<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    Hi Marek,<br>
    <br>
    <div class="moz-cite-prefix">Am 10.05.22 um 22:43 schrieb Marek
      Olšák:<br>
    </div>
    <blockquote type="cite"
cite="mid:CAAxE2A7wOfoWWh5VUFmnhyhNeCQ086trJR2BgT+nAmsYZJTbVg@mail.gmail.com">
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      <div dir="ltr">
        <div>A better flag name would be:</div>
        <div>AMDGPU_GEM_CREATE_BEST_PLACEMENT_OR_DISCARD</div>
      </div>
    </blockquote>
    <br>
    A bit long for my taste and I think the best placement is just a
    side effect.<br>
    <br>
    <blockquote type="cite"
cite="mid:CAAxE2A7wOfoWWh5VUFmnhyhNeCQ086trJR2BgT+nAmsYZJTbVg@mail.gmail.com">
      <div dir="ltr">
        <div><br>
        </div>
        <div>Marek<br>
        </div>
      </div>
      <br>
      <div class="gmail_quote">
        <div dir="ltr" class="gmail_attr">On Tue, May 10, 2022 at 4:13
          PM Marek Olšák <<a href="mailto:maraeo@gmail.com"
            moz-do-not-send="true" class="moz-txt-link-freetext">maraeo@gmail.com</a>>
          wrote:<br>
        </div>
        <blockquote class="gmail_quote" style="margin:0px 0px 0px
          0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
          <div dir="ltr">
            <div>Does this really guarantee VRAM placement? The code
              doesn't say anything about that.</div>
          </div>
        </blockquote>
      </div>
    </blockquote>
    <br>
    Yes, see the code here:<br>
    <br>
    <blockquote type="cite"
cite="mid:CAAxE2A7wOfoWWh5VUFmnhyhNeCQ086trJR2BgT+nAmsYZJTbVg@mail.gmail.com">
      <div class="gmail_quote">
        <blockquote class="gmail_quote" style="margin:0px 0px 0px
          0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><br>
          <div class="gmail_quote">
            <blockquote class="gmail_quote" style="margin:0px 0px 0px
              0.8ex;border-left:1px solid
              rgb(204,204,204);padding-left:1ex">
              diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
              b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c<br>
              index 8b7ee1142d9a..1944ef37a61e 100644<br>
              --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c<br>
              +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c<br>
              @@ -567,6 +567,7 @@ int amdgpu_bo_create(struct
              amdgpu_device *adev,<br>
                              bp->domain;<br>
                      bo->allowed_domains = bo->preferred_domains;<br>
                      if (bp->type != ttm_bo_type_kernel &&<br>
              +           !(bp->flags &
              AMDGPU_GEM_CREATE_DISCARDABLE) &&<br>
                          bo->allowed_domains ==
              AMDGPU_GEM_DOMAIN_VRAM)<br>
                              bo->allowed_domains |=
              AMDGPU_GEM_DOMAIN_GTT;<br>
            </blockquote>
          </div>
        </blockquote>
      </div>
    </blockquote>
    <br>
    The only case where this could be circumvented is when you try to
    allocate more than physically available on an APU.<br>
    <br>
    E.g. you only have something like 32 MiB VRAM and request 64 MiB,
    then the GEM code will catch the error and fallback to GTT (IIRC).<br>
    <br>
    Regards,<br>
    Christian.<br>
  </body>
</html>