RISC-V: Fix testsuite regression due to recent IRA changes.
authorKito Cheng <kito.cheng@sifive.com>
Fri, 6 Mar 2020 08:30:48 +0000 (16:30 +0800)
committerKito Cheng <kito.cheng@sifive.com>
Fri, 6 Mar 2020 10:31:11 +0000 (18:31 +0800)
gcc/testsuite/ChangeLog
gcc/testsuite/gcc.target/riscv/pr93304.c

index 09d597307308ba160368b1d70e8a849f25a96783..6c9206aeb27a9fc8d852755517c5281c4be0fa0c 100644 (file)
@@ -1,3 +1,7 @@
+2020-03-06  Kito Cheng  <kito.cheng@sifive.com>
+
+       * gcc.target/riscv/pr93304.c: Update expected output and comment.
+
 2020-03-06  Delia Burduv  <delia.burduv@arm.com>
 
        * gcc.target/aarch64/advsimd-intrinsics/bfcvt-compile.c: New test.
index f771e4859a96f70cb8adf35e158835de730dcd31..248f205e0d2d7f1bf931933003aff1d7465bcd83 100644 (file)
@@ -13,7 +13,6 @@ foo (void)
 
 /* Register rename will try to use registers from the lower register
    regradless of the REG_ALLOC_ORDER.
-   In theory, t0-t6 should not used in such small program if regrename
-   not executed incorrectly, because a5-a0 has higher priority in
-   REG_ALLOC_ORDER.  */
-/* { dg-final { scan-assembler-not "t\[0-6\]" } } */
+   In theory, t2 should not used in such small program if regrename
+   not executed incorrectly, because t0-a2 should be enough.  */
+/* { dg-final { scan-assembler-not "t2" } } */