Mesa (main): clover: fix api zero sized enqueue

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: 3106a340a01cf4a760d828e38d560ed6044e100a
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=3106a340a01cf4a760d828e38d560ed6044e100a

Author: Dave Airlie <airlied at redhat.com>
Date:   Tue Oct 13 14:17:00 2020 +1000

clover: fix api zero sized enqueue

Reviewed-by: Karol Herbst <kherbst at redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12225>

---

 src/gallium/frontends/clover/api/kernel.cpp  | 3 ---
 src/gallium/frontends/clover/core/kernel.cpp | 7 +++++++
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/src/gallium/frontends/clover/api/kernel.cpp b/src/gallium/frontends/clover/api/kernel.cpp
index d172750481e..c937642fccf 100644
--- a/src/gallium/frontends/clover/api/kernel.cpp
+++ b/src/gallium/frontends/clover/api/kernel.cpp
@@ -271,9 +271,6 @@ namespace {
       if (dims < 1 || dims > q.device().max_block_size().size())
          throw error(CL_INVALID_WORK_DIMENSION);
 
-      if (!d_grid_size || any_of(is_zero(), grid_size))
-         throw error(CL_INVALID_GLOBAL_WORK_SIZE);
-
       return grid_size;
    }
 
diff --git a/src/gallium/frontends/clover/core/kernel.cpp b/src/gallium/frontends/clover/core/kernel.cpp
index 5d8b0452563..894b3bfffe2 100644
--- a/src/gallium/frontends/clover/core/kernel.cpp
+++ b/src/gallium/frontends/clover/core/kernel.cpp
@@ -67,6 +67,10 @@ kernel::launch(command_queue &q,
    const auto m = program().build(q.device()).binary;
    const auto reduced_grid_size =
       map(divides(), grid_size, block_size);
+
+   if (any_of(is_zero(), grid_size))
+      return;
+
    void *st = exec.bind(&q, grid_offset);
    struct pipe_grid_info info = {};
 
@@ -137,6 +141,9 @@ kernel::name() const {
 std::vector<size_t>
 kernel::optimal_block_size(const command_queue &q,
                            const std::vector<size_t> &grid_size) const {
+   if (any_of(is_zero(), grid_size))
+      return grid_size;
+
    return factor::find_grid_optimal_factor<size_t>(
       q.device().max_threads_per_block(), q.device().max_block_size(),
       grid_size);