[gcc(refs/vendors/ARM/heads/arm-struct-reorg-wip)] internal-fn: Avoid dropping the lhs of some calls [PR94941]
Tamar Christina
tnfchris@gcc.gnu.org
Fri Jul 17 14:56:15 GMT 2020
https://gcc.gnu.org/g:688c8da3eb4cc7482f9844e0e87c11f5003bfbef
commit 688c8da3eb4cc7482f9844e0e87c11f5003bfbef
Author: Richard Sandiford <richard.sandiford@arm.com>
Date: Mon May 4 21:22:16 2020 +0100
internal-fn: Avoid dropping the lhs of some calls [PR94941]
create_output_operand coerces an output operand to the insn's
predicates, using a suggested rtx location if convenient.
But if that rtx location is actually required rather than
optional, the builder of the insn has to emit a move afterwards.
(We could instead add a new interface that does this automatically,
but that's future work.)
This PR shows that we were failing to emit the move for some of the
vector load internal functions. I think there are other routines in
internal-fn.c that potentially have the same problem, but this patch is
supposed to be a conservative subset suitable for backporting to GCC 10.
2020-05-04 Richard Sandiford <richard.sandiford@arm.com>
gcc/
PR middle-end/94941
* internal-fn.c (expand_load_lanes_optab_fn): Emit a move if the
chosen lhs is different from the gcall lhs.
(expand_mask_load_optab_fn): Likewise.
(expand_gather_load_optab_fn): Likewise.
gcc/testsuite/
PR middle-end/94941
* gcc.target/aarch64/sve/acle/general/unoptimized_1.c: New test.
Diff:
---
gcc/ChangeLog | 8 ++++++++
gcc/internal-fn.c | 6 ++++++
gcc/testsuite/ChangeLog | 5 +++++
.../aarch64/sve/acle/general/unoptimized_1.c | 21 +++++++++++++++++++++
4 files changed, 40 insertions(+)
diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index f25256d189f..2e4651174d2 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,11 @@
+2020-05-04 Richard Sandiford <richard.sandiford@arm.com>
+
+ PR middle-end/94941
+ * internal-fn.c (expand_load_lanes_optab_fn): Emit a move if the
+ chosen lhs is different from the gcall lhs.
+ (expand_mask_load_optab_fn): Likewise.
+ (expand_gather_load_optab_fn): Likewise.
+
2020-05-04 Marek Polacek <polacek@redhat.com>
Revert:
diff --git a/gcc/internal-fn.c b/gcc/internal-fn.c
index 52d1638917a..5e9aa60721e 100644
--- a/gcc/internal-fn.c
+++ b/gcc/internal-fn.c
@@ -167,6 +167,8 @@ expand_load_lanes_optab_fn (internal_fn, gcall *stmt, convert_optab optab)
create_output_operand (&ops[0], target, TYPE_MODE (type));
create_fixed_operand (&ops[1], mem);
expand_insn (get_multi_vector_move (type, optab), 2, ops);
+ if (!rtx_equal_p (target, ops[0].value))
+ emit_move_insn (target, ops[0].value);
}
/* Expand STORE_LANES call STMT using optab OPTAB. */
@@ -2507,6 +2509,8 @@ expand_mask_load_optab_fn (internal_fn, gcall *stmt, convert_optab optab)
create_fixed_operand (&ops[1], mem);
create_input_operand (&ops[2], mask, TYPE_MODE (TREE_TYPE (maskt)));
expand_insn (icode, 3, ops);
+ if (!rtx_equal_p (target, ops[0].value))
+ emit_move_insn (target, ops[0].value);
}
#define expand_mask_load_lanes_optab_fn expand_mask_load_optab_fn
@@ -2827,6 +2831,8 @@ expand_gather_load_optab_fn (internal_fn, gcall *stmt, direct_optab optab)
insn_code icode = convert_optab_handler (optab, TYPE_MODE (TREE_TYPE (lhs)),
TYPE_MODE (TREE_TYPE (offset)));
expand_insn (icode, i, ops);
+ if (!rtx_equal_p (lhs_rtx, ops[0].value))
+ emit_move_insn (lhs_rtx, ops[0].value);
}
/* Expand DIVMOD() using:
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index c6025647734..9109924e2e4 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2020-05-04 Richard Sandiford <richard.sandiford@arm.com>
+
+ PR middle-end/94941
+ * gcc.target/aarch64/sve/acle/general/unoptimized_1.c: New test.
+
2020-04-30 Martin Jambor <mjambor@suse.cz>
PR ipa/94856
diff --git a/gcc/testsuite/gcc.target/aarch64/sve/acle/general/unoptimized_1.c b/gcc/testsuite/gcc.target/aarch64/sve/acle/general/unoptimized_1.c
new file mode 100644
index 00000000000..18d73e21a83
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/sve/acle/general/unoptimized_1.c
@@ -0,0 +1,21 @@
+/* { dg-do run { target aarch64_sve_hw } } */
+
+#include <arm_sve.h>
+
+svfloat32_t
+foo (float *ptr)
+{
+ svbool_t pg = svptrue_pat_b32 (SV_VL1);
+ svfloat32_t res = svld1 (pg, ptr);
+ return res;
+}
+
+int
+main (void)
+{
+ svbool_t pg = svptrue_pat_b32 (SV_VL1);
+ float x[1] = { 1 };
+ if (svptest_any (pg, svcmpne (pg, foo (x), 1.0)))
+ __builtin_abort ();
+ return 0;
+}
More information about the Gcc-cvs
mailing list