diff mbox series

rs6000: Fix vector homogeneous aggregates (PR86197)

Message ID 91ad08f6bdab72e72c606211a2b0803399382e72.1529404593.git.segher@kernel.crashing.org
State New
Headers show
Series rs6000: Fix vector homogeneous aggregates (PR86197) | expand

Commit Message

Segher Boessenkool June 19, 2018, 10:45 a.m. UTC 
The existing code allows only 4 vectors worth of ieee128 homogeneous
aggregates, but it should be 8.  This happens because at one spot it
is mistakenly qualified as being passed in floating point registers.

This patch fixes it and makes the code easier to read.  Committing to
trunk; needs backports too.


Segher


2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>

	* config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
	ieee128 argument takes up only one (vector) register, not two (floating
	point) registers.

---
 gcc/config/rs6000/rs6000.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

Comments

Segher Boessenkool June 25, 2018, 11:32 a.m. UTC | #1 
On Tue, Jun 19, 2018 at 10:45:59AM +0000, Segher Boessenkool wrote:
> The existing code allows only 4 vectors worth of ieee128 homogeneous
> aggregates, but it should be 8.  This happens because at one spot it
> is mistakenly qualified as being passed in floating point registers.
> 
> This patch fixes it and makes the code easier to read.  Committing to
> trunk; needs backports too.

Backported to 8 now.


Segher
Segher Boessenkool Aug. 10, 2018, 8:46 p.m. UTC | #2 
On Mon, Jun 25, 2018 at 06:32:27AM -0500, Segher Boessenkool wrote:
> On Tue, Jun 19, 2018 at 10:45:59AM +0000, Segher Boessenkool wrote:
> > The existing code allows only 4 vectors worth of ieee128 homogeneous
> > aggregates, but it should be 8.  This happens because at one spot it
> > is mistakenly qualified as being passed in floating point registers.
> > 
> > This patch fixes it and makes the code easier to read.  Committing to
> > trunk; needs backports too.
> 
> Backported to 8 now.

And to 7 and 6.


Segher
diff mbox series

Patch

diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index 307c88e..78f6ffe 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -10867,12 +10867,12 @@  rs6000_discover_homogeneous_aggregate (machine_mode mode, const_tree type,
 
       if (field_count > 0)
 	{
-	  int n_regs = (SCALAR_FLOAT_MODE_P (field_mode) ?
-			(GET_MODE_SIZE (field_mode) + 7) >> 3 : 1);
+	  int reg_size = ALTIVEC_OR_VSX_VECTOR_MODE (field_mode) ? 16 : 8;
+	  int field_size = ROUND_UP (GET_MODE_SIZE (field_mode), reg_size);
 
 	  /* The ELFv2 ABI allows homogeneous aggregates to occupy
 	     up to AGGR_ARG_NUM_REG registers.  */
-	  if (field_count * n_regs <= AGGR_ARG_NUM_REG)
+	  if (field_count * field_size <= AGGR_ARG_NUM_REG * reg_size)
 	    {
 	      if (elt_mode)
 		*elt_mode = field_mode;