https://gcc.gnu.org/g:cc33cf487524e738e3bb723154ce315dda4fefdd

commit r16-8383-gcc33cf487524e738e3bb723154ce315dda4fefdd
Author: H.J. Lu <[email protected]>
Date:   Tue Mar 31 08:59:50 2026 -0700

    Check BIGGEST_ALIGNMENT when forcing a local copy
    
    BIGGEST_ALIGNMENT is the biggest alignment that any data type can require
    on the target machine.  It is not the biggest alignment that is supported,
    just the biggest alignment that, when violated, may cause a fault.
    
    MAX_SUPPORTED_STACK_ALIGNMENT is the maximum stack alignment guaranteed by
    the backend.  If SUPPORTS_STACK_ALIGNMENT is false, it is defined as
    PREFERRED_STACK_BOUNDARY.
    
    Adjust alignment check forcing a local copy if the argument on stack is
    not aligned to its type and the current alignment is less than the
    minimum of BIGGEST_ALIGNMENT and MAX_SUPPORTED_STACK_ALIGNMENT.
    
    gcc/
    
            PR middle-end/124697
            * function.cc (assign_parm_adjust_stack_rtl): Force a local copy
            if the current alignment is less than the minimum of
            BIGGEST_ALIGNMENT and MAX_SUPPORTED_STACK_ALIGNMENT.
    
    gcc/testsuite/
    
            PR middle-end/124697
            * gcc.target/i386/pr120839-1b.c: Adjusted.
            * gcc.target/i386/pr124697-1a.c: New test.
            * gcc.target/i386/pr124697-1b.c: Likewise.
            * gcc.target/i386/pr124697-2a.c: Likewise.
            * gcc.target/i386/pr124697-2b.c: Likewise.
            * gcc.target/i386/pr124697-3a.c: Likewise.
            * gcc.target/i386/pr124697-3b.c: Likewise.
    
    Signed-off-by: H.J. Lu <[email protected]>

Diff:
---
 gcc/function.cc                             |  3 ++-
 gcc/testsuite/gcc.target/i386/pr120839-1b.c |  2 +-
 gcc/testsuite/gcc.target/i386/pr124697-1a.c | 31 +++++++++++++++++++++++++++++
 gcc/testsuite/gcc.target/i386/pr124697-1b.c | 19 ++++++++++++++++++
 gcc/testsuite/gcc.target/i386/pr124697-2a.c | 30 ++++++++++++++++++++++++++++
 gcc/testsuite/gcc.target/i386/pr124697-2b.c | 18 +++++++++++++++++
 gcc/testsuite/gcc.target/i386/pr124697-3a.c | 31 +++++++++++++++++++++++++++++
 gcc/testsuite/gcc.target/i386/pr124697-3b.c | 19 ++++++++++++++++++
 8 files changed, 151 insertions(+), 2 deletions(-)

diff --git a/gcc/function.cc b/gcc/function.cc
index 46c0d8b54c2f..109821c16fc9 100644
--- a/gcc/function.cc
+++ b/gcc/function.cc
@@ -2840,7 +2840,8 @@ assign_parm_adjust_stack_rtl (tree parm, struct 
assign_parm_data_one *data)
                                                 MEM_ALIGN (stack_parm))))
          || (data->nominal_type
              && TYPE_ALIGN (data->nominal_type) > MEM_ALIGN (stack_parm)
-             && (MEM_ALIGN (stack_parm) < PREFERRED_STACK_BOUNDARY
+             && ((MEM_ALIGN (stack_parm)
+                  < MIN (BIGGEST_ALIGNMENT, MAX_SUPPORTED_STACK_ALIGNMENT))
                  /* If its address is taken, make a local copy whose
                     maximum alignment is MAX_SUPPORTED_STACK_ALIGNMENT.
                   */
diff --git a/gcc/testsuite/gcc.target/i386/pr120839-1b.c 
b/gcc/testsuite/gcc.target/i386/pr120839-1b.c
index d9dafc19afb7..cd7701f1d895 100644
--- a/gcc/testsuite/gcc.target/i386/pr120839-1b.c
+++ b/gcc/testsuite/gcc.target/i386/pr120839-1b.c
@@ -1,5 +1,5 @@
 /* { dg-do compile } */
 /* { dg-options "-O2 -mavx2" } */
-/* { dg-final { scan-assembler-not "and\[lq\]?\[\\t \]*\\$-32,\[\\t 
\]*%\[re\]?sp" { xfail ia32 } } } */
+/* { dg-final { scan-assembler "and\[lq\]?\[\\t \]*\\$-32,\[\\t \]*%\[re\]?sp" 
} } */
 
