https://bugs.freedesktop.org/show_bug.cgi?id=88444
--- Comment #13 from evgeniy.tyu...@intel.com ---
By the way,
I've studied our log (attached cl_3880.log) more closely, and that's what I
found:
On line 283 we can see tmp0_0_0_SumIfs_reduction kernel is being queried for
arg info. SetKernelArg (uiIndex=4) on line 297, which is 5th argument (i.e.
tmp0_0_4), is being initialized with buffer handle 363061576. Above in the line
194 this handle can be seen returned from the CreateBuffer (line 185). The
debug output line 168 shows it was created with size 8.
Just a bit below (line 208) the same buffer with handle 363061576 is being set
as arg5 of DynamicKernel_nop_SumIfs kernel.
In other words, the buffer is created with size 8 for one kernel -
DynamicKernel_nop_SumIfs (but actually tmp0_0_4 is not used there). And then
reused in tmp0_0_0_SumIfs_reduction without re-allocating. All kernels seems to
be enqeued with { 256, 100 }, { 256, 1 }.
May be this will give you guys a hint.
Please correct me in case I'm mistaken.
--
You are receiving this mail because:
You are the assignee for the bug.
_______________________________________________
Libreoffice-bugs mailing list
Libreoffice-bugs@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice-bugs
