public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Xi Ruoyao <xry111@xry111.site>
To: gcc-patches@gcc.gnu.org
Cc: chenglulu <chenglulu@loongson.cn>,
	i@xen0n.name, xuchenghua@loongson.cn, hev <wangrui@loongson.cn>,
	Xi Ruoyao <xry111@xry111.site>
Subject: [PATCH] LoongArch: Remove redundant barrier instructions before LL-SC loops
Date: Mon,  6 Nov 2023 19:36:04 +0800	[thread overview]
Message-ID: <20231106113809.1193236-1-xry111@xry111.site> (raw)

This is isomorphic to the LLVM changes [1-2].

On LoongArch, the LL and SC instructions has memory barrier semantics:

- LL: <memory-barrier> + <load-exclusive>
- SC: <store-conditional> + <memory-barrier>

But the compare and swap operation is allowed to fail, and if it fails
the SC instruction is not executed, thus the guarantee of acquiring
semantics cannot be ensured. Therefore, an acquire barrier needs to be
generated when failure_memorder includes an acquire operation.

On CPUs implementing LoongArch v1.10 or later, "dbar 0b10100" is an
acquire barrier; on CPUs implementing LoongArch v1.00, it is a full
barrier.  So it's always enough for acquire semantics.  OTOH if an
acquire semantic is not needed, we still needs the "dbar 0x700" as the
load-load barrier like all LL-SC loops.

[1]:https://github.com/llvm/llvm-project/pull/67391
[2]:https://github.com/llvm/llvm-project/pull/69339

gcc/ChangeLog:

	* config/loongarch/loongarch.cc
	(loongarch_memmodel_needs_release_fence): Remove.
	(loongarch_cas_failure_memorder_needs_acquire): New static
	function.
	(loongarch_print_operand): Redefine 'G' for the barrier on CAS
	failure.
	* config/loongarch/sync.md (atomic_cas_value_strong<mode>):
	Remove the redundant barrier before the LL instruction, and
	emit an acquire barrier on failure if needed by
	failure_memorder.
	(atomic_cas_value_cmp_and_7_<mode>): Likewise.
	(atomic_cas_value_add_7_<mode>): Remove the unnecessary barrier
	before the LL instruction.
	(atomic_cas_value_sub_7_<mode>): Likewise.
	(atomic_cas_value_and_7_<mode>): Likewise.
	(atomic_cas_value_xor_7_<mode>): Likewise.
	(atomic_cas_value_or_7_<mode>): Likewise.
	(atomic_cas_value_nand_7_<mode>): Likewise.
	(atomic_cas_value_exchange_7_<mode>): Likewise.

gcc/testsuite/ChangeLog:

	* gcc.target/loongarch/cas-acquire.c: New test.
---

Bootstrapped and regtested on loongarch64-linux-gnu.  Ok for trunk
and/or GCC 12/13 (for fixing the acquire semantics in failure_memorder)?

 gcc/config/loongarch/loongarch.cc             | 27 +++---
 gcc/config/loongarch/sync.md                  | 49 +++++------
 .../gcc.target/loongarch/cas-acquire.c        | 84 +++++++++++++++++++
 3 files changed, 118 insertions(+), 42 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/loongarch/cas-acquire.c

diff --git a/gcc/config/loongarch/loongarch.cc b/gcc/config/loongarch/loongarch.cc
index 9b63f0dc322..d9b7a1076a2 100644
--- a/gcc/config/loongarch/loongarch.cc
+++ b/gcc/config/loongarch/loongarch.cc
@@ -5833,25 +5833,22 @@ loongarch_memmodel_needs_rel_acq_fence (enum memmodel model)
     }
 }
 
-/* Return true if a FENCE should be emitted to before a memory access to
-   implement the release portion of memory model MODEL.  */
+/* Return true if a FENCE should be emitted after a failed CAS to
+   implement the acquire semantic of failure_memorder.  */
 
 static bool
-loongarch_memmodel_needs_release_fence (enum memmodel model)
+loongarch_cas_failure_memorder_needs_acquire (enum memmodel model)
 {
-  switch (model)
+  switch (memmodel_base (model))
     {
+    case MEMMODEL_ACQUIRE:
     case MEMMODEL_ACQ_REL:
+    case MEMMODEL_CONSUME:
     case MEMMODEL_SEQ_CST:
-    case MEMMODEL_SYNC_SEQ_CST:
-    case MEMMODEL_RELEASE:
-    case MEMMODEL_SYNC_RELEASE:
       return true;
 
