[ARM] Implement support for ACLE Coprocessor LDC and STC intrinsics

gcc/ChangeLog:
2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>

	* config/arm/arm.md (*ldc): New.
	(*stc): New.
	(<ldc>): New.
	(<stc>): New.
	* config/arm/arm.c (arm_coproc_builtin_available): Add
	support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
	(arm_coproc_ldc_stc_legitimate_address): New.
	* config/arm/arm-builtins.c (arm_type_qualifiers): Add
	'qualifier_const_pointer'.
	(LDC_QUALIFIERS): Define to...
	(arm_ldc_qualifiers): ... this. New.
	(STC_QUALIFIERS): Define to...
	(arm_stc_qualifiers): ... this. New.
	* config/arm/arm-protos.h
	(arm_coproc_ldc_stc_legitimate_address): New.
	* config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
	__arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
	* config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
	stc2, stcl, stc2l): New.
	* config/arm/constraints.md (Uz): New.
	* config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
	* config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
	VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
	VUNSPEC_STC2L): New.

gcc/testsuite/ChangeLog:
2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>

	* gcc.target/arm/acle/ldc: New.
	* gcc.target/arm/acle/ldc2: New.
	* gcc.target/arm/acle/ldcl: New.
	* gcc.target/arm/acle/ldc2l: New.
	* gcc.target/arm/acle/stc: New.
	* gcc.target/arm/acle/stc2: New.
	* gcc.target/arm/acle/stcl: New.
	* gcc.target/arm/acle/stc2l: New.

From-SVN: r244173

gcc/ChangeLog

 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
 
+	* config/arm/arm.md (*ldc): New.
+	(*stc): New.
+	(<ldc>): New.
+	(<stc>): New.
+	* config/arm/arm.c (arm_coproc_builtin_available): Add
+	support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
+	(arm_coproc_ldc_stc_legitimate_address): New.
+	* config/arm/arm-builtins.c (arm_type_qualifiers): Add
+	'qualifier_const_pointer'.
+	(LDC_QUALIFIERS): Define to...
+	(arm_ldc_qualifiers): ... this. New.
+	(STC_QUALIFIERS): Define to...
+	(arm_stc_qualifiers): ... this. New.
+	* config/arm/arm-protos.h
+	(arm_coproc_ldc_stc_legitimate_address): New.
+	* config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
+	__arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
+	* config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
+	stc2, stcl, stc2l): New.
+	* config/arm/constraints.md (Uz): New.
+	* config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
+	* config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
+	VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
+	VUNSPEC_STC2L): New.
+
+2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
+
 	* config/arm/arm.md (<cdp>): New.
 	* config/arm/arm.c (neon_const_bounds): Rename this ...
 	(arm_const_bounds): ... this.

gcc/config/arm/arm-builtins.c

@@ -51,6 +51,8 @@ enum arm_type_qualifiers
   qualifier_const = 0x2, /* 1 << 1  */
   /* T *foo.  */
   qualifier_pointer = 0x4, /* 1 << 2  */
+  /* const T * foo.  */
+  qualifier_const_pointer = 0x6,
   /* Used when expanding arguments if an operand could
      be an immediate.  */
   qualifier_immediate = 0x8, /* 1 << 3  */
@@ -178,6 +180,23 @@ arm_cdp_qualifiers[SIMD_MAX_BUILTIN_ARGS]
       qualifier_unsigned_immediate };
 #define CDP_QUALIFIERS \
   (arm_cdp_qualifiers)
+
+/* void (unsigned immediate, unsigned immediate,  const void *).  */
+static enum arm_type_qualifiers
+arm_ldc_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_void, qualifier_unsigned_immediate,
+      qualifier_unsigned_immediate, qualifier_const_pointer };
+#define LDC_QUALIFIERS \
+  (arm_ldc_qualifiers)
+
+/* void (unsigned immediate, unsigned immediate,  void *).  */
+static enum arm_type_qualifiers
+arm_stc_qualifiers[SIMD_MAX_BUILTIN_ARGS]
+  = { qualifier_void, qualifier_unsigned_immediate,
+      qualifier_unsigned_immediate, qualifier_pointer };
+#define STC_QUALIFIERS \
+  (arm_stc_qualifiers)
+
 /* The first argument (return type) of a store should be void type,
    which we represent with qualifier_void.  Their first operand will be
    a DImode pointer to the location to store to, so we must use

