Represents the cost of a rdffr.
gcc/ChangeLog:
* config/aarch64/aarch64-protos.h (struct sve_vec_cost): Add
read_ffr_cost member.
* config/aarch64/aarch64.cc (aarch64_builtin_vectorization_cost):
Add handling for hssr_read using read_ffr_cost.
(aarch64_vector_costs::count_ops):
Add handling for hssr_read.
* config/aarch64/tuning_models/a64fx.h: Add entry for
read_ffr_cost.
* config/aarch64/tuning_models/cortexa320.h: Likewise.
* config/aarch64/tuning_models/cortexx925.h: Likewise.
* config/aarch64/tuning_models/generic.h: Likewise.
* config/aarch64/tuning_models/generic_armv8_a.h: Likewise.
* config/aarch64/tuning_models/generic_armv9_a.h: Likewise.
* config/aarch64/tuning_models/hip12.h: Likewise.
* config/aarch64/tuning_models/neoverse512tvb.h: Likewise.
* config/aarch64/tuning_models/neoversen2.h: Likewise.
* config/aarch64/tuning_models/neoversen3.h: Likewise.
* config/aarch64/tuning_models/neoversev1.h: Likewise.
* config/aarch64/tuning_models/neoversev2.h: Likewise.
* config/aarch64/tuning_models/neoversev3.h: Likewise.
* config/aarch64/tuning_models/neoversev3ae.h: Likewise.
* config/aarch64/tuning_models/olympus.h: Likewise.
---
gcc/config/aarch64/aarch64-protos.h | 10 ++++++++--
gcc/config/aarch64/aarch64.cc | 12 ++++++++++++
gcc/config/aarch64/tuning_models/a64fx.h | 3 ++-
gcc/config/aarch64/tuning_models/cortexa320.h | 3 ++-
gcc/config/aarch64/tuning_models/cortexx925.h | 3 ++-
gcc/config/aarch64/tuning_models/generic.h | 3 ++-
gcc/config/aarch64/tuning_models/generic_armv8_a.h | 3 ++-
gcc/config/aarch64/tuning_models/generic_armv9_a.h | 3 ++-
gcc/config/aarch64/tuning_models/hip12.h | 3 ++-
gcc/config/aarch64/tuning_models/neoverse512tvb.h | 3 ++-
gcc/config/aarch64/tuning_models/neoversen2.h | 3 ++-
gcc/config/aarch64/tuning_models/neoversen3.h | 3 ++-
gcc/config/aarch64/tuning_models/neoversev1.h | 3 ++-
gcc/config/aarch64/tuning_models/neoversev2.h | 3 ++-
gcc/config/aarch64/tuning_models/neoversev3.h | 3 ++-
gcc/config/aarch64/tuning_models/neoversev3ae.h | 3 ++-
gcc/config/aarch64/tuning_models/olympus.h | 3 ++-
17 files changed, 50 insertions(+), 17 deletions(-)
diff --git a/gcc/config/aarch64/aarch64-protos.h
b/gcc/config/aarch64/aarch64-protos.h
index 22fd1c68b05..c21611f97f7 100644
--- a/gcc/config/aarch64/aarch64-protos.h
+++ b/gcc/config/aarch64/aarch64-protos.h
@@ -266,7 +266,9 @@ struct sve_vec_cost : simd_vec_cost
unsigned int gather_load_x64_cost,
unsigned int gather_load_x32_init_cost,
unsigned int gather_load_x64_init_cost,
- unsigned int scatter_store_elt_cost)
+ unsigned int scatter_store_elt_cost,
+ unsigned int read_ffr_cost
+ )
: simd_vec_cost (base),
clast_cost (clast_cost),
fadda_f16_cost (fadda_f16_cost),
@@ -276,7 +278,8 @@ struct sve_vec_cost : simd_vec_cost
gather_load_x64_cost (gather_load_x64_cost),
gather_load_x32_init_cost (gather_load_x32_init_cost),
gather_load_x64_init_cost (gather_load_x64_init_cost),
- scatter_store_elt_cost (scatter_store_elt_cost)
+ scatter_store_elt_cost (scatter_store_elt_cost),
+ read_ffr_cost (read_ffr_cost)
{}
/* The cost of a vector-to-scalar CLASTA or CLASTB instruction,
@@ -303,6 +306,9 @@ struct sve_vec_cost : simd_vec_cost
/* The per-element cost of a scatter store. */