-    case MEMMODEL_ACQUIRE:
-    case MEMMODEL_CONSUME:
-    case MEMMODEL_SYNC_ACQUIRE:
     case MEMMODEL_RELAXED:
+    case MEMMODEL_RELEASE:
       return false;
 
     default:
@@ -5966,7 +5963,8 @@ loongarch_print_operand_reloc (FILE *file, rtx op, bool hi64_part,
    'd'	Print CONST_INT OP in decimal.
    'E'	Print CONST_INT OP element 0 of a replicated CONST_VECTOR in decimal.
    'F'	Print the FPU branch condition for comparison OP.
-   'G'	Print a DBAR insn if the memory model requires a release.
+   'G'	Print a DBAR insn for CAS failure (with an acquire semantic if
+	needed, otherwise a simple load-load barrier).
    'H'  Print address 52-61bit relocation associated with OP.
    'h'  Print the high-part relocation associated with OP.
    'i'	Print i if the operand is not a register.
@@ -6057,8 +6055,11 @@ loongarch_print_operand (FILE *file, rtx op, int letter)
       break;
 
     case 'G':
-      if (loongarch_memmodel_needs_release_fence ((enum memmodel) INTVAL (op)))
-	fputs ("dbar\t0", file);
+      if (loongarch_cas_failure_memorder_needs_acquire (
+	    memmodel_from_int (INTVAL (op))))
+	fputs ("dbar\t0b10100", file);
+      else
+	fputs ("dbar\t0x700", file);
       break;
 
     case 'h':
diff --git a/gcc/config/loongarch/sync.md b/gcc/config/loongarch/sync.md
index 9924d522bcd..db3a21690b8 100644
--- a/gcc/config/loongarch/sync.md
+++ b/gcc/config/loongarch/sync.md
@@ -129,19 +129,18 @@ (define_insn "atomic_cas_value_strong<mode>"
    (clobber (match_scratch:GPR 6 "=&r"))]
   ""
 {
-  return "%G5\\n\\t"
-	 "1:\\n\\t"
+  return "1:\\n\\t"
 	 "ll.<amo>\\t%0,%1\\n\\t"
 	 "bne\\t%0,%z2,2f\\n\\t"
 	 "or%i3\\t%6,$zero,%3\\n\\t"
 	 "sc.<amo>\\t%6,%1\\n\\t"
-	 "beq\\t$zero,%6,1b\\n\\t"
+	 "beqz\\t%6,1b\\n\\t"
 	 "b\\t3f\\n\\t"
 	 "2:\\n\\t"
-	 "dbar\\t0x700\\n\\t"
+	 "%G5\\n\\t"
 	 "3:\\n\\t";
 }
-  [(set (attr "length") (const_int 32))])
+  [(set (attr "length") (const_int 28))])
 
 (define_expand "atomic_compare_and_swap<mode>"
   [(match_operand:SI 0 "register_operand" "")   ;; bool output
@@ -234,8 +233,7 @@ (define_insn "atomic_cas_value_cmp_and_7_<mode>"
    (clobber (match_scratch:GPR 7 "=&r"))]
   ""
 {
-  return "%G6\\n\\t"
-	 "1:\\n\\t"
+  return "1:\\n\\t"
 	 "ll.<amo>\\t%0,%1\\n\\t"
 	 "and\\t%7,%0,%2\\n\\t"
 	 "bne\\t%7,%z4,2f\\n\\t"
@@ -245,10 +243,10 @@ (define_insn "atomic_cas_value_cmp_and_7_<mode>"
 	 "beq\\t$zero,%7,1b\\n\\t"
 	 "b\\t3f\\n\\t"
 	 "2:\\n\\t"
-	 "dbar\\t0x700\\n\\t"
+	 "%G6\\n\\t"
 	 "3:\\n\\t";
 }
-  [(set (attr "length") (const_int 40))])
+  [(set (attr "length") (const_int 36))])
 
 (define_expand "atomic_compare_and_swap<mode>"
   [(match_operand:SI 0 "register_operand" "")   ;; bool output
@@ -303,8 +301,7 @@ (define_insn "atomic_cas_value_add_7_<mode>"
    (clobber (match_scratch:GPR 8 "=&r"))]
   ""
 {
-  return "%G6\\n\\t"
-	 "1:\\n\\t"
+  return "1:\\n\\t"
 	 "ll.<amo>\\t%0,%1\\n\\t"
 	 "and\\t%7,%0,%3\\n\\t"
 	 "add.w\\t%8,%0,%z5\\n\\t"
@@ -314,7 +311,7 @@ (define_insn "atomic_cas_value_add_7_<mode>"
 	 "beq\\t$zero,%7,1b";
 }
 
