Andrew Pinski wrote:
>
> Note this caused a few testsuite failures:
Confirmed. The diffs show the new sequence is always better. I've
committed this:
Update vmov_n_1.c now we are generating better code for dup:
ldr s0, [x0]
dup v0.2s, v0.s[0]
ret
Instead of:
ldr w0, [x0]
dup v0.2s, w0
ret
ChangeLog:
2017-09-13 Wilco Dijkstra <wdijk...@arm.com>
* gcc.target/aarch64/vmov_n_1.c: Update dup scan-assembler.
--
diff --git a/gcc/testsuite/gcc.target/aarch64/vmov_n_1.c
b/gcc/testsuite/gcc.target/aarch64/vmov_n_1.c
index
485a1a970bbcd30cb45a2f69bbd9f62f8258d3df..d0c284296a5e256be5cf5f859b113cdd62f929ba
100644
--- a/gcc/testsuite/gcc.target/aarch64/vmov_n_1.c
+++ b/gcc/testsuite/gcc.target/aarch64/vmov_n_1.c
@@ -190,8 +190,9 @@ TESTFUNC_NAME (reg_len, data_type, data_len) () \
OBSCURE_FUNC (64, 32, f)
TESTFUNC (64, 32, f)
-/* "dup Vd.2s, Rn" is less preferable then "dup Vd.2s, Vn.s[lane]". */
-/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.2s,
v\[0-9\]+\.s\\\[\[0-9\]+\\\]" 1 } } */
+/* "dup Vd.2s, Rn" is less preferable than "dup Vd.2s, Vn.s[lane]". */
+/* { dg-final { scan-assembler-not "dup\\tv\[0-9\]+\.2s, w\[0-9\]+" } } */
+/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.2s,
v\[0-9\]+\.s\\\[\[0-9\]+\\\]" 3 } } */
OBSCURE_FUNC (64, 64, f)
TESTFUNC (64, 64, f)
@@ -216,7 +217,9 @@ TESTFUNC (64, 16, s)
OBSCURE_FUNC (64, 32, s)
TESTFUNC (64, 32, s)
-/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.2s, w\[0-9\]+" 2 } } */
+/* "dup Vd.2s, Rn" is less preferable than "dup Vd.2s, Vn.s[lane]". */
+/* { dg-final { scan-assembler-not "dup\\tv\[0-9\]+\.2s, w\[0-9\]+" } } */
+/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.2s,
v\[0-9\]+\.s\\\[\[0-9\]+\\\]" 3 } } */
OBSCURE_FUNC (64, 64, s)
TESTFUNC (64, 64, s)
@@ -242,13 +245,15 @@ TESTFUNC (64, 64, u)
OBSCURE_FUNC (128, 32, f)
TESTFUNC (128, 32, f)
-/* "dup Vd.4s, Rn" is less preferable then "dup Vd.4s, Vn.s[lane]". */
-/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.4s,
v\[0-9\]+\.s\\\[\[0-9\]+\\\]" 1 } } */
+/* "dup Vd.4s, Rn" is less preferable than "dup Vd.4s, Vn.s[lane]". */
+/* { dg-final { scan-assembler-not "dup\\tv\[0-9\]+\.4s, w\[0-9\]+" } } */
+/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.4s,
v\[0-9\]+\.s\\\[\[0-9\]+\\\]" 3 } } */
OBSCURE_FUNC (128, 64, f)
TESTFUNC (128, 64, f)
-/* "dup Vd.2d, Rn" is less preferable then "dup Vd.2d, Vn.d[lane]". */
-/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.2d,
v\[0-9\]+\.d\\\[\[0-9\]+\\\]" 1 } } */
+/* "dup Vd.2d, Rn" is less preferable than "dup Vd.2d, Vn.d[lane]". */
+/* { dg-final { scan-assembler-not "dup\\tv\[0-9\]+\.2d, x\[0-9\]+" } } */
+/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.2d,
v\[0-9\]+\.d\\\[\[0-9\]+\\\]" 3 } } */
OBSCURE_FUNC (128, 8, p)
TESTFUNC (128, 8, p)
@@ -268,11 +273,15 @@ TESTFUNC (128, 16, s)
OBSCURE_FUNC (128, 32, s)
TESTFUNC (128, 32, s)
-/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.4s, w\[0-9\]+" 2 } } */
+/* "dup Vd.4s, Rn" is less preferable than "dup Vd.4s, Vn.s[lane]". */
+/* { dg-final { scan-assembler-not "dup\\tv\[0-9\]+\.4s, w\[0-9\]+" } } */
+/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.4s,
v\[0-9\]+\.s\\\[\[0-9\]+\\\]" 3 } } */
OBSCURE_FUNC (128, 64, s)
TESTFUNC (128, 64, s)
-/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.2d, x\[0-9\]+" 2 } } */
+/* "dup Vd.2d, Rn" is less preferable than "dup Vd.2d, Vn.d[lane]". */
+/* { dg-final { scan-assembler-not "dup\\tv\[0-9\]+\.2d, x\[0-9\]+" } } */
+/* { dg-final { scan-assembler-times "dup\\tv\[0-9\]+\.2d,
v\[0-9\]+\.d\\\[\[0-9\]+\\\]" 3 } } */
OBSCURE_FUNC (128, 8, u)
TESTFUNC (128, 8, u)
