OpenE2K
/
gcc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

aarch64: Revert bogus fix for PR105254 

In f2ebf2d98e I'd forced the
chosen unroll factor to be a factor of the VF, in order to
work around an exact_div ICE in PR105254.  This was completely
bogus -- clearly I didn't look in enough detail at why we ended
up with an unrolled VF that wasn't a multiple of the UF.

Kewen has since fixed the bug properly for PR105940, so this
patch reverts my earlier attempt.  Sorry for the stupidity.

gcc/
	PR tree-optimization/105254
	PR tree-optimization/105940

	Revert:

	* config/aarch64/aarch64.cc
	(aarch64_vector_costs::determine_suggested_unroll_factor): Take a
	loop_vec_info as argument.  Restrict the unroll factor to values
	that divide the VF.
	(aarch64_vector_costs::finish_cost): Update call accordingly.

gcc/testsuite/
	* gcc.target/aarch64/sve/cost_model_14.c: New test.
This commit is contained in:
Richard Sandiford 2022-06-15 11:12:51 +01:00
parent 183a4f3829
commit 2636660b6f
2 changed files with 17 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
gcc/config/aarch64/aarch64.cc
View File

@ -15631,7 +15631,7 @@ private:
unsigned int adjust_body_cost (loop_vec_info, const aarch64_vector_costs *,
unsigned int);
bool prefer_unrolled_loop () const;
unsigned int determine_suggested_unroll_factor (loop_vec_info);
unsigned int determine_suggested_unroll_factor ();
/* True if we have performed one-time initialization based on the
vec_info. */
@ -16740,8 +16740,7 @@ adjust_body_cost_sve (const aarch64_vec_op_count *ops,
}
unsigned int
aarch64_vector_costs::
determine_suggested_unroll_factor (loop_vec_info loop_vinfo)
aarch64_vector_costs::determine_suggested_unroll_factor ()
{
bool sve = m_vec_flags & VEC_ANY_SVE;
/* If we are trying to unroll an Advanced SIMD main loop that contains
@ -16755,7 +16754,6 @@ determine_suggested_unroll_factor (loop_vec_info loop_vinfo)
return 1;
unsigned int max_unroll_factor = 1;
auto vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
for (auto vec_ops : m_ops)
{
aarch64_simd_vec_issue_info const *vec_issue
@ -16764,8 +16762,7 @@ determine_suggested_unroll_factor (loop_vec_info loop_vinfo)
return 1;
/* Limit unroll factor to a value adjustable by the user, the default
value is 4. */
unsigned int unroll_factor = MIN (aarch64_vect_unroll_limit,
(int) known_alignment (vf));
unsigned int unroll_factor = aarch64_vect_unroll_limit;
unsigned int factor
= vec_ops.reduction_latency > 1 ? vec_ops.reduction_latency : 1;
unsigned int temp;
@ -16943,8 +16940,7 @@ aarch64_vector_costs::finish_cost (const vector_costs *uncast_scalar_costs)
{
m_costs[vect_body] = adjust_body_cost (loop_vinfo, scalar_costs,
m_costs[vect_body]);
m_suggested_unroll_factor
= determine_suggested_unroll_factor (loop_vinfo);
m_suggested_unroll_factor = determine_suggested_unroll_factor ();
}
/* Apply the heuristic described above m_stp_sequence_cost. Prefer

13
gcc/testsuite/gcc.target/aarch64/sve/cost_model_14.c Normal file
View File

@ -0,0 +1,13 @@
/* { dg-options "-O3 -mtune=neoverse-v1" } */
#include <stdint.h>
uint64_t f2(uint64_t *ptr, int n) {
uint64_t res = 0;
for (int i = 0; i < n; ++i)
res += ptr[i];
return res;
}
/* { dg-final { scan-assembler-times {\tld1d\tz[0-9]+\.d,} 5 } } */
/* { dg-final { scan-assembler-times {\tadd\tz[0-9]+\.d,} 8 } } */