int scatter_store_elt_cost;
+
+ /* Cost of an rdffrs. */
+ int read_ffr_cost;
};
/* Base information about how the CPU issues code, containing
diff --git a/gcc/config/aarch64/aarch64.cc b/gcc/config/aarch64/aarch64.cc
index d8bd0cc74f1..00df040c8dc 100644
--- a/gcc/config/aarch64/aarch64.cc
+++ b/gcc/config/aarch64/aarch64.cc
@@ -18007,6 +18007,17 @@ aarch64_builtin_vectorization_cost (enum
vect_cost_for_stmt type_of_cost,
elements = estimated_poly_value (TYPE_VECTOR_SUBPARTS (vectype));
return elements / 2 + 1;
+ case hssr_read:
+ {
+ const sve_vec_cost *sve_costs = aarch64_tune_params.vec_costs->sve;
+ /* We need SVE to have HSSR so we surely will have sve costs defined
+ too. */
+ gcc_assert (sve_costs);
+
+ /* This will be a rdffr + b.nlast. */
+ return sve_costs->read_ffr_cost + costs->cond_not_taken_branch_cost;
+ }
+
default:
gcc_unreachable ();
}
@@ -18636,6 +18647,7 @@ aarch64_vector_costs::count_ops (unsigned int count,
vect_cost_for_stmt kind,
case scalar_to_vec:
case vector_stmt:
case scalar_stmt:
+ case hssr_read:
ops->general_ops += count;
break;
diff --git a/gcc/config/aarch64/tuning_models/a64fx.h
b/gcc/config/aarch64/tuning_models/a64fx.h
index c3222960f17..d457e8e29a8 100644
--- a/gcc/config/aarch64/tuning_models/a64fx.h
+++ b/gcc/config/aarch64/tuning_models/a64fx.h
@@ -106,7 +106,8 @@ static const sve_vec_cost a64fx_sve_vector_cost =
32, /* gather_load_x64_cost */
0, /* gather_load_x32_init_cost */
0, /* gather_load_x64_init_cost */
- 1 /* scatter_store_elt_cost */
+ 1, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const struct cpu_vector_cost a64fx_vector_cost =
diff --git a/gcc/config/aarch64/tuning_models/cortexa320.h
b/gcc/config/aarch64/tuning_models/cortexa320.h
index d5e1f2b2993..749d2afd879 100644
--- a/gcc/config/aarch64/tuning_models/cortexa320.h
+++ b/gcc/config/aarch64/tuning_models/cortexa320.h
@@ -137,7 +137,8 @@ static const sve_vec_cost cortexa320_sve_vector_cost =
19, /* gather_load_x64_cost */
48, /* gather_load_x32_init_cost */
38, /* gather_load_x64_init_cost */
- 0 /* scatter_store_elt_cost */
+ 0, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
/* Cortexa320 costs for vector insn classes. */
diff --git a/gcc/config/aarch64/tuning_models/cortexx925.h
b/gcc/config/aarch64/tuning_models/cortexx925.h
index 2a73b369706..46e2f653812 100644
--- a/gcc/config/aarch64/tuning_models/cortexx925.h
+++ b/gcc/config/aarch64/tuning_models/cortexx925.h
@@ -119,7 +119,8 @@ static const sve_vec_cost cortexx925_sve_vector_cost =
12, /* gather_load_x64_cost */
42, /* gather_load_x32_init_cost */
24, /* gather_load_x64_init_cost */
- 1 /* scatter_store_elt_cost */
+ 1, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_scalar_vec_issue_info cortexx925_scalar_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/generic.h
b/gcc/config/aarch64/tuning_models/generic.h
index af013e478d4..e163b60f405 100644
--- a/gcc/config/aarch64/tuning_models/generic.h
+++ b/gcc/config/aarch64/tuning_models/generic.h
@@ -107,7 +107,8 @@ static const sve_vec_cost generic_sve_vector_cost =
2, /* gather_load_x64_cost */
0, /* gather_load_x32_init_cost */
0, /* gather_load_x64_init_cost */
- 1 /* scatter_store_elt_cost */
