On Mon, Aug 24, 2020 at 6:12 PM Jeff Law <l...@redhat.com> wrote:
> On Thu, 2020-08-06 at 12:42 +0000, Pip Cet via Gcc-patches wrote:
> > I've bootstrapped and run the test suite with the patch, without
> > differences.
> So it looks like Richard has given you some feedback and you've got some
> further
> work to do. I just wanted to chime in and say thanks for tackling this.
Thanks for the encouragement!
Richard wrote:
> We should be able to test whether the rest of the parallel is suitable by
> adding a single_set test (after testing everything else).
It seems to me the existing two checks using single_set are sufficient
to ensure this.
Revised patch attached. It bootstraps and causes no test regressions,
and it appears to work.
From 72e05ac3a570f3e6ca9e54599db1b3b4daa84e90 Mon Sep 17 00:00:00 2001
From: Pip Cet <pip...@gmail.com>
Date: Tue, 25 Aug 2020 14:23:58 +0000
Subject: [PATCH] cmpelim: handle extra clobbers
Handle extra clobbers in CC-clobbering insns when attempting to
recognize the corresponding CC-setting insn.
This is for the AVR CCmode conversion. AVR has insns like
(define_insn "andhi3"
[(set (match_operand:HI 0 ...)
(and:HI (match_operand:HI 1 ...)
(match_operand:HI 2 ...)))
(clobber (reg:CC REG_CC))
(clobber (match_scratch:QI 3 ...))] ...)
which can be profitably converted into CC-setting variants.
2020-08-28 Pip Cet <pip...@gmail.com>
gcc/ChangeLog:
* compare-elim.c (arithmetic_flags_clobber_p): Allow extra
clobbers. (try_validate_parallel): (try_eliminate_compare):
Likewise.
---
gcc/compare-elim.c | 20 +++++++++++++++++---
1 file changed, 17 insertions(+), 3 deletions(-)
diff --git a/gcc/compare-elim.c b/gcc/compare-elim.c
index 85f3e344074..8e5c9a05fa8 100644
--- a/gcc/compare-elim.c
+++ b/gcc/compare-elim.c
@@ -202,7 +202,7 @@ arithmetic_flags_clobber_p (rtx_insn *insn)
if (asm_noperands (pat) >= 0)
return false;
- if (GET_CODE (pat) == PARALLEL && XVECLEN (pat, 0) == 2)
+ if (GET_CODE (pat) == PARALLEL && XVECLEN (pat, 0) >= 2)
{
x = XVECEXP (pat, 0, 0);
if (GET_CODE (x) != SET)
@@ -663,6 +663,17 @@ static rtx
try_validate_parallel (rtx set_a, rtx set_b)
{
rtx par = gen_rtx_PARALLEL (VOIDmode, gen_rtvec (2, set_a, set_b));
+ if (GET_CODE (set_b) == PARALLEL)
+ {
+ int len = XVECLEN (set_b, 0);
+ rtvec v = rtvec_alloc (len);
+ RTVEC_ELT (v, 0) = set_a;
+ RTVEC_ELT (v, 1) = XVECEXP (set_b, 0, 0);
+ for (int i = 2; i < len; i++)
+ RTVEC_ELT (v, i) = XVECEXP (set_b, 0, i);
+
+ par = gen_rtx_PARALLEL (VOIDmode, v);
+ }
rtx_insn *insn = make_insn_raw (par);
if (insn_invalid_p (insn, false))
@@ -873,10 +884,13 @@ try_eliminate_compare (struct comparison *cmp)
[(set (reg:CCM) (compare:CCM (operation) (immediate)))
(set (reg) (operation)] */
- rtvec v = rtvec_alloc (2);
+ int len = XVECLEN (PATTERN (insn), 0);
+ rtvec v = rtvec_alloc (len);
RTVEC_ELT (v, 0) = y;
RTVEC_ELT (v, 1) = x;
-
+ for (int i = 2; i < len; i++)
+ RTVEC_ELT (v, i) = XVECEXP (PATTERN (insn), 0, i);
+
rtx pat = gen_rtx_PARALLEL (VOIDmode, v);
/* Succeed if the new instruction is valid. Note that we may have started
--
2.28.0
