From 29f86ee2c44ab74c91034b209c6d5e81bc8eabb5 Mon Sep 17 00:00:00 2001 From: Kyrylo Tkachov Date: Thu, 8 Oct 2020 15:36:25 +0100 Subject: [PATCH] AArch64: Add Neoverse N2 tuning model This is the GCC 10 version of the Neoverse N2 tuning struct patch. It's more or less identical. Bootstrapped and tested on the branch. gcc/ * config/aarch64/aarch64.c (neoversen2_tunings): Define. * config/aarch64/aarch64-cores.def (neoverse-n2): Use it. --- gcc/config/aarch64/aarch64-cores.def | 2 +- gcc/config/aarch64/aarch64.c | 26 ++++++++++++++++++++++++++ 2 files changed, 27 insertions(+), 1 deletion(-) diff --git a/gcc/config/aarch64/aarch64-cores.def b/gcc/config/aarch64/aarch64-cores.def index ac528285adb1..9c290292479b 100644 --- a/gcc/config/aarch64/aarch64-cores.def +++ b/gcc/config/aarch64/aarch64-cores.def @@ -140,7 +140,7 @@ AARCH64_CORE("neoverse-v1", neoversev1, cortexa57, 8_4A, AARCH64_FL_FOR_ARCH8_4 AARCH64_CORE("saphira", saphira, saphira, 8_4A, AARCH64_FL_FOR_ARCH8_4 | AARCH64_FL_CRYPTO | AARCH64_FL_RCPC, saphira, 0x51, 0xC01, -1) /* Armv8.5-A Architecture Processors. */ -AARCH64_CORE("neoverse-n2", neoversen2, cortexa57, 8_5A, AARCH64_FL_FOR_ARCH8_5 | AARCH64_FL_I8MM | AARCH64_FL_BF16 | AARCH64_FL_F16 | AARCH64_FL_SVE | AARCH64_FL_SVE2 | AARCH64_FL_SVE2_BITPERM | AARCH64_FL_RNG | AARCH64_FL_MEMTAG, neoversen1, 0x41, 0xd49, -1) +AARCH64_CORE("neoverse-n2", neoversen2, cortexa57, 8_5A, AARCH64_FL_FOR_ARCH8_5 | AARCH64_FL_I8MM | AARCH64_FL_BF16 | AARCH64_FL_F16 | AARCH64_FL_SVE | AARCH64_FL_SVE2 | AARCH64_FL_SVE2_BITPERM | AARCH64_FL_RNG | AARCH64_FL_MEMTAG, neoversen2, 0x41, 0xd49, -1) /* ARMv8-A big.LITTLE implementations. */ diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c index fdfe00026f73..df973165165e 100644 --- a/gcc/config/aarch64/aarch64.c +++ b/gcc/config/aarch64/aarch64.c @@ -1362,6 +1362,32 @@ static const struct tune_params neoversev1_tunings = &generic_prefetch_tune }; +static const struct tune_params neoversen2_tunings = +{ + &cortexa57_extra_costs, + &generic_addrcost_table, + &generic_regmove_cost, + &cortexa57_vector_cost, + &generic_branch_cost, + &generic_approx_modes, + SVE_128, /* sve_width */ + 4, /* memmov_cost */ + 3, /* issue_rate */ + (AARCH64_FUSE_AES_AESMC | AARCH64_FUSE_CMP_BRANCH), /* fusible_ops */ + "32:16", /* function_align. */ + "4", /* jump_align. */ + "32:16", /* loop_align. */ + 2, /* int_reassoc_width. */ + 4, /* fp_reassoc_width. */ + 2, /* vec_reassoc_width. */ + 2, /* min_div_recip_mul_sf. */ + 2, /* min_div_recip_mul_df. */ + 0, /* max_case_values. */ + tune_params::AUTOPREFETCHER_WEAK, /* autoprefetcher_model. */ + (AARCH64_EXTRA_TUNE_NONE), /* tune_flags. */ + &generic_prefetch_tune +}; + static const struct tune_params a64fx_tunings = { &generic_extra_costs, -- 2.47.2