[Beignet] [PATCH V2 07/17] Backend: Fix workgroup op masked elements by proper init value

[Grigore Lupescu]

From: Grigore Lupescu <grigore.lupescu at intel.com>

Signed-off-by: Grigore Lupescu <grigore.lupescu at intel.com>
---
 backend/src/backend/gen_context.cpp | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)

diff --git a/backend/src/backend/gen_context.cpp b/backend/src/backend/gen_context.cpp
index cac78a9..38bd0f2 100644
--- a/backend/src/backend/gen_context.cpp
+++ b/backend/src/backend/gen_context.cpp
@@ -2395,8 +2395,7 @@ namespace gbe
     if (wg_op == ir::WORKGROUP_OP_ALL)
     {
       if (dataReg.type == GEN_TYPE_D
-          || dataReg.type == GEN_TYPE_UD
-          || dataReg.type == GEN_TYPE_F)
+          || dataReg.type == GEN_TYPE_UD)
         p->MOV(dataReg, GenRegister::immd(0xFFFFFFFF));
       else if(dataReg.type == GEN_TYPE_L ||
           dataReg.type == GEN_TYPE_UL)
@@ -2619,8 +2618,8 @@ namespace gbe
  */
   void GenContext::emitWorkGroupOpInstruction(const SelectionInstruction &insn){
     const GenRegister dst = ra->genReg(insn.dst(0));
-    const GenRegister tmp = ra->genReg(insn.dst(1));
-    const GenRegister theVal = ra->genReg(insn.src(2));
+    const GenRegister tmp = GenRegister::retype(ra->genReg(insn.dst(1)), dst.type);
+    const GenRegister theVal = GenRegister::retype(ra->genReg(insn.src(2)), dst.type);
     GenRegister threadData = ra->genReg(insn.src(3));
     GenRegister partialData = GenRegister::toUniform(threadData, dst.type);
     GenRegister threadId = ra->genReg(insn.src(0));
@@ -2631,6 +2630,16 @@ namespace gbe
     uint32_t simd = p->curr.execWidth;
     int32_t jip0, jip1;
 
+    /* Masked elements should be properly set to init value */
+    p->push(); {
+      p->curr.noMask = 1;
+      initValue(p, tmp, wg_op);
+      p->curr.noMask = 0;
+      p->MOV(tmp, theVal);
+      p->curr.noMask = 1;
+      p->MOV(theVal, tmp);
+    } p->pop();
+
     threadId = GenRegister::toUniform(threadId, GEN_TYPE_UD);
 
     /* Use of continuous GRF allocation from insn selection */
-- 
2.5.0