AMDGPU/GlobalISel: Start matching s_lshlN_add_u32 instructions

Use a hack to only enable this for GlobalISel.

Technically this also works with SelectionDAG, but the divergence
selection isn't reliable enough and a few cases fail, but I have no
desire to spend time writing the manual expansion code for it. The DAG
actually does a better job since it catches using v_add_lshl_u32 in
the mixed SGPR/VGPR cases.

Author: arsenm

llvm/lib/Target/AMDGPU/SIInstrInfo.td

@@ -630,6 +630,16 @@ def add_ctpop : PatFrag <
   (add (ctpop $src0), $src1)
 >;
 
+foreach I = 1-4 in {
+def shl#I#_add : PatFrag <
+  (ops node:$src0, node:$src1),
+  (add (shl_oneuse $src0, (i32 I)), $src1)> {
+  // FIXME: Poor substitute for disabling pattern in SelectionDAG
+  let PredicateCode = [{return false;}];
+  let GISelPredicateCode = [{return true;}];
+}
+}
+
 multiclass SIAtomicM0Glue2 <string op_name, bit is_amdgpu = 0,
                             SDTypeProfile tc = SDTAtomic2,
                             bit IsInt = 1> {

llvm/lib/Target/AMDGPU/SOPInstructions.td

@@ -613,10 +613,18 @@ let SubtargetPredicate = isGFX9Plus in {
   def S_PACK_HH_B32_B16 : SOP2_32<"s_pack_hh_b32_b16">;
 
   let Defs = [SCC] in {
-    def S_LSHL1_ADD_U32 : SOP2_32<"s_lshl1_add_u32">;
-    def S_LSHL2_ADD_U32 : SOP2_32<"s_lshl2_add_u32">;
-    def S_LSHL3_ADD_U32 : SOP2_32<"s_lshl3_add_u32">;
-    def S_LSHL4_ADD_U32 : SOP2_32<"s_lshl4_add_u32">;
+    def S_LSHL1_ADD_U32 : SOP2_32<"s_lshl1_add_u32",
+      [(set i32:$sdst, (shl1_add SSrc_b32:$src0, SSrc_b32:$src1))]
+    >;
+    def S_LSHL2_ADD_U32 : SOP2_32<"s_lshl2_add_u32",
+      [(set i32:$sdst, (shl2_add SSrc_b32:$src0, SSrc_b32:$src1))]
+    >;
+    def S_LSHL3_ADD_U32 : SOP2_32<"s_lshl3_add_u32",
+      [(set i32:$sdst, (shl3_add SSrc_b32:$src0, SSrc_b32:$src1))]
+    >;
+    def S_LSHL4_ADD_U32 : SOP2_32<"s_lshl4_add_u32",
+      [(set i32:$sdst, (shl4_add SSrc_b32:$src0, SSrc_b32:$src1))]
+    >;
   } // End Defs = [SCC]
 
   let isCommutable = 1 in {