This patch generalizes the code alignment and lets each CPU set function, jump and loop alignment independently. The defaults for A53/A57 are based the original patch by James Greenhalgh.
OK for trunk? ChangeLog: 2014-12-13 Wilco Dijkstra wdijk...@arm.com * gcc/config/aarch64/aarch64-protos.h (tune-params): Add code alignment tuning parameters. * gcc/config/aarch64/aarch64.c (generic_tunings) Add code alignment tuning parameters. (cortexa53_tunings): Likewise. (cortexa57_tunings): Likewise. (thunderx_tunings): Likewise. (aarch64_override_options): Use new alignment tunings. --- gcc/config/aarch64/aarch64-protos.h | 4 +++- gcc/config/aarch64/aarch64.c | 22 +++++++++++++++------- 2 files changed, 18 insertions(+), 8 deletions(-) diff --git a/gcc/config/aarch64/aarch64-protos.h b/gcc/config/aarch64/aarch64-protos.h index 234efcb..f22573b 100644 --- a/gcc/config/aarch64/aarch64-protos.h +++ b/gcc/config/aarch64/aarch64-protos.h @@ -170,8 +170,10 @@ struct tune_params const struct cpu_vector_cost *const vec_costs; const int memmov_cost; const int issue_rate; - const int align; const unsigned int fuseable_ops; + const int function_align; + const int jump_align; + const int loop_align; const int int_reassoc_width; const int fp_reassoc_width; const int vec_reassoc_width; diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c index 5f51b97..db42164 100644 --- a/gcc/config/aarch64/aarch64.c +++ b/gcc/config/aarch64/aarch64.c @@ -323,8 +323,10 @@ static const struct tune_params generic_tunings = &generic_vector_cost, NAMED_PARAM (memmov_cost, 4), NAMED_PARAM (issue_rate, 2), - NAMED_PARAM (align, 4), NAMED_PARAM (fuseable_ops, AARCH64_FUSE_NOTHING), + 8, /* function_align. */ + 8, /* jump_align. */ + 4, /* loop_align. */ 2, /* int_reassoc_width. */ 4, /* fp_reassoc_width. */ 1 /* vec_reassoc_width. */ @@ -338,9 +340,11 @@ static const struct tune_params cortexa53_tunings = &generic_vector_cost, NAMED_PARAM (memmov_cost, 4), NAMED_PARAM (issue_rate, 2), - NAMED_PARAM (align, 8), NAMED_PARAM (fuseable_ops, (AARCH64_FUSE_MOV_MOVK | AARCH64_FUSE_ADRP_ADD | AARCH64_FUSE_MOVK_MOVK | AARCH64_FUSE_ADRP_LDR)), + 8, /* function_align. */ + 8, /* jump_align. */ + 4, /* loop_align. */ 2, /* int_reassoc_width. */ 4, /* fp_reassoc_width. */ 1 /* vec_reassoc_width. */ @@ -354,8 +358,10 @@ static const struct tune_params cortexa57_tunings = &cortexa57_vector_cost, NAMED_PARAM (memmov_cost, 4), NAMED_PARAM (issue_rate, 3), - NAMED_PARAM (align, 8), NAMED_PARAM (fuseable_ops, (AARCH64_FUSE_MOV_MOVK | AARCH64_FUSE_ADRP_ADD | AARCH64_FUSE_MOVK_MOVK)), + 16, /* function_align. */ + 8, /* jump_align. */ + 4, /* loop_align. */ 2, /* int_reassoc_width. */ 4, /* fp_reassoc_width. */ 1 /* vec_reassoc_width. */ @@ -369,8 +375,10 @@ static const struct tune_params thunderx_tunings = &generic_vector_cost, NAMED_PARAM (memmov_cost, 6), NAMED_PARAM (issue_rate, 2), - NAMED_PARAM (align, 8), NAMED_PARAM (fuseable_ops, AARCH64_FUSE_CMP_BRANCH), + 8, /* function_align. */ + 8, /* jump_align. */ + 8, /* loop_align. */ 2, /* int_reassoc_width. */ 4, /* fp_reassoc_width. */ 1 /* vec_reassoc_width. */ @@ -6773,11 +6781,11 @@ aarch64_override_options (void) if (!optimize_size) { if (align_loops <= 0) - align_loops = aarch64_tune_params->align; + align_loops = aarch64_tune_params->loop_align; if (align_jumps <= 0) - align_jumps = aarch64_tune_params->align; + align_jumps = aarch64_tune_params->jump_align; if (align_functions <= 0) - align_functions = aarch64_tune_params->align; + align_functions = aarch64_tune_params->function_align; } aarch64_override_options_after_change (); -- 1.9.1