+ 1, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
/* Generic costs for vector insn classes. */
diff --git a/gcc/config/aarch64/tuning_models/generic_armv8_a.h
b/gcc/config/aarch64/tuning_models/generic_armv8_a.h
index 29cf3a5a7c7..4de03a3770e 100644
--- a/gcc/config/aarch64/tuning_models/generic_armv8_a.h
+++ b/gcc/config/aarch64/tuning_models/generic_armv8_a.h
@@ -108,7 +108,8 @@ static const sve_vec_cost generic_armv8_a_sve_vector_cost =
2, /* gather_load_x64_cost */
12, /* gather_load_x32_init_cost */
4, /* gather_load_x64_init_cost */
- 1 /* scatter_store_elt_cost */
+ 1, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
/* Generic costs for vector insn classes. */
diff --git a/gcc/config/aarch64/tuning_models/generic_armv9_a.h
b/gcc/config/aarch64/tuning_models/generic_armv9_a.h
index 05e777d20ef..d0131c168d9 100644
--- a/gcc/config/aarch64/tuning_models/generic_armv9_a.h
+++ b/gcc/config/aarch64/tuning_models/generic_armv9_a.h
@@ -138,7 +138,8 @@ static const sve_vec_cost generic_armv9_a_sve_vector_cost =
12, /* gather_load_x64_cost */
42, /* gather_load_x32_init_cost */
24, /* gather_load_x64_init_cost */
- 3 /* scatter_store_elt_cost */
+ 3, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_scalar_vec_issue_info generic_armv9_a_scalar_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/hip12.h
b/gcc/config/aarch64/tuning_models/hip12.h
index cfb204b2e53..64e08de3da0 100644
--- a/gcc/config/aarch64/tuning_models/hip12.h
+++ b/gcc/config/aarch64/tuning_models/hip12.h
@@ -117,7 +117,8 @@ static const sve_vec_cost hip12_sve_vector_cost =
16, /* gather_load_x64_cost */
96, /* gather_load_x32_init_cost */
32, /* gather_load_x64_init_cost */
- 3 /* scatter_store_elt_cost */
+ 3, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_scalar_vec_issue_info hip12_scalar_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/neoverse512tvb.h
b/gcc/config/aarch64/tuning_models/neoverse512tvb.h
index d6e2c2e8120..910eb0d52cf 100644
--- a/gcc/config/aarch64/tuning_models/neoverse512tvb.h
+++ b/gcc/config/aarch64/tuning_models/neoverse512tvb.h
@@ -81,7 +81,8 @@ static const sve_vec_cost neoverse512tvb_sve_vector_cost =
12, /* gather_load_x64_cost */
42, /* gather_load_x32_init_cost */
24, /* gather_load_x64_init_cost */
- 3 /* scatter_store_elt_cost */
+ 3, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_sve_vec_issue_info neoverse512tvb_sve_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/neoversen2.h
b/gcc/config/aarch64/tuning_models/neoversen2.h
index 21597e6b50f..abbdc98e068 100644
--- a/gcc/config/aarch64/tuning_models/neoversen2.h
+++ b/gcc/config/aarch64/tuning_models/neoversen2.h
@@ -119,7 +119,8 @@ static const sve_vec_cost neoversen2_sve_vector_cost =
12, /* gather_load_x64_cost */
42, /* gather_load_x32_init_cost */
24, /* gather_load_x64_init_cost */
- 3 /* scatter_store_elt_cost */
+ 3, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_scalar_vec_issue_info neoversen2_scalar_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/neoversen3.h
b/gcc/config/aarch64/tuning_models/neoversen3.h
index 37f9819148f..e17af8c259e 100644
--- a/gcc/config/aarch64/tuning_models/neoversen3.h
+++ b/gcc/config/aarch64/tuning_models/neoversen3.h