-  [(set (attr "length") (const_int 32))])
+  [(set (attr "length") (const_int 28))])
 
 (define_insn "atomic_cas_value_sub_7_<mode>"
   [(set (match_operand:GPR 0 "register_operand" "=&r")				;; res
@@ -330,8 +327,7 @@ (define_insn "atomic_cas_value_sub_7_<mode>"
    (clobber (match_scratch:GPR 8 "=&r"))]
   ""
 {
-  return "%G6\\n\\t"
-	 "1:\\n\\t"
+  return "1:\\n\\t"
 	 "ll.<amo>\\t%0,%1\\n\\t"
 	 "and\\t%7,%0,%3\\n\\t"
 	 "sub.w\\t%8,%0,%z5\\n\\t"
@@ -340,7 +336,7 @@ (define_insn "atomic_cas_value_sub_7_<mode>"
 	 "sc.<amo>\\t%7,%1\\n\\t"
 	 "beq\\t$zero,%7,1b";
 }
-  [(set (attr "length") (const_int 32))])
+  [(set (attr "length") (const_int 28))])
 
 (define_insn "atomic_cas_value_and_7_<mode>"
   [(set (match_operand:GPR 0 "register_operand" "=&r")				;; res
@@ -356,8 +352,7 @@ (define_insn "atomic_cas_value_and_7_<mode>"
    (clobber (match_scratch:GPR 8 "=&r"))]
   ""
 {
-  return "%G6\\n\\t"
-	 "1:\\n\\t"
+  return "1:\\n\\t"
 	 "ll.<amo>\\t%0,%1\\n\\t"
 	 "and\\t%7,%0,%3\\n\\t"
 	 "and\\t%8,%0,%z5\\n\\t"
@@ -366,7 +361,7 @@ (define_insn "atomic_cas_value_and_7_<mode>"
 	 "sc.<amo>\\t%7,%1\\n\\t"
 	 "beq\\t$zero,%7,1b";
 }
-  [(set (attr "length") (const_int 32))])
+  [(set (attr "length") (const_int 28))])
 
 (define_insn "atomic_cas_value_xor_7_<mode>"
   [(set (match_operand:GPR 0 "register_operand" "=&r")				;; res
@@ -382,8 +377,7 @@ (define_insn "atomic_cas_value_xor_7_<mode>"
    (clobber (match_scratch:GPR 8 "=&r"))]
   ""
 {
-  return "%G6\\n\\t"
-	 "1:\\n\\t"
+  return "1:\\n\\t"
 	 "ll.<amo>\\t%0,%1\\n\\t"
 	 "and\\t%7,%0,%3\\n\\t"
 	 "xor\\t%8,%0,%z5\\n\\t"
@@ -393,7 +387,7 @@ (define_insn "atomic_cas_value_xor_7_<mode>"
 	 "beq\\t$zero,%7,1b";
 }
 
-  [(set (attr "length") (const_int 32))])
+  [(set (attr "length") (const_int 28))])
 
 (define_insn "atomic_cas_value_or_7_<mode>"
   [(set (match_operand:GPR 0 "register_operand" "=&r")				;; res
@@ -409,8 +403,7 @@ (define_insn "atomic_cas_value_or_7_<mode>"
    (clobber (match_scratch:GPR 8 "=&r"))]
   ""
 {
-  return "%G6\\n\\t"
-	 "1:\\n\\t"
+  return "1:\\n\\t"
 	 "ll.<amo>\\t%0,%1\\n\\t"
 	 "and\\t%7,%0,%3\\n\\t"
 	 "or\\t%8,%0,%z5\\n\\t"
@@ -420,7 +413,7 @@ (define_insn "atomic_cas_value_or_7_<mode>"
 	 "beq\\t$zero,%7,1b";
 }
 
-  [(set (attr "length") (const_int 32))])
+  [(set (attr "length") (const_int 28))])
 
