htm.md (UNSPEC_HTM_FENCE): New.

	* config/rs6000/htm.md (UNSPEC_HTM_FENCE): New.
	(tabort, tabort<wd>c, tabort<wd>ci, tbegin, tcheck, tend,
	trechkpt, treclaim, tsr, ttest): Rename define_insns from this...
	(*tabort, *tabort<wd>c, *tabort<wd>ci, *tbegin, *tcheck, *tend,
	*trechkpt, *treclaim, *tsr, *ttest): ...to this.  Add memory barrier.
	(tabort, tabort<wd>c, tabort<wd>ci, tbegin, tcheck, tend,
	trechkpt, treclaim, tsr, ttest): New define_expands.
	* config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
	__TM_FENCE__ for htm.
	* doc/extend.texi: Update documentation for htm builtins.

Co-Authored-By: Torvald Riegel <triegel@redhat.com>

From-SVN: r228827

4 files changed, 199 insertions, 24 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 8b6ed9a..b2a38da 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,17 @@
+2015-10-14  Peter Bergner  <bergner@vnet.ibm.com>
+	    Torvald Riegel  <triegel@redhat.com>
+
+	* config/rs6000/htm.md (UNSPEC_HTM_FENCE): New.
+	(tabort, tabort<wd>c, tabort<wd>ci, tbegin, tcheck, tend,
+	trechkpt, treclaim, tsr, ttest): Rename define_insns from this...
+	(*tabort, *tabort<wd>c, *tabort<wd>ci, *tbegin, *tcheck, *tend,
+	*trechkpt, *treclaim, *tsr, *ttest): ...to this.  Add memory barrier.
+	(tabort, tabort<wd>c, tabort<wd>ci, tbegin, tcheck, tend,
+	trechkpt, treclaim, tsr, ttest): New define_expands.
+	* config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
+	__TM_FENCE__ for htm.
+	* doc/extend.texi: Update documentation for htm builtins.
+
 2015-10-14  Uros Bizjak  <ubizjak@gmail.com>
 
 	PR target/67967
diff --git a/gcc/config/rs6000/htm.md b/gcc/config/rs6000/htm.md
index cec2538..098723f 100644
--- a/gcc/config/rs6000/htm.md
+++ b/gcc/config/rs6000/htm.md
@@ -27,6 +27,14 @@
   ])
 
 ;;
+;; UNSPEC usage
+;;
+
+(define_c_enum "unspec"
+  [UNSPEC_HTM_FENCE
+  ])
+
+;;
 ;; UNSPEC_VOLATILE usage
 ;;
 
@@ -45,96 +53,223 @@
    UNSPECV_HTM_MTSPR
   ])
 
+(define_expand "tabort"
+  [(parallel
+     [(set (match_operand:CC 1 "cc_reg_operand" "=x")
+	   (unspec_volatile:CC [(match_operand:SI 0 "base_reg_operand" "b")]
+			       UNSPECV_HTM_TABORT))
+      (set (match_dup 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[2] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[2]) = 1;
+})
 
-(define_insn "tabort"
+(define_insn "*tabort"
   [(set (match_operand:CC 1 "cc_reg_operand" "=x")
 	(unspec_volatile:CC [(match_operand:SI 0 "base_reg_operand" "b")]
-			    UNSPECV_HTM_TABORT))]
+			    UNSPECV_HTM_TABORT))
+   (set (match_operand:BLK 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "tabort. %0"
   [(set_attr "type" "htm")
    (set_attr "length" "4")])
 
-(define_insn "tabort<wd>c"
+(define_expand "tabort<wd>c"
+  [(parallel
+     [(set (match_operand:CC 3 "cc_reg_operand" "=x")
+	   (unspec_volatile:CC [(match_operand 0 "u5bit_cint_operand" "n")
+				(match_operand:GPR 1 "gpc_reg_operand" "r")
+				(match_operand:GPR 2 "gpc_reg_operand" "r")]
+			       UNSPECV_HTM_TABORTXC))
+      (set (match_dup 4) (unspec:BLK [(match_dup 4)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[4] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[4]) = 1;
+})
+
+(define_insn "*tabort<wd>c"
   [(set (match_operand:CC 3 "cc_reg_operand" "=x")
 	(unspec_volatile:CC [(match_operand 0 "u5bit_cint_operand" "n")
 			     (match_operand:GPR 1 "gpc_reg_operand" "r")
 			     (match_operand:GPR 2 "gpc_reg_operand" "r")]
-			    UNSPECV_HTM_TABORTXC))]
+			    UNSPECV_HTM_TABORTXC))
+   (set (match_operand:BLK 4) (unspec:BLK [(match_dup 4)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "tabort<wd>c. %0,%1,%2"
   [(set_attr "type" "htm")
    (set_attr "length" "4")])
 
