This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PING^2] [PATCH][5a/5] Postpone expanding va_arg until pass_stdarg
- From: Tom de Vries <Tom_deVries at mentor dot com>
- To: Richard Biener <rguenther at suse dot de>, Alan Lawrence <alan dot lawrence at arm dot com>
- Cc: Jakub Jelinek <jakub at redhat dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>, Michael Matz <matz at suse dot de>
- Date: Tue, 9 Jun 2015 14:56:51 +0200
- Subject: Re: [PING^2] [PATCH][5a/5] Postpone expanding va_arg until pass_stdarg
- Authentication-results: sourceware.org; auth=none
- References: <54E5BB06 dot 2080102 at mentor dot com> <54E5D0D2 dot 8080306 at mentor dot com> <alpine dot LSU dot 2 dot 11 dot 1502191400500 dot 28824 at zhemvz dot fhfr dot qr> <54E9D5E2 dot 3070805 at mentor dot com> <54FF0E2C dot 6080602 at mentor dot com> <552F77B4 dot 7050406 at mentor dot com> <alpine dot LSU dot 2 dot 11 dot 1504161055040 dot 6786 at zhemvz dot fhfr dot qr> <5576C67E dot 9070202 at arm dot com> <alpine dot LSU dot 2 dot 11 dot 1506091301120 dot 30088 at zhemvz dot fhfr dot qr>
On 09/06/15 13:03, Richard Biener wrote:
On Tue, 9 Jun 2015, Alan Lawrence wrote:
Hmmm. One side effect of this is that the line number information available in
the target hook gimplify_va_arg_expr, is now just the name of the containing
function, rather than the specific use of va_arg. Is there some way to get
this more precise location (e.g. gimple_location(stmt) in expand_ifn_va_arg_1,
the only caller of said hook)? I don't really want to have to add an extra
parameter to the target hook...
The x86 variant doesn't use any locations but if then the caller of
the target hook (expand_ifn_va_arg_1) should assign the IFNs location
to all statements expanded from it (it could set input_location to
that during the target hook call...)
That seems to work.
The scan-assembler-not test in the testcase in attached patch:
- fails without the expand_ifn_va_arg_1 patch hunk, and
- passes with that hunk.
I'll put it through bootstrap and reg-test on x86_64.
OK for trunk if that goes well?
Thanks,
- Tom
Handle location in expand_ifn_va_arg_1
2015-06-09 Tom de Vries <tom@codesourcery.com>
* tree-stdarg.c (expand_ifn_va_arg_1): Handle location.
* gcc.target/i386/vararg-loc.c: New test.
---
gcc/testsuite/gcc.target/i386/vararg-loc.c | 27 +++++++++++++++++++++++++++
gcc/tree-stdarg.c | 4 ++++
2 files changed, 31 insertions(+)
create mode 100644 gcc/testsuite/gcc.target/i386/vararg-loc.c
diff --git a/gcc/testsuite/gcc.target/i386/vararg-loc.c b/gcc/testsuite/gcc.target/i386/vararg-loc.c
new file mode 100644
index 0000000..f236fe3
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/vararg-loc.c
@@ -0,0 +1,27 @@
+/* { dg-do compile } */
+/* { dg-options "-g -O0" } */
+
+#include <stdarg.h>
+
+int /* 6. */
+ /* 7. */
+f (int a, ...) /* 8. */
+ /* 9. */
+{
+
+ int sum = a;
+
+ va_list ap;
+
+ va_start (ap, a);
+
+ sum += va_arg (ap, int); /* 18. */
+
+ sum += va_arg (ap, int); /* 20. */
+
+ return sum;
+}
+
+/* { dg-final { scan-assembler-not "\\.loc 1 \[6789\] 0" } } */
+/* { dg-final { scan-assembler-times "\\.loc 1 18 0" 1 } } */
+/* { dg-final { scan-assembler-times "\\.loc 1 20 0" 1 } } */
diff --git a/gcc/tree-stdarg.c b/gcc/tree-stdarg.c
index 08d10b5..65fe9f9 100644
--- a/gcc/tree-stdarg.c
+++ b/gcc/tree-stdarg.c
@@ -1031,6 +1031,7 @@ expand_ifn_va_arg_1 (function *fun)
bool modified = false;
basic_block bb;
gimple_stmt_iterator i;
+ location_t saved_location;
FOR_EACH_BB_FN (bb, fun)
for (i = gsi_start_bb (bb); !gsi_end_p (i); gsi_next (&i))
@@ -1051,6 +1052,8 @@ expand_ifn_va_arg_1 (function *fun)
ap = build_fold_indirect_ref (ap);
push_gimplify_context (false);
+ saved_location = input_location;
+ input_location = gimple_location (stmt);
/* Make it easier for the backends by protecting the valist argument
from multiple evaluations. */
@@ -1081,6 +1084,7 @@ expand_ifn_va_arg_1 (function *fun)
else
gimplify_expr (&expr, &pre, &post, is_gimple_lvalue, fb_lvalue);
+ input_location = saved_location;
pop_gimplify_context (NULL);
gimple_seq_add_seq (&pre, post);
--
1.9.1