[PATCH] D103634: [BitcodeWriter][PowerPC] Avoid clearing lower bits for ppc_fp128 NullValues

Jinsong Ji via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu Jun 3 12:35:35 PDT 2021


jsji updated this revision to Diff 349648.
jsji added a comment.
Herald added a subscriber: dexonsmith.

Fix isNullValue for ppc_fp128.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D103634/new/

https://reviews.llvm.org/D103634

Files:
  llvm/lib/IR/Constants.cpp
  llvm/test/CodeGen/PowerPC/ppc_fp128-bcwriter.ll


Index: llvm/test/CodeGen/PowerPC/ppc_fp128-bcwriter.ll
===================================================================
--- llvm/test/CodeGen/PowerPC/ppc_fp128-bcwriter.ll
+++ llvm/test/CodeGen/PowerPC/ppc_fp128-bcwriter.ll
@@ -1,7 +1,7 @@
 ; RUN: llvm-as < %s -o - | llvm-dis - | FileCheck %s
 
 ;CHECK-LABEL: main
-;CHECK: store ppc_fp128 0xM0000000000000000000000000000000
+;CHECK: store ppc_fp128 0xM0000000000000000FFFFFFFFFFFFFFFF
 
 define i32 @main() local_unnamed_addr {
 _main_entry:
Index: llvm/lib/IR/Constants.cpp
===================================================================
--- llvm/lib/IR/Constants.cpp
+++ llvm/lib/IR/Constants.cpp
@@ -89,8 +89,13 @@
     return CI->isZero();
 
   // +0.0 is null.
-  if (const ConstantFP *CFP = dyn_cast<ConstantFP>(this))
+  if (const ConstantFP *CFP = dyn_cast<ConstantFP>(this)) {
+    // ppc_fp128 determine isZero using high order double only
+    // Should check the bitwise value to make sure all bits are zero.
+    if (CFP->getType()->isPPC_FP128Ty())
+      return CFP->isExactlyValue(+0.0);
     return CFP->isZero() && !CFP->isNegative();
+  }
 
   // constant zero is zero for aggregates, cpnull is null for pointers, none for
   // tokens.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D103634.349648.patch
Type: text/x-patch
Size: 1215 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210603/7b310e1e/attachment.bin>


More information about the llvm-commits mailing list