-(define_insn "tabort<wd>ci"
+(define_expand "tabort<wd>ci"
+  [(parallel
+     [(set (match_operand:CC 3 "cc_reg_operand" "=x")
+	   (unspec_volatile:CC [(match_operand 0 "u5bit_cint_operand" "n")
+				(match_operand:GPR 1 "gpc_reg_operand" "r")
+				(match_operand 2 "s5bit_cint_operand" "n")]
+			       UNSPECV_HTM_TABORTXCI))
+      (set (match_dup 4) (unspec:BLK [(match_dup 4)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[4] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[4]) = 1;
+})
+
+(define_insn "*tabort<wd>ci"
   [(set (match_operand:CC 3 "cc_reg_operand" "=x")
 	(unspec_volatile:CC [(match_operand 0 "u5bit_cint_operand" "n")
 			     (match_operand:GPR 1 "gpc_reg_operand" "r")
 			     (match_operand 2 "s5bit_cint_operand" "n")]
-			    UNSPECV_HTM_TABORTXCI))]
+			    UNSPECV_HTM_TABORTXCI))
+   (set (match_operand:BLK 4) (unspec:BLK [(match_dup 4)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "tabort<wd>ci. %0,%1,%2"
   [(set_attr "type" "htm")
    (set_attr "length" "4")])
 
-(define_insn "tbegin"
+(define_expand "tbegin"
+  [(parallel
+     [(set (match_operand:CC 1 "cc_reg_operand" "=x")
+	   (unspec_volatile:CC [(match_operand 0 "const_0_to_1_operand" "n")]
+			       UNSPECV_HTM_TBEGIN))
+      (set (match_dup 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[2] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[2]) = 1;
+})
+
+(define_insn "*tbegin"
   [(set (match_operand:CC 1 "cc_reg_operand" "=x")
 	(unspec_volatile:CC [(match_operand 0 "const_0_to_1_operand" "n")]
-			    UNSPECV_HTM_TBEGIN))]
+			    UNSPECV_HTM_TBEGIN))
+   (set (match_operand:BLK 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "tbegin. %0"
   [(set_attr "type" "htm")
    (set_attr "length" "4")])
 
-(define_insn "tcheck"
+(define_expand "tcheck"
+  [(parallel
+     [(set (match_operand:CC 0 "cc_reg_operand" "=y")
+	   (unspec_volatile:CC [(const_int 0)] UNSPECV_HTM_TCHECK))
+      (set (match_dup 1) (unspec:BLK [(match_dup 1)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[1] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[1]) = 1;
+})
+
+(define_insn "*tcheck"
   [(set (match_operand:CC 0 "cc_reg_operand" "=y")
-	(unspec_volatile:CC [(const_int 0)]
-			    UNSPECV_HTM_TCHECK))]
+	(unspec_volatile:CC [(const_int 0)] UNSPECV_HTM_TCHECK))
+   (set (match_operand:BLK 1) (unspec:BLK [(match_dup 1)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "tcheck %0"
   [(set_attr "type" "htm")
    (set_attr "length" "4")])
 
-(define_insn "tend"
+(define_expand "tend"
+  [(parallel
+     [(set (match_operand:CC 1 "cc_reg_operand" "=x")
+	   (unspec_volatile:CC [(match_operand 0 "const_0_to_1_operand" "n")]
+			       UNSPECV_HTM_TEND))
+      (set (match_dup 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[2] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[2]) = 1;
+})
+
+(define_insn "*tend"
   [(set (match_operand:CC 1 "cc_reg_operand" "=x")
 	(unspec_volatile:CC [(match_operand 0 "const_0_to_1_operand" "n")]
-			    UNSPECV_HTM_TEND))]
+			    UNSPECV_HTM_TEND))
+   (set (match_operand:BLK 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "tend. %0"
   [(set_attr "type" "htm")
    (set_attr "length" "4")])
 
-(define_insn "trechkpt"
+(define_expand "trechkpt"
+  [(parallel
+     [(set (match_operand:CC 0 "cc_reg_operand" "=x")
+	   (unspec_volatile:CC [(const_int 0)] UNSPECV_HTM_TRECHKPT))
+      (set (match_dup 1) (unspec:BLK [(match_dup 1)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[1] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[1]) = 1;
+})
+
+(define_insn "*trechkpt"
   [(set (match_operand:CC 0 "cc_reg_operand" "=x")
-	(unspec_volatile:CC [(const_int 0)]
-			    UNSPECV_HTM_TRECHKPT))]
+	(unspec_volatile:CC [(const_int 0)] UNSPECV_HTM_TRECHKPT))
+   (set (match_operand:BLK 1) (unspec:BLK [(match_dup 1)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "trechkpt."
   [(set_attr "type" "htm")
    (set_attr "length" "4")])
 
-(define_insn "treclaim"
+(define_expand "treclaim"
+  [(parallel
+     [(set (match_operand:CC 1 "cc_reg_operand" "=x")
+	   (unspec_volatile:CC [(match_operand:SI 0 "gpc_reg_operand" "r")]
+			       UNSPECV_HTM_TRECLAIM))
+      (set (match_dup 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[2] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[2]) = 1;
+})
+
+(define_insn "*treclaim"
   [(set (match_operand:CC 1 "cc_reg_operand" "=x")
 	(unspec_volatile:CC [(match_operand:SI 0 "gpc_reg_operand" "r")]
-			    UNSPECV_HTM_TRECLAIM))]
+			    UNSPECV_HTM_TRECLAIM))
+   (set (match_operand:BLK 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "treclaim. %0"
   [(set_attr "type" "htm")
    (set_attr "length" "4")])
 