@@ -119,7 +119,8 @@ static const sve_vec_cost neoversen3_sve_vector_cost =
12, /* gather_load_x64_cost */
42, /* gather_load_x32_init_cost */
24, /* gather_load_x64_init_cost */
- 1 /* scatter_store_elt_cost */
+ 1, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_scalar_vec_issue_info neoversen3_scalar_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/neoversev1.h
b/gcc/config/aarch64/tuning_models/neoversev1.h
index 004676a8dba..68a82e42dc2 100644
--- a/gcc/config/aarch64/tuning_models/neoversev1.h
+++ b/gcc/config/aarch64/tuning_models/neoversev1.h
@@ -128,7 +128,8 @@ static const sve_vec_cost neoversev1_sve_vector_cost =
16, /* gather_load_x64_cost */
96, /* gather_load_x32_init_cost */
32, /* gather_load_x64_init_cost */
- 3 /* scatter_store_elt_cost */
+ 3, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_scalar_vec_issue_info neoversev1_scalar_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/neoversev2.h
b/gcc/config/aarch64/tuning_models/neoversev2.h
index 73408977251..da20d3724cf 100644
--- a/gcc/config/aarch64/tuning_models/neoversev2.h
+++ b/gcc/config/aarch64/tuning_models/neoversev2.h
@@ -121,7 +121,8 @@ static const sve_vec_cost neoversev2_sve_vector_cost =
12, /* gather_load_x64_cost */
42, /* gather_load_x32_init_cost */
24, /* gather_load_x64_init_cost */
- 3 /* scatter_store_elt_cost */
+ 3, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_scalar_vec_issue_info neoversev2_scalar_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/neoversev3.h
b/gcc/config/aarch64/tuning_models/neoversev3.h
index 7e0144032dc..a6a25a3425a 100644
--- a/gcc/config/aarch64/tuning_models/neoversev3.h
+++ b/gcc/config/aarch64/tuning_models/neoversev3.h
@@ -119,7 +119,8 @@ static const sve_vec_cost neoversev3_sve_vector_cost =
12, /* gather_load_x64_cost */
42, /* gather_load_x32_init_cost */
24, /* gather_load_x64_init_cost */
- 1 /* scatter_store_elt_cost */
+ 1, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_scalar_vec_issue_info neoversev3_scalar_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/neoversev3ae.h
b/gcc/config/aarch64/tuning_models/neoversev3ae.h
index bbe4dc54701..86484268985 100644
--- a/gcc/config/aarch64/tuning_models/neoversev3ae.h
+++ b/gcc/config/aarch64/tuning_models/neoversev3ae.h
@@ -119,7 +119,8 @@ static const sve_vec_cost neoversev3ae_sve_vector_cost =
12, /* gather_load_x64_cost */
42, /* gather_load_x32_init_cost */
24, /* gather_load_x64_init_cost */
- 1 /* scatter_store_elt_cost */
+ 1, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static const aarch64_scalar_vec_issue_info neoversev3ae_scalar_issue_info =
diff --git a/gcc/config/aarch64/tuning_models/olympus.h
b/gcc/config/aarch64/tuning_models/olympus.h
index 1e701722e22..41bd70f0147 100644
--- a/gcc/config/aarch64/tuning_models/olympus.h
+++ b/gcc/config/aarch64/tuning_models/olympus.h
@@ -90,7 +90,8 @@ static sve_vec_cost olympus_sve_vector_cost =
12, /* gather_load_x64_cost */
42, /* gather_load_x32_init_cost */
24, /* gather_load_x64_init_cost */
- 1 /* scatter_store_elt_cost */
+ 1, /* scatter_store_elt_cost */
+ 5 /* read_ffr_cost */
};
static aarch64_scalar_vec_issue_info olympus_scalar_issue_info =
--
2.43.0
