Re: [PATCH v1] LoongArch: Adjust the vector cost model for better performance

[chenglulu]

Hi,liwei:

 It is best to add a test case. If the test case is really difficult to 
extract, then enrich the description information.

Thanks!

在 2023/9/18 下午5:23, Li Wei 写道:
gcc/ChangeLog:

        * config/loongarch/loongarch.cc (loongarch_builtin_vectorization_cost):
---
  gcc/config/loongarch/loongarch.cc | 21 ++++++++++++++-------
  1 file changed, 14 insertions(+), 7 deletions(-)

diff --git a/gcc/config/loongarch/loongarch.cc 
b/gcc/config/loongarch/loongarch.cc
index 845fad5a8e8..d24fc30570d 100644
--- a/gcc/config/loongarch/loongarch.cc
+++ b/gcc/config/loongarch/loongarch.cc
@@ -3860,17 +3860,18 @@ loongarch_builtin_vectorization_cost (enum 
vect_cost_for_stmt type_of_cost,
    switch (type_of_cost)
      {
        case scalar_stmt:
-      case scalar_load:
        case vector_stmt:
-      case vector_load:
        case vec_to_scalar:
        case scalar_to_vec:
-      case cond_branch_not_taken:
        case vec_promote_demote:
        case scalar_store:
-      case vector_store:
+      case scalar_load:
        return 1;
  
+      case vector_store:
+      case vector_load:
+       return 2;
+
        case vec_perm:
        return LASX_SUPPORTED_MODE_P (mode)
          && !LSX_SUPPORTED_MODE_P (mode) ? 2 : 1;
@@ -3881,14 +3882,20 @@ loongarch_builtin_vectorization_cost (enum 
vect_cost_for_stmt type_of_cost,
  
        case unaligned_store:
        case vector_scatter_store:
-       return 10;
+       return 5;
  
        case cond_branch_taken:
-       return 3;
+       return 4;
+
+      case cond_branch_not_taken:
+       return 2;
  
        case vec_construct:
        elements = TYPE_VECTOR_SUBPARTS (vectype);
-       return elements / 2 + 1;
+       if (ISA_HAS_LASX)
+         return elements + 1;
+       else
+         return elements;
  
        default:
        gcc_unreachable ();