https://gcc.gnu.org/g:8f0ff6b998748f3581e0f06e3108193866b1209d
commit r14-9824-g8f0ff6b998748f3581e0f06e3108193866b1209d
Author: Lulu Cheng <chengl...@loongson.cn>
Date: Tue Apr 2 14:29:08 2024 +0800
LoongArch: Set default alignment for functions jumps and loops [PR112919].
Xi Ruoyao set the alignment rules under LA464 in commit r14-1839,
but the macro ASM_OUTPUT_ALIGN_WITH_NOP was removed in R14-4674,
which affected the alignment rules.
So I set different aligns on LA464 and LA664 again to test the
performance of spec2006, and modify the alignment based on the test
results.
gcc/ChangeLog:
PR target/112919
* config/loongarch/loongarch-def.cc (la664_align): Newly defined
function that sets alignment rules under the LA664
microarchitecture.
* config/loongarch/loongarch-opts.cc
(loongarch_target_option_override): If not optimizing for size, set
the default alignment to what the target wants.
* config/loongarch/loongarch-tune.h (struct loongarch_align): Add
new member variables jump and loop.
Diff:
---
gcc/config/loongarch/loongarch-def.cc | 11 ++++++++---
gcc/config/loongarch/loongarch-opts.cc | 19 +++++++++++++------
gcc/config/loongarch/loongarch-tune.h | 22 +++++++++++++++-------
3 files changed, 36 insertions(+), 16 deletions(-)
diff --git a/gcc/config/loongarch/loongarch-def.cc
b/gcc/config/loongarch/loongarch-def.cc
index e8c129ce643..63a8f108f4e 100644
--- a/gcc/config/loongarch/loongarch-def.cc
+++ b/gcc/config/loongarch/loongarch-def.cc
@@ -81,14 +81,19 @@ array_tune<loongarch_cache> loongarch_cpu_cache =
static inline loongarch_align la464_align ()
{
- return loongarch_align ().function_ ("32").label_ ("16");
+ return loongarch_align ().function_ ("32").loop_ ("16").jump_ ("16");
+}
+
+static inline loongarch_align la664_align ()
+{
+ return loongarch_align ().function_ ("8").loop_ ("8").jump_ ("32");
}
array_tune<loongarch_align> loongarch_cpu_align =
array_tune<loongarch_align> ()
- .set (CPU_LOONGARCH64, la464_align ())
+ .set (CPU_LOONGARCH64, la664_align ())
.set (CPU_LA464, la464_align ())
- .set (CPU_LA664, la464_align ());
+ .set (CPU_LA664, la664_align ());
/* Default RTX cost initializer. */
loongarch_rtx_cost_data::loongarch_rtx_cost_data ()
diff --git a/gcc/config/loongarch/loongarch-opts.cc
b/gcc/config/loongarch/loongarch-opts.cc
index 2a6fc41b247..7b21cc311a8 100644
--- a/gcc/config/loongarch/loongarch-opts.cc
+++ b/gcc/config/loongarch/loongarch-opts.cc
@@ -922,13 +922,20 @@ loongarch_target_option_override (struct loongarch_target
*target,
{
loongarch_update_gcc_opt_status (target, opts, opts_set);
- /* alignments */
- if (opts->x_flag_align_functions && !opts->x_str_align_functions)
- opts->x_str_align_functions
- = loongarch_cpu_align[target->cpu_tune].function;
+ /* If not optimizing for size, set the default
+ alignment to what the target wants. */
+ if (!opts->x_optimize_size)
+ {
+ if (opts->x_flag_align_functions && !opts->x_str_align_functions)
+ opts->x_str_align_functions
+ = loongarch_cpu_align[target->cpu_tune].function;
+
+ if (opts->x_flag_align_loops && !opts->x_str_align_loops)
+ opts->x_str_align_loops = loongarch_cpu_align[target->cpu_tune].loop;
- if (opts->x_flag_align_labels && !opts->x_str_align_labels)
- opts->x_str_align_labels = loongarch_cpu_align[target->cpu_tune].label;
+ if (opts->x_flag_align_jumps && !opts->x_str_align_jumps)
+ opts->x_str_align_jumps = loongarch_cpu_align[target->cpu_tune].jump;
+ }
/* Set up parameters to be used in prefetching algorithm. */
int simultaneous_prefetches
diff --git a/gcc/config/loongarch/loongarch-tune.h
b/gcc/config/loongarch/loongarch-tune.h
index 72b75f6de3f..cec9deb541a 100644
--- a/gcc/config/loongarch/loongarch-tune.h
+++ b/gcc/config/loongarch/loongarch-tune.h
@@ -162,14 +162,16 @@ struct loongarch_cache {
}
};
-/* Alignment for functions and labels for best performance. For new uarchs
- the value should be measured via benchmarking. See the documentation for
- -falign-functions and -falign-labels in invoke.texi for the format. */
+/* Alignment for functions loops and jumps for best performance. For new
+ uarchs the value should be measured via benchmarking. See the
+ documentation for -falign-functions, -falign-loops, and -falign-jumps in
+ invoke.texi for the format. */
struct loongarch_align {
const char *function; /* default value for -falign-functions */
- const char *label; /* default value for -falign-labels */
+ const char *loop; /* default value for -falign-loops */
+ const char *jump; /* default value for -falign-jumps */
- loongarch_align () : function (nullptr), label (nullptr) {}
+ loongarch_align () : function (nullptr), loop (nullptr), jump (nullptr) {}
loongarch_align function_ (const char *_function)
{
@@ -177,9 +179,15 @@ struct loongarch_align {
return *this;
}
- loongarch_align label_ (const char *_label)
+ loongarch_align loop_ (const char *_loop)
{
- label = _label;
+ loop = _loop;
+ return *this;
+ }
+
+ loongarch_align jump_ (const char *_jump)
+ {
+ jump = _jump;
return *this;
}
};
