diff options
| author | Alejandro Martinez <alejandro.martinezvicente@arm.com> | 2019-05-07 16:34:20 +0000 |
|---|---|---|
| committer | Alejandro Martinez <alejandro@gcc.gnu.org> | 2019-05-07 16:34:20 +0000 |
| commit | a9fad8fe6c84de272f2a56d462e67d53c9f4a73d (patch) | |
| tree | 9f0ff9561477c22ee099d09330ce742f5830d9a3 /gcc/tree-vect-loop.c | |
| parent | 0a59215131c02dee4c8829f93d1ee678647614da (diff) | |
| download | gcc-a9fad8fe6c84de272f2a56d462e67d53c9f4a73d.zip gcc-a9fad8fe6c84de272f2a56d462e67d53c9f4a73d.tar.gz gcc-a9fad8fe6c84de272f2a56d462e67d53c9f4a73d.tar.bz2 | |
This patch adds support to vectorize sum of abslolute differences (SAD_EXPR)
using SVE.
Given this input code:
int
sum_abs (uint8_t *restrict x, uint8_t *restrict y, int n)
{
int sum = 0;
for (int i = 0; i < n; i++)
{
sum += __builtin_abs (x[i] - y[i]);
}
return sum;
}
The resulting SVE code is:
0000000000000000 <sum_abs>:
0: 7100005f cmp w2, #0x0
4: 5400026d b.le 50 <sum_abs+0x50>
8: d2800003 mov x3, #0x0 // #0
c: 93407c42 sxtw x2, w2
10: 2538c002 mov z2.b, #0
14: 25221fe0 whilelo p0.b, xzr, x2
18: 2538c023 mov z3.b, #1
1c: 2518e3e1 ptrue p1.b
20: a4034000 ld1b {z0.b}, p0/z, [x0, x3]
24: a4034021 ld1b {z1.b}, p0/z, [x1, x3]
28: 0430e3e3 incb x3
2c: 0520c021 sel z1.b, p0, z1.b, z0.b
30: 25221c60 whilelo p0.b, x3, x2
34: 040d0420 uabd z0.b, p1/m, z0.b, z1.b
38: 44830402 udot z2.s, z0.b, z3.b
3c: 54ffff21 b.ne 20 <sum_abs+0x20> // b.any
40: 2598e3e0 ptrue p0.s
44: 04812042 uaddv d2, p0, z2.s
48: 1e260040 fmov w0, s2
4c: d65f03c0 ret
50: 1e2703e2 fmov s2, wzr
54: 1e260040 fmov w0, s2
58: d65f03c0 ret
Notice how udot is used inside a fully masked loop.
gcc/Changelog:
2019-05-07 Alejandro Martinez <alejandro.martinezvicente@arm.com>
* config/aarch64/aarch64-sve.md (<su>abd<mode>_3): New define_expand.
(aarch64_<su>abd<mode>_3): Likewise.
(*aarch64_<su>abd<mode>_3): New define_insn.
(<sur>sad<vsi2qi>): New define_expand.
* config/aarch64/iterators.md: Added MAX_OPP attribute.
* tree-vect-loop.c (use_mask_by_cond_expr_p): Add SAD_EXPR.
(build_vect_cond_expr): Likewise.
gcc/testsuite/Changelog:
2019-05-07 Alejandro Martinez <alejandro.martinezvicente@arm.com>
* gcc.target/aarch64/sve/sad_1.c: New test for sum of absolute
differences.
From-SVN: r270975
Diffstat (limited to 'gcc/tree-vect-loop.c')
| -rw-r--r-- | gcc/tree-vect-loop.c | 12 |
1 files changed, 12 insertions, 0 deletions
diff --git a/gcc/tree-vect-loop.c b/gcc/tree-vect-loop.c index 493c1ab..057a874 100644 --- a/gcc/tree-vect-loop.c +++ b/gcc/tree-vect-loop.c @@ -5973,6 +5973,7 @@ use_mask_by_cond_expr_p (enum tree_code code, internal_fn cond_fn, switch (code) { case DOT_PROD_EXPR: + case SAD_EXPR: return true; default: @@ -6002,6 +6003,17 @@ build_vect_cond_expr (enum tree_code code, tree vop[3], tree mask, break; } + case SAD_EXPR: + { + tree vectype = TREE_TYPE (vop[1]); + tree masked_op1 = make_temp_ssa_name (vectype, NULL, "masked_op1"); + gassign *select = gimple_build_assign (masked_op1, VEC_COND_EXPR, + mask, vop[1], vop[0]); + gsi_insert_before (gsi, select, GSI_SAME_STMT); + vop[1] = masked_op1; + break; + } + default: gcc_unreachable (); } |