Age | Commit message (Collapse) | Author | Files | Lines |
|
Simple shader that write the thread x id to buffer[threadid.x] ie pseudo code:
buffer[threadid.x] = threadid.x;
In the old way the buffer is allocated with regular GPU memory allocation and
everything is pin while in the new way this is simply malloc memory and the GPU
might trigger a page fault if there is no page backing the buffer addresses yet.
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
|
|
old way is use device specific memory allocator
new way is to use malloc (here we have a limitation due to nouveau
the address must be in below 1 << 40 hence why the custom malloc)
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
|
|
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
|
|
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
|
|
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
|
|
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
|
|
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
|
|
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
|