-(define_insn "tsr"
+(define_expand "tsr"
+  [(parallel
+     [(set (match_operand:CC 1 "cc_reg_operand" "=x")
+	   (unspec_volatile:CC [(match_operand 0 "const_0_to_1_operand" "n")]
+			       UNSPECV_HTM_TSR))
+      (set (match_dup 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[2] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[2]) = 1;
+})
+
+(define_insn "*tsr"
   [(set (match_operand:CC 1 "cc_reg_operand" "=x")
 	(unspec_volatile:CC [(match_operand 0 "const_0_to_1_operand" "n")]
-			    UNSPECV_HTM_TSR))]
+			    UNSPECV_HTM_TSR))
+   (set (match_operand:BLK 2) (unspec:BLK [(match_dup 2)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "tsr. %0"
   [(set_attr "type" "htm")
    (set_attr "length" "4")])
 
-(define_insn "ttest"
+(define_expand "ttest"
+  [(parallel
+     [(set (match_operand:CC 0 "cc_reg_operand" "=x")
+	   (unspec_volatile:CC [(const_int 0)] UNSPECV_HTM_TTEST))
+      (set (match_dup 1) (unspec:BLK [(match_dup 1)] UNSPEC_HTM_FENCE))])]
+  "TARGET_HTM"
+{
+  operands[1] = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode));
+  MEM_VOLATILE_P (operands[1]) = 1;
+})
+
+(define_insn "*ttest"
   [(set (match_operand:CC 0 "cc_reg_operand" "=x")
-	(unspec_volatile:CC [(const_int 0)]
-			    UNSPECV_HTM_TTEST))]
+	(unspec_volatile:CC [(const_int 0)] UNSPECV_HTM_TTEST))
+   (set (match_operand:BLK 1) (unspec:BLK [(match_dup 1)] UNSPEC_HTM_FENCE))]
   "TARGET_HTM"
   "tabortwci. 0,1,0"
   [(set_attr "type" "htm")
diff --git a/gcc/config/rs6000/rs6000-c.c b/gcc/config/rs6000/rs6000-c.c
index 5fc2b53..f8dfa31 100644
--- a/gcc/config/rs6000/rs6000-c.c
+++ b/gcc/config/rs6000/rs6000-c.c
@@ -372,7 +372,11 @@ rs6000_target_modify_macros (bool define_p, HOST_WIDE_INT flags,
   if ((flags & OPTION_MASK_VSX) != 0)
     rs6000_define_or_undefine_macro (define_p, "__VSX__");
   if ((flags & OPTION_MASK_HTM) != 0)
-    rs6000_define_or_undefine_macro (define_p, "__HTM__");
+    {
+      rs6000_define_or_undefine_macro (define_p, "__HTM__");
+      /* Tell the user that our HTM insn patterns act as memory barriers.  */
+      rs6000_define_or_undefine_macro (define_p, "__TM_FENCE__");
+    }
   if ((flags & OPTION_MASK_P8_VECTOR) != 0)
     rs6000_define_or_undefine_macro (define_p, "__POWER8_VECTOR__");
   if ((flags & OPTION_MASK_QUAD_MEMORY) != 0)
diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi
index 79440d3..5def659 100644
--- a/gcc/doc/extend.texi
+++ b/gcc/doc/extend.texi
@@ -16092,6 +16092,28 @@ unsigned int __builtin_tresume (void)
 unsigned int __builtin_tsuspend (void)
 @end smallexample
 
+Note that the semantics of the above HTM builtins are required to mimic
+the locking semantics used for critical sections.  Builtins that are used
+to create a new transaction or restart a suspended transaction must have
+lock acquisition like semantics while those builtins that end or suspend a
+transaction must have lock release like semantics.  Specifically, this must
+mimic lock semantics as specified by C++11, for example: Lock acquisition is
+as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
+that returns 0, and lock release is as-if an execution of
+__atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
+implicit implementation-defined lock used for all transactions.  The HTM
+instructions associated with with the builtins inherently provide the
+correct acquisition and release hardware barriers required.  However,
+the compiler must also be prohibited from moving loads and stores across
+the builtins in a way that would violate their semantics.  This has been
+accomplished by adding memory barriers to the associated HTM instructions
+(which is a conservative approach to provide acquire and release semantics).
+Earlier versions of the compiler did not treat the HTM instructions as
+memory barriers.  A @code{__TM_FENCE__} macro has been added, which can
+be used to determine whether the current compiler treats HTM instructions
+as memory barriers or not.  This allows the user to explicitly add memory
+barriers to their code when using an older version of the compiler.
+
 The following set of built-in functions are available to gain access
 to the HTM specific special purpose registers.