Add support for non-decomposed complex operations

Added xbit 201 0x4000000 in order to control alignment assumption for complex arrays. When enabled, it is assumed that single complex arrays are 8-byte aligned and that double complex arrays are 16-byte aligned.
flang-compiler · Sep 3, 2019 · 7a7aae1 · 7a7aae1
1 parent 5625686
commit 7a7aae1
Show file tree

Hide file tree

Showing 3 changed files with 34 additions and 1 deletion.
diff --git a/tools/flang2/docs/xflag.n b/tools/flang2/docs/xflag.n
@@ -5282,6 +5282,8 @@ but the innermost loop ones.
 .XB 0x1000000:
 Only find ACIV induction variables for innermost loops.
 reserved
+.XB 0x2000000:
+Assume that complex arrays on GPU are aligned as follows: complex:8-byte dcmplx:16-byte
 
 .XF "202:"
 Set number of bigbuffers for multi-buffer memory management for AMD GPU.

diff --git a/tools/flang2/flang2exe/ll_structure.h b/tools/flang2/flang2exe/ll_structure.h
@@ -40,7 +40,8 @@ typedef enum LL_Op {
   LL_ICMP,     LL_FCMP,        LL_BR,       LL_UBR,          LL_SELECT,
   LL_GEP,      LL_BITCAST,     LL_INTTOPTR, LL_PTRTOINT,     LL_ALLOCA,
   LL_TEXTCALL, LL_UNREACHABLE, LL_SWITCH,   LL_EXTRACTVALUE, LL_INSERTVALUE,
-  LL_ATOMICRMW, LL_CMPXCHG,    LL_ASM,	LL_NONE
+  LL_ATOMICRMW, LL_CMPXCHG,    LL_ASM,	    LL_EXTRACTELEM,  LL_INSERTELEM,
+  LL_FNEG,     LL_NONE
 } LL_Op;
 
 /* clang-format on */

diff --git a/tools/flang2/flang2exe/ll_write.cpp b/tools/flang2/flang2exe/ll_write.cpp
@@ -185,6 +185,14 @@ get_op_name(enum LL_Op op)
     return "cmpxchg";
   case LL_EXTRACTVALUE:
     return "extractvalue";
+  case LL_INSERTVALUE:
+    return "insertvalue";
+  case LL_EXTRACTELEM:
+    return "extractelement";
+  case LL_INSERTELEM:
+    return "insertelement";
+  case LL_FNEG:
+    return "fneg";
   default:
     return "thisisnotacceptable";
   }
@@ -450,6 +458,23 @@ ll_write_instruction(FILE *out, LL_Instruction *inst, LL_Module *module, int no_
             inst->operands[1]->type_struct->str, inst->operands[1]->data,
             inst->operands[2]->data);
   } break;
+  case LL_INSERTVALUE: {
+    fprintf(out, "%s%s = %s %s %s, %s %s, %s", SPACES, inst->operands[0]->data, opname,
+            inst->operands[1]->type_struct->str, inst->operands[1]->data,
+            inst->operands[2]->type_struct->str, inst->operands[2]->data,
+            inst->operands[3]->data);
+  } break;
+  case LL_EXTRACTELEM: {
+    fprintf(out, "%s%s = %s %s %s, %s %s", SPACES, inst->operands[0]->data, opname,
+            inst->operands[1]->type_struct->str, inst->operands[1]->data,
+            inst->operands[2]->type_struct->str, inst->operands[2]->data);
+  } break;
+  case LL_INSERTELEM: {
+    fprintf(out, "%s%s = %s %s %s, %s %s, %s %s", SPACES, inst->operands[0]->data, opname,
+            inst->operands[1]->type_struct->str, inst->operands[1]->data,
+            inst->operands[2]->type_struct->str, inst->operands[2]->data,
+            inst->operands[3]->type_struct->str, inst->operands[3]->data);
+  } break;
   case LL_ADD:
   case LL_FADD:
   case LL_SUB:
@@ -472,6 +497,11 @@ ll_write_instruction(FILE *out, LL_Instruction *inst, LL_Module *module, int no_
             inst->operands[1]->type_struct->str, inst->operands[1]->data,
             inst->operands[2]->data);
     break;
+  case LL_FNEG:
+    /* unary ops */
+    fprintf(out, "%s%s = %s %s %s", SPACES, inst->operands[0]->data, opname,
+            inst->operands[1]->type_struct->str, inst->operands[1]->data);
+    break;
   case LL_STORE:
     render_store(out, inst);
     break;