<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    On 23.02.2012 18:00, Jerome Glisse wrote:
    <blockquote
cite="mid:CAH3drwYGubUtLhFGsq7W6+E06Av2WpVyWZsYzJbsrU_4RqVLrw@mail.gmail.com"
      type="cite"><br>
      <div class="gmail_quote">2012/2/23 Christian K&ouml;nig <span
          dir="ltr">&lt;<a moz-do-not-send="true"
            href="mailto:deathsimple@vodafone.de">deathsimple@vodafone.de</a>&gt;</span><br>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          So don't confuse devs by doing so.<br>
          <br>
          Signed-off-by: Christian K&ouml;nig &lt;<a moz-do-not-send="true"
            href="mailto:deathsimple@vodafone..de">deathsimple@vodafone.de</a>&gt;<br>
          ---<br>
          &nbsp;drivers/gpu/drm/radeon/r600.c | &nbsp; 15 +--------------<br>
          &nbsp;1 files changed, 1 insertions(+), 14 deletions(-)<br>
          <br>
          diff --git a/drivers/gpu/drm/radeon/r600.c
          b/drivers/gpu/drm/radeon/r600.c<br>
          index 4f08e5e..4a4ac8f 100644<br>
          --- a/drivers/gpu/drm/radeon/r600.c<br>
          +++ b/drivers/gpu/drm/radeon/r600.c<br>
          @@ -2719,20 +2719,7 @@ int r600_ib_test(struct radeon_device
          *rdev, int ring)<br>
          &nbsp; &nbsp; &nbsp; &nbsp;ib-&gt;ptr[0] = PACKET3(PACKET3_SET_CONFIG_REG, 1);<br>
          &nbsp; &nbsp; &nbsp; &nbsp;ib-&gt;ptr[1] = ((scratch -
          PACKET3_SET_CONFIG_REG_OFFSET) &gt;&gt; 2);<br>
          &nbsp; &nbsp; &nbsp; &nbsp;ib-&gt;ptr[2] = 0xDEADBEEF;<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[3] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[4] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[5] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[6] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[7] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[8] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[9] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[10] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[11] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[12] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[13] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[14] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;ptr[15] = PACKET2(0);<br>
          - &nbsp; &nbsp; &nbsp; ib-&gt;length_dw = 16;<br>
          + &nbsp; &nbsp; &nbsp; ib-&gt;length_dw = 3;<br>
          &nbsp; &nbsp; &nbsp; &nbsp;r = radeon_ib_schedule(rdev, ib);<br>
          &nbsp; &nbsp; &nbsp; &nbsp;if (r) {<br>
          &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;radeon_scratch_free(rdev, scratch);<br>
          <span class="HOEnZb"><font color="#888888">--<br>
              1.7.5.4</font></span><br>
        </blockquote>
      </div>
      <br>
      You sure about that ? I remember this helped with GPU lockup and i
      also seen fglrx aligning IB.<br>
    </blockquote>
    Yeah, pretty much. Well I searched for halve an hour for the
    corresponding IB alignment in mesa/the CS ioctl until I finally
    figured out that there isn't any. <br>
    <br>
    So IBs submitted by usermode aren't aligned in any way.... So it
    really seems to work fine and I couldn't find any reason why we
    should align an IB for the GFX ring in our docs also.<br>
    <br>
    Christian.<br>
  </body>
</html>