rusticl/kernel: fix clGetKernelSuggestedLocalWorkSizeKHR implementation

There were two issues:
1. The global_work_offset parameter is optional but we errored on NULL
2. We didn't return the reqd_work_group_size when set on the kernel.

Fixes: 376d1e6667 ("rusticl: implement cl_khr_suggested_local_work_size")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38375>

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

@@ -857,10 +857,6 @@ fn get_kernel_suggested_local_work_size_khr(
         return Err(CL_INVALID_GLOBAL_WORK_SIZE);
     }
 
-    if global_work_offset.is_null() {
-        return Err(CL_INVALID_GLOBAL_OFFSET);
-    }
-
     // CL_INVALID_VALUE if suggested_local_work_size is NULL.
     if suggested_local_work_size.is_null() {
         return Err(CL_INVALID_VALUE);

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

@@ -1440,6 +1440,15 @@ impl Kernel {
         grid: &mut [usize],
         block: &mut [usize],
     ) {
+        // We have to use the required workgroup size if specified.
+        if self.work_group_size() != [0; 3] {
+            for i in 0..work_dim {
+                block[i] = self.work_group_size()[i];
+                grid[i] /= block[i];
+            }
+            return;
+        }
+
         let mut threads = self.max_threads_per_block(d);
         let dim_threads = d.max_block_sizes();
         let subgroups = self.preferred_simd_size(d);