[Beignet] [PATCH 2/2] GBE: optimize scalar data type conversion.

[Zhigang Gong]

If the dst is scalar, the register region restrication is relaxed.
we can save one instruction as below:

    (12      )  mov.sat(1)      g127.24<4>:B    g1.3<0,1,0>:D              { align1 WE_all };
    (14      )  mov(1)          g127.28<1>:B    g127.24<0,1,4>:D           { align1 WE_all };

Optimized to:

    (12      )  mov.sat(1)      g128.28<4>:B    g1.3<0,1,0>:D              { align1 WE_all };

No need to create a temporary register g127.24.

Signed-off-by: Zhigang Gong <zhigang.gong at intel.com>
---
 backend/src/backend/gen_insn_selection.cpp | 17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/backend/src/backend/gen_insn_selection.cpp b/backend/src/backend/gen_insn_selection.cpp
index 19921d4..db20162 100644
--- a/backend/src/backend/gen_insn_selection.cpp
+++ b/backend/src/backend/gen_insn_selection.cpp
@@ -3175,12 +3175,18 @@ namespace gbe
         GenRegister unpacked;
         if (dstFamily == FAMILY_WORD) {
           const uint32_t type = dstType == TYPE_U16 ? GEN_TYPE_UW : GEN_TYPE_W;
-          unpacked = sel.unpacked_uw(sel.reg(FAMILY_DWORD, sel.isScalarReg(insn.getSrc(0))));
-          unpacked = GenRegister::retype(unpacked, type);
+          if (!sel.isScalarReg(dst.reg())) {
+            unpacked = sel.unpacked_uw(sel.reg(FAMILY_DWORD, sel.isScalarReg(insn.getSrc(0))));
+            unpacked = GenRegister::retype(unpacked, type);
+          } else
+            unpacked = GenRegister::retype(sel.unpacked_uw(dst.reg()), type);
         } else {
           const uint32_t type = dstType == TYPE_U8 ? GEN_TYPE_UB : GEN_TYPE_B;
-          unpacked = sel.unpacked_ub(sel.reg(FAMILY_DWORD, sel.isScalarReg(insn.getSrc(0))));
-          unpacked = GenRegister::retype(unpacked, type);
+          if (!sel.isScalarReg(dst.reg())) {
+            unpacked = sel.unpacked_ub(sel.reg(FAMILY_DWORD, sel.isScalarReg(insn.getSrc(0))));
+            unpacked = GenRegister::retype(unpacked, type);
+          } else
+            unpacked = GenRegister::retype(sel.unpacked_ub(dst.reg()), type);
         }
         if(srcFamily == FAMILY_QWORD) {
           GenRegister tmp = sel.selReg(sel.reg(FAMILY_DWORD));
@@ -3197,7 +3203,8 @@ namespace gbe
             sel.MOV(unpacked, src);
           sel.pop();
         }
-        sel.MOV(dst, unpacked);
+        if (unpacked.reg() != dst.reg())
+          sel.MOV(dst, unpacked);
       } else if ((dstType == ir::TYPE_S32 || dstType == ir::TYPE_U32) && srcFamily == FAMILY_QWORD) {
         sel.CONVI64_TO_I(dst, src);
       } else if (dstType == ir::TYPE_FLOAT && srcFamily == FAMILY_QWORD) {
-- 
1.8.3.2