R600/SI: implement SI.load.const intrinsic

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177273 91177308-0d34-0410-b5e6-96231b3b80d8

Author: ChristianKoenigAMD

lib/Target/R600/SIInstructions.td

@@ -1403,6 +1403,18 @@ def : Pat <
   (V_CNDMASK_B32_e64 (i32 0), (i32 -1), SReg_64:$src0)
 >;
 
+// 1. Offset as 8bit DWORD immediate
+def : Pat <
+  (int_SI_load_const SReg_128:$sbase, IMM8bitDWORD:$offset),
+  (S_BUFFER_LOAD_DWORD_IMM SReg_128:$sbase, IMM8bitDWORD:$offset)
+>;
+
+// 2. Offset loaded in an 32bit SGPR
+def : Pat <
+  (int_SI_load_const SReg_128:$sbase, imm:$offset),
+  (S_BUFFER_LOAD_DWORD_SGPR SReg_128:$sbase, (S_MOV_B32 imm:$offset))
+>;
+
 /********** ================== **********/
 /**********   VOP3 Patterns    **********/
 /********** ================== **********/

lib/Target/R600/SIIntrinsics.td

@@ -16,8 +16,7 @@ let TargetPrefix = "SI", isTarget = 1 in {
 
   def int_SI_packf16 : Intrinsic <[llvm_i32_ty], [llvm_float_ty, llvm_float_ty], [IntrNoMem]>;
   def int_SI_export : Intrinsic <[], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_float_ty, llvm_float_ty, llvm_float_ty, llvm_float_ty], []>;
-  /* XXX: We may need a seperate intrinsic here for loading integer values */
-  def int_SI_load_const : Intrinsic <[llvm_float_ty], [llvm_i64_ty, llvm_i32_ty], []>;
+  def int_SI_load_const : Intrinsic <[llvm_float_ty], [llvm_v16i8_ty, llvm_i32_ty], [IntrReadMem]>;
   def int_SI_vs_load_input : Intrinsic <[llvm_v4f32_ty], [llvm_v16i8_ty, llvm_i16_ty, llvm_i32_ty], [IntrReadMem]> ;
   def int_SI_wqm : Intrinsic <[], [], []>;