clover, gallium: add PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK

This is not necessarily the product of MAX_BLOCK_SIZE[i].

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>

src/gallium/docs/source/screen.rst

@@ -212,6 +212,11 @@ pipe_screen::get_compute_param.
   units.  Value type: ``uint64_t []``.
 * ``PIPE_COMPUTE_CAP_MAX_BLOCK_SIZE``: Maximum block size in thread
   units.  Value type: ``uint64_t []``.
+* ``PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK``: Maximum number of threads that
+  a single block can contain.  Value type: ``uint64_t``.
+  This may be less than the product of the components of MAX_BLOCK_SIZE and is
+  usually limited by the number of threads that can be resident simultaneously
+  on a compute unit.
 * ``PIPE_COMPUTE_CAP_MAX_GLOBAL_SIZE``: Maximum size of the GLOBAL
   resource.  Value type: ``uint64_t``.
 * ``PIPE_COMPUTE_CAP_MAX_LOCAL_SIZE``: Maximum size of the LOCAL

src/gallium/include/pipe/p_defines.h

@@ -548,6 +548,7 @@ enum pipe_compute_cap
    PIPE_COMPUTE_CAP_GRID_DIMENSION,
    PIPE_COMPUTE_CAP_MAX_GRID_SIZE,
    PIPE_COMPUTE_CAP_MAX_BLOCK_SIZE,
+   PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK,
    PIPE_COMPUTE_CAP_MAX_GLOBAL_SIZE,
    PIPE_COMPUTE_CAP_MAX_LOCAL_SIZE,
    PIPE_COMPUTE_CAP_MAX_PRIVATE_SIZE,

src/gallium/state_trackers/clover/api/device.cpp

@@ -87,7 +87,8 @@ clGetDeviceInfo(cl_device_id dev, cl_device_info param,
                                      dev->max_block_size());
 
    case CL_DEVICE_MAX_WORK_GROUP_SIZE:
-      return scalar_property<size_t>(buf, size, size_ret, SIZE_MAX);
+      return scalar_property<size_t>(buf, size, size_ret,
+                                     dev->max_threads_per_block());
 
    case CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR:
       return scalar_property<cl_uint>(buf, size, size_ret, 16);

src/gallium/state_trackers/clover/core/device.cpp

@@ -138,6 +138,12 @@ _cl_device_id::max_const_buffers() const {
                                  PIPE_SHADER_CAP_MAX_CONST_BUFFERS);
 }
 
+size_t
+_cl_device_id::max_threads_per_block() const {
+   return get_compute_param<uint64_t>(
+      pipe, PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK)[0];
+}
+
 std::vector<size_t>
 _cl_device_id::max_block_size() const {
    return get_compute_param<uint64_t>(pipe, PIPE_COMPUTE_CAP_MAX_BLOCK_SIZE);

src/gallium/state_trackers/clover/core/device.hpp

@@ -55,6 +55,7 @@ public:
    cl_ulong max_mem_input() const;
    cl_ulong max_const_buffer_size() const;
    cl_uint max_const_buffers() const;
+   size_t max_threads_per_block() const;
 
    std::vector<size_t> max_block_size() const;
    std::string device_name() const;