 (define_insn "atomic_cas_value_nand_7_<mode>"
   [(set (match_operand:GPR 0 "register_operand" "=&r")				;; res
@@ -436,8 +429,7 @@ (define_insn "atomic_cas_value_nand_7_<mode>"
    (clobber (match_scratch:GPR 8 "=&r"))]
   ""
 {
-  return "%G6\\n\\t"
-	 "1:\\n\\t"
+  return "1:\\n\\t"
 	 "ll.<amo>\\t%0,%1\\n\\t"
 	 "and\\t%7,%0,%3\\n\\t"
 	 "and\\t%8,%0,%z5\\n\\t"
@@ -446,7 +438,7 @@ (define_insn "atomic_cas_value_nand_7_<mode>"
 	 "sc.<amo>\\t%7,%1\\n\\t"
 	 "beq\\t$zero,%7,1b";
 }
-  [(set (attr "length") (const_int 32))])
+  [(set (attr "length") (const_int 28))])
 
 (define_insn "atomic_cas_value_exchange_7_<mode>"
   [(set (match_operand:GPR 0 "register_operand" "=&r")
@@ -461,8 +453,7 @@ (define_insn "atomic_cas_value_exchange_7_<mode>"
    (clobber (match_scratch:GPR 7 "=&r"))]
   ""
 {
-  return "%G6\\n\\t"
-	 "1:\\n\\t"
+  return "1:\\n\\t"
 	 "ll.<amo>\\t%0,%1\\n\\t"
 	 "and\\t%7,%0,%z3\\n\\t"
 	 "or%i5\\t%7,%7,%5\\n\\t"
diff --git a/gcc/testsuite/gcc.target/loongarch/cas-acquire.c b/gcc/testsuite/gcc.target/loongarch/cas-acquire.c
new file mode 100644
index 00000000000..cd3ba06e258
--- /dev/null
+++ b/gcc/testsuite/gcc.target/loongarch/cas-acquire.c
@@ -0,0 +1,84 @@
+/* { dg-do run } */
+/* { dg-require-effective-target c99_runtime } */
+/* { dg-require-effective-target pthread } */
+/* { dg-options "-std=c99 -pthread" } */
+
+/* https://github.com/llvm/llvm-project/pull/67391#issuecomment-1752403934
+   reported that this had failed with GCC and 3A6000.  */
+
+#include <pthread.h>
+#include <stdatomic.h>
+#include <stdbool.h>
+#include <stdlib.h>
+
+static unsigned int tags[32];
+static unsigned int vals[32];
+
+static void *
+writer_entry (void *data)
+{
+  atomic_uint *pt = (atomic_uint *)tags;
+  atomic_uint *pv = (atomic_uint *)vals;
+  unsigned int n;
+
+  for (n = 1; n < 10000000; n++)
+    {
+      atomic_store_explicit (&pv[n & 31], n, memory_order_release);
+      atomic_store_explicit (&pt[n & 31], n, memory_order_release);
+      n++;
+    }
+
+  return NULL;
+}
+
+static void *
+reader_entry (void *data)
+{
+  atomic_uint *pt = (atomic_uint *)tags;
+  atomic_uint *pv = (atomic_uint *)vals;
+  int i;
+
+  for (;;)
+    {
+      for (i = 0; i < 32; i++)
+        {
+          unsigned int tag = 0;
+          bool res;
+
+          res = atomic_compare_exchange_weak_explicit (
+              &pt[i], &tag, 0, memory_order_acquire, memory_order_acquire);
+          if (!res)
+            {
+              unsigned int val;
+
+              val = atomic_load_explicit (&pv[i], memory_order_relaxed);
+              if (val < tag)
+                abort ();
+            }
+        }
+    }
+
+  return NULL;
+}
+
+int
+main (int argc, char *argv[])
+{
+  pthread_t writer;
+  pthread_t reader;
+  int res;
+
+  res = pthread_create (&writer, NULL, writer_entry, NULL);
+  if (res < 0)
+    abort ();
+
+  res = pthread_create (&reader, NULL, reader_entry, NULL);
+  if (res < 0)
+    abort ();
+
+  res = pthread_join (writer, NULL);
+  if (res < 0)
+    abort ();
+
+  return 0;
+}
-- 
2.42.1


             reply	other threads:[~2023-11-06 11:40 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-11-06 11:36 Xi Ruoyao [this message]
2023-11-08  1:49 ` chenglulu
2023-11-08  8:27   ` Xi Ruoyao
2023-11-13 15:32     ` Xi Ruoyao

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20231106113809.1193236-1-xry111@xry111.site \
    --to=xry111@xry111.site \
    --cc=chenglulu@loongson.cn \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=i@xen0n.name \
    --cc=wangrui@loongson.cn \
    --cc=xuchenghua@loongson.cn \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).