rusticl: Fix work group size validation

For each dimension, we `threads *= lws`.. which is still zero if threads
is initialized to zero.

Fixes: eca4f0f632 ("rusticl/kernel: check that local size on dispatch doesn't exceed limits")
Signed-off-by: Rob Clark <rob.clark@oss.qualcomm.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35864>

src/amd/ci/radeonsi-raven-fails.txt

@@ -68,6 +68,10 @@ program@execute@global-offset,Fail
 program@execute@global-offset@3d- input dependent,Fail
 program@execute@vector-conversion,Fail
 
+# piglit bug, pigligt checks the global limits, but radeonsi advertises
+# only half the threads if the work group size is variable...
+program@run kernel with max work item sizes,Fail
+
 # uprev Piglit in Mesa
 spec@ext_external_objects@vk-semaphores,Crash
 spec@ext_external_objects@vk-semaphores-2,Crash

src/gallium/frontends/rusticl/api/kernel.rs

@@ -694,7 +694,7 @@ fn enqueue_ndrange_kernel(
     let device_bits = q.device.address_bits();
     let device_max = u64::MAX >> (u64::BITS - device_bits);
 
-    let mut threads = 0;
+    let mut threads = 1;
     for i in 0..work_dim as usize {
         let lws = local_work_size[i];
         let gws = global_work_size[i];