedicated to IO buffers? ... > available KVM rather for other purposes (kernel resources). ... that he'd noticed a reduction in disk IO when he did that. ...
    (freebsd-performance)
  • [PATCH 2/2] ext3: add data=guarded mode
    ... a workqueue where the real work of updating the on disk i_size is done. ... end_io handler for buffers that are marked as guarded. ... When we start tracking guarded buffers on a given inode, ... and it also takes a reference on the buffer head. ...
    (Linux-Kernel)