From bbeac6ee9249e664a33343562d1e8f89b6185c07 Mon Sep 17 00:00:00 2001
From: Yang Rong <rong.r.yang@intel.com>
Date: Thu, 16 Oct 2014 15:10:35 +0800
Subject: BDW: Fix load/store half error.

BDW support HF data type, so use mov directly to convert between Half Float/Float.

Signed-off-by: Yang Rong <rong.r.yang@intel.com>
Reviewed-by: Zhigang Gong <zhigang.gong@linux.intel.com>
---
 backend/src/backend/gen8_encoder.cpp |  8 ++++++++
 backend/src/backend/gen8_encoder.hpp |  2 ++
 backend/src/backend/gen_defs.hpp     |  2 +-
 backend/src/backend/gen_encoder.cpp  | 10 ++++++++--
 backend/src/backend/gen_encoder.hpp  |  5 +++--
 5 files changed, 22 insertions(+), 5 deletions(-)

diff --git a/backend/src/backend/gen8_encoder.cpp b/backend/src/backend/gen8_encoder.cpp
index 4ca900de..749bc521 100644
--- a/backend/src/backend/gen8_encoder.cpp
+++ b/backend/src/backend/gen8_encoder.cpp
@@ -96,6 +96,14 @@ namespace gbe
     gen8_insn->bits3.gen7_typed_rw.slot = 1;
   }
 
+  void Gen8Encoder::F16TO32(GenRegister dest, GenRegister src0) {
+    MOV(GenRegister::retype(dest, GEN_TYPE_F), GenRegister::retype(src0, GEN_TYPE_HF));
+  }
+
+  void Gen8Encoder::F32TO16(GenRegister dest, GenRegister src0) {
+    MOV(GenRegister::retype(dest, GEN_TYPE_HF), GenRegister::retype(src0, GEN_TYPE_F));
+  }
+
   void Gen8Encoder::ATOMIC(GenRegister dst, uint32_t function, GenRegister src, uint32_t bti, uint32_t srcNum) {
     GenNativeInstruction *insn = this->next(GEN_OPCODE_SEND);
     Gen8NativeInstruction *gen8_insn = &insn->gen8_insn;
diff --git a/backend/src/backend/gen8_encoder.hpp b/backend/src/backend/gen8_encoder.hpp
index 6ca3b416..970258be 100644
--- a/backend/src/backend/gen8_encoder.hpp
+++ b/backend/src/backend/gen8_encoder.hpp
@@ -44,6 +44,8 @@ namespace gbe
     virtual void patchJMPI(uint32_t insnID, int32_t jip, int32_t uip);
     /*! Get double/long exec width */
     virtual int getDoubleExecWidth(void) { return GEN8_DOUBLE_EXEC_WIDTH; }
+    virtual void F16TO32(GenRegister dest, GenRegister src0);
+    virtual void F32TO16(GenRegister dest, GenRegister src0);
     virtual void MOV_DF(GenRegister dest, GenRegister src0, GenRegister tmp = GenRegister::null());
     virtual void LOAD_DF_IMM(GenRegister dest, GenRegister tmp, double value);
     virtual void ATOMIC(GenRegister dst, uint32_t function, GenRegister src, uint32_t bti, uint32_t srcNum);
diff --git a/backend/src/backend/gen_defs.hpp b/backend/src/backend/gen_defs.hpp
index 385fdfaf..3773ed98 100644
--- a/backend/src/backend/gen_defs.hpp
+++ b/backend/src/backend/gen_defs.hpp
@@ -246,12 +246,12 @@ enum GenMessageTarget {
 #define GEN_TYPE_UB  4
 #define GEN_TYPE_B   5
 #define GEN_TYPE_VF  5 /* packed float vector, immediates only? */
-#define GEN_TYPE_HF  6
 #define GEN_TYPE_V   6 /* packed int vector, immediates only, uword dest only */
 #define GEN_TYPE_DF  6
 #define GEN_TYPE_F   7
 #define GEN_TYPE_UL  8
 #define GEN_TYPE_L   9
+#define GEN_TYPE_HF  10
 
 #define GEN_ARF_NULL                  0x00
 #define GEN_ARF_ADDRESS               0x10
diff --git a/backend/src/backend/gen_encoder.cpp b/backend/src/backend/gen_encoder.cpp
index 847ab7ba..1ae4cc5d 100644
--- a/backend/src/backend/gen_encoder.cpp
+++ b/backend/src/backend/gen_encoder.cpp
@@ -678,6 +678,14 @@ namespace gbe
     }
   }
 
+  void GenEncoder::F16TO32(GenRegister dest, GenRegister src0) {
+    alu1(this, GEN_OPCODE_F16TO32, dest, src0);
+  }
+
+  void GenEncoder::F32TO16(GenRegister dest, GenRegister src0) {
+    alu1(this, GEN_OPCODE_F32TO16, dest, src0);
+  }
+
   ALU1(MOV)
   ALU1(RNDZ)
   ALU1(RNDE)
@@ -686,8 +694,6 @@ namespace gbe
   ALU1(FBH)
   ALU1(FBL)
   ALU1(CBIT)
-  ALU1(F16TO32)
-  ALU1(F32TO16)
   ALU2(SEL)
   ALU1(NOT)
   ALU2_MOD(AND)
diff --git a/backend/src/backend/gen_encoder.hpp b/backend/src/backend/gen_encoder.hpp
index 3df7a57d..d4692e1b 100644
--- a/backend/src/backend/gen_encoder.hpp
+++ b/backend/src/backend/gen_encoder.hpp
@@ -107,8 +107,6 @@ namespace gbe
     ALU1(RNDE)
     ALU1(RNDD)
     ALU1(RNDU)
-    ALU1(F16TO32)
-    ALU1(F32TO16)
     ALU2(SEL)
     ALU1(NOT)
     ALU2_MOD(AND)
@@ -136,6 +134,9 @@ namespace gbe
 #undef ALU2
 #undef ALU2_MOD
 #undef ALU3
+
+    virtual void F16TO32(GenRegister dest, GenRegister src0);
+    virtual void F32TO16(GenRegister dest, GenRegister src0);
     /*! Get double/long exec width */
     virtual int getDoubleExecWidth(void) = 0;
     virtual void MOV_DF(GenRegister dest, GenRegister src0, GenRegister tmp = GenRegister::null());
-- 
cgit v1.2.3