 #include "pr120839-1a.c"
diff --git a/gcc/testsuite/gcc.target/i386/pr124697-1a.c 
b/gcc/testsuite/gcc.target/i386/pr124697-1a.c
new file mode 100644
index 000000000000..63d5cc7862c9
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr124697-1a.c
@@ -0,0 +1,31 @@
+/* { dg-do run  { target avx2_runtime } }  */
+/* { dg-options "-O2 -mavx2" } */
+
+typedef double v4df __attribute__((vector_size(32)));
+typedef double v2df __attribute__((vector_size(16)));
+typedef struct {
+  v2df a[2];
+} c __attribute__((aligned(32)));
+v4df d;
+c v;
+
+__attribute__ ((noipa))
+void
+foo (float a1, float a2, float a3, float a4, float a5, float a6, c f)
+{
+  d = *(v4df *) &f;
+}
+
+__attribute__ ((noipa))
+void
+bar (void)
+{
+  foo (0, 0, 0, 0, 0, 0, v);
+}
+
+int
+main ()
+{
+  bar ();
+  return 0;
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr124697-1b.c 
b/gcc/testsuite/gcc.target/i386/pr124697-1b.c
new file mode 100644
index 000000000000..3b74f62b508c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr124697-1b.c
@@ -0,0 +1,19 @@
+/* { dg-do compile }  */
+/* { dg-options "-O2 -mno-avx512f -mavx2" } */
+/* { dg-final { scan-assembler-times "vmovupd.*, %ymm0" 1 } } */
+/* { dg-final { scan-assembler "and\[lq\]?\[\\t \]*\\$-32,\[\\t \]*%\[re\]?sp" 
} } */
+/* { dg-final { scan-assembler-not "vmovapd.*, %ymm0" } } */
+
+typedef double v4df __attribute__((vector_size(32)));
+typedef double v2df __attribute__((vector_size(16)));
+typedef struct {
+  v2df a[2];
+} c __attribute__((aligned(32)));
+v4df d;
+c v;
+
+void
+foo (float a1, float a2, float a3, float a4, float a5, float a6, c f)
+{
+  d = *(v4df *) &f;
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr124697-2a.c 
b/gcc/testsuite/gcc.target/i386/pr124697-2a.c
new file mode 100644
index 000000000000..46a1f953290f
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr124697-2a.c
@@ -0,0 +1,30 @@
+/* { dg-do run  { target avx2_runtime } }  */
+/* { dg-options "-O2 -mavx2" } */
+
+typedef double v4df __attribute__((vector_size(32)));
+typedef struct {
+  v4df a[2];
+} c __attribute__((aligned(64)));
+v4df d;
+c v;
+
+__attribute__ ((noipa))
+void
+foo (float a1, float a2, float a3, float a4, float a5, float a6, c f)
+{
+  d = *(v4df *) &f;
+}
+
+__attribute__ ((noipa))
+void
+bar (void)
+{
+  foo (0, 0, 0, 0, 0, 0, v);
+}
+
+int
+main ()
+{
+  bar ();
+  return 0;
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr124697-2b.c 
b/gcc/testsuite/gcc.target/i386/pr124697-2b.c
new file mode 100644
index 000000000000..1c2256df9b9e
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr124697-2b.c
@@ -0,0 +1,18 @@
+/* { dg-do compile }  */
+/* { dg-options "-O2 -mno-avx512f -mavx2" } */
+/* { dg-final { scan-assembler-times "vmovapd.*, %ymm0" 1 } } */
+/* { dg-final { scan-assembler-not "vmovupd.*, %ymm0" } } */
+/* { dg-final { scan-assembler-not "and\[lq\]?\[\\t \]*\\$-64,\[\\t 
\]*%\[re\]?sp" } } */
+
+typedef double v4df __attribute__((vector_size(32)));
+typedef struct {
+  v4df a[2];
+} c __attribute__((aligned(64)));
+v4df d;
+c v;
+
+void
+foo (float a1, float a2, float a3, float a4, float a5, float a6, c f)
+{
+  d = *(v4df *) &f;
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr124697-3a.c 
b/gcc/testsuite/gcc.target/i386/pr124697-3a.c
new file mode 100644
index 000000000000..87e6f6f75204
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr124697-3a.c
@@ -0,0 +1,31 @@
+/* { dg-do run  { target avx512f_runtime } }  */
+/* { dg-options "-O2 -mavx512f" } */
+
+typedef double v4df __attribute__((vector_size(32)));
+typedef double v8df __attribute__((vector_size(64)));
+typedef struct {
+  v4df a[2];
+} c __attribute__((aligned(64)));
+v8df d;
+c v;
+
+__attribute__ ((noipa))
+void
+foo (float a1, float a2, float a3, float a4, float a5, float a6, c f)
+{
+  d = *(v8df *) &f;
+}
+
+__attribute__ ((noipa))
+void
+bar (void)
+{
+  foo (0, 0, 0, 0, 0, 0, v);
+}
+
+int
+main ()
+{
+  bar ();
+  return 0;
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr124697-3b.c 
b/gcc/testsuite/gcc.target/i386/pr124697-3b.c
new file mode 100644
index 000000000000..31d52f83ad5e
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr124697-3b.c
@@ -0,0 +1,19 @@
+/* { dg-do compile }  */
+/* { dg-options "-O2 -mavx512f" } */
+/* { dg-final { scan-assembler-times "vmovupd.*, %zmm0" 1 } } */
+/* { dg-final { scan-assembler "and\[lq\]?\[\\t \]*\\$-64,\[\\t \]*%\[re\]?sp" 
} } */
+/* { dg-final { scan-assembler-not "vmovapd.*, %zmm0" } } */
+
+typedef double v4df __attribute__((vector_size(32)));
+typedef double v8df __attribute__((vector_size(64)));
+typedef struct {
+  v4df a[2];
+} c __attribute__((aligned(64)));
+v8df d;
+c v;
+
+void
+foo (float a1, float a2, float a3, float a4, float a5, float a6, c f)
+{
+  d = *(v8df *) &f;
+}

Reply via email to