RE: [PATCH PR94442] [AArch64] Redundant ldp/stp instructions emitted at -O3

[xiezhiheng]

> -----Original Message-----
> From: Richard Sandiford [mailto:richard.sandif...@arm.com]
> Sent: Friday, August 21, 2020 5:02 PM
> To: xiezhiheng <xiezhih...@huawei.com>
> Cc: Richard Biener <richard.guent...@gmail.com>; gcc-patches@gcc.gnu.org
> Subject: Re: [PATCH PR94442] [AArch64] Redundant ldp/stp instructions
> emitted at -O3

Cut...
 
> Looks like the saturating intrinsics might need a bit more thought.
> Would you mind submitting the patch with just the other parts?
> Those were uncontroversial and it would be a shame to hold them
> up over this.

Okay, I reorganized the existing patch and finished the first half of the 
intrinsics
except saturating intrinsics and load intrinsics.

Bootstrapped and tested on aarch64 Linux platform.

For load intrinsics, I have one problem when I set FLAG_READ_MEMORY for them,
some test cases like
gcc.target/aarch64/advsimd-intrinsics/vld2_lane_p8_indices_1.c
  #include <arm_neon.h>

  /* { dg-do compile } */
  /* { dg-skip-if "" { *-*-* } { "-fno-fat-lto-objects" } } */

  poly8x8x2_t
  f_vld2_lane_p8 (poly8_t * p, poly8x8x2_t v)
  {
    poly8x8x2_t res;
    /* { dg-error "lane 8 out of range 0 - 7" "" { target *-*-* } 0 } */
    res = vld2_lane_p8 (p, v, 8);
    /* { dg-error "lane -1 out of range 0 - 7" "" { target *-*-* } 0 } */
    res = vld2_lane_p8 (p, v, -1);
    return res;
  }
would fail in regression.  Because the first statement
  res = vld2_lane_p8 (p, v, 8);
would be eliminated as dead code in gimple phase but the error message is
generated in expand pass.  So I am going to replace the second statement
  res = vld2_lane_p8 (p, v, -1);
with
  res = vld2_lane_p8 (p, res, -1);
or do you have any other suggestions?

And for test case gcc.target/aarch64/arg-type-diagnostics-1.c, I return the 
result
to prevent the statement
  result = vrsra_n_s32 (arg1, arg2, a);
from being eliminated by treated as dead code.

Thanks,
Xie Zhiheng


diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 7a71b4367d4..217344d7d1f 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,8 @@
+2020-08-25  Zhiheng Xie  <xiezhih...@huawei.com>
+
+       * config/aarch64/aarch64-simd-builtins.def: Add proper FLAGS
+       for intrinsic functions.
+

diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index b9562e67883..e10bcc9b28a 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2020-08-25  Zhiheng Xie  <xiezhih...@huawei.com>
+
+       * gcc.target/aarch64/arg-type-diagnostics-1.c: Return result
+       to prevent statement from being eliminated.
+

pr94442-v1.patch
Description: pr94442-v1.patch