gpu-compute: Added method to compute the actual workgroup size

This patch adds a method to the Wavefront class to compute the actual workgroup
size. This can be different from the maximum workgroup size specified when
launching the kernel through the NDRange object. Current solution is still not
optimal, as we are computing these for each wavefront and the dispatcher also
needs to have this information and can't actually call
Wavefront::computeActuallWgSz before the wavefronts are being created. A long
term solution would be to have a Workgroup class that deals with all these
details.

1 files changed, 6 insertions, 0 deletions

diff --git a/src/gpu-compute/wavefront.hh b/src/gpu-compute/wavefront.hh
index ef8c80989..0df8a6c82 100644
--- a/src/gpu-compute/wavefront.hh
+++ b/src/gpu-compute/wavefront.hh
@@ -47,6 +47,7 @@
 #include "gpu-compute/condition_register_state.hh"
 #include "gpu-compute/lds_state.hh"
 #include "gpu-compute/misc.hh"
+#include "gpu-compute/ndrange.hh"
 #include "params/Wavefront.hh"
 #include "sim/sim_object.hh"
 
@@ -189,11 +190,16 @@ class Wavefront : public SimObject
     std::vector<Addr> lastAddr;
     std::vector<uint32_t> workItemId[3];
     std::vector<uint32_t> workItemFlatId;
+    /* kernel launch parameters */
     uint32_t workGroupId[3];
     uint32_t workGroupSz[3];
     uint32_t gridSz[3];
     uint32_t wgId;
     uint32_t wgSz;
+    /* the actual WG size can differ than the maximum size */
+    uint32_t actualWgSz[3];
+    uint32_t actualWgSzTotal;
+    void computeActualWgSz(NDRange *ndr);
     // wavefront id within a workgroup
     uint32_t wfId;
     uint32_t maxDynWaveId;