gcc/config/arm/arm-protos.h

@@ -177,6 +177,7 @@ extern void arm_split_compare_and_swap (rtx op[]);
 extern void arm_split_atomic_op (enum rtx_code, rtx, rtx, rtx, rtx, rtx, rtx);
 extern rtx arm_load_tp (rtx);
 extern bool arm_coproc_builtin_available (enum unspecv);
+extern bool arm_coproc_ldc_stc_legitimate_address (rtx);
 
 #if defined TREE_CODE
 extern void arm_init_cumulative_args (CUMULATIVE_ARGS *, tree, rtx, tree);

gcc/config/arm/arm.c

@@ -30904,10 +30904,18 @@ arm_coproc_builtin_available (enum unspecv builtin)
   switch (builtin)
     {
       case VUNSPEC_CDP:
+      case VUNSPEC_LDC:
+      case VUNSPEC_LDCL:
+      case VUNSPEC_STC:
+      case VUNSPEC_STCL:
 	if (arm_arch4)
 	  return true;
 	break;
       case VUNSPEC_CDP2:
+      case VUNSPEC_LDC2:
+      case VUNSPEC_LDC2L:
+      case VUNSPEC_STC2:
+      case VUNSPEC_STC2L:
 	/* Only present in ARMv5*, ARMv6 (but not ARMv6-M), ARMv7* and
 	   ARMv8-{A,M}.  */
 	if (arm_arch5)
@@ -30918,4 +30926,55 @@ arm_coproc_builtin_available (enum unspecv builtin)
     }
   return false;
 }
+
+/* This function returns true if OP is a valid memory operand for the ldc and
+   stc coprocessor instructions and false otherwise.  */
+
+bool
+arm_coproc_ldc_stc_legitimate_address (rtx op)
+{
+  HOST_WIDE_INT range;
+  /* Has to be a memory operand.  */
+  if (!MEM_P (op))
+    return false;
+
+  op = XEXP (op, 0);
+
+  /* We accept registers.  */
+  if (REG_P (op))
+    return true;
+
+  switch GET_CODE (op)
+    {
+      case PLUS:
+	{
+	  /* Or registers with an offset.  */
+	  if (!REG_P (XEXP (op, 0)))
+	    return false;
+
+	  op = XEXP (op, 1);
+
+	  /* The offset must be an immediate though.  */
+	  if (!CONST_INT_P (op))
+	    return false;
+
+	  range = INTVAL (op);
+
+	  /* Within the range of [-1020,1020].  */
+	  if (!IN_RANGE (range, -1020, 1020))
+	    return false;
+
+	  /* And a multiple of 4.  */
+	  return (range % 4) == 0;
+	}
+      case PRE_INC:
+      case POST_INC:
+      case PRE_DEC:
+      case POST_DEC:
+	return REG_P (XEXP (op, 0));
+      default:
+	gcc_unreachable ();
+    }
+  return false;
+}
 #include "gt-arm.h"

gcc/config/arm/arm.md

@@ -11939,6 +11939,44 @@
   [(set_attr "length" "4")
    (set_attr "type" "coproc")])
 
+(define_insn "*ldc"
+  [(unspec_volatile [(match_operand:SI 0 "immediate_operand" "n")
+		     (match_operand:SI 1 "immediate_operand" "n")
+		     (match_operand:SI 2 "memory_operand" "Uz")] LDCI)]
+  "arm_coproc_builtin_available (VUNSPEC_<LDC>)"
+{
+  arm_const_bounds (operands[0], 0, 16);
+  arm_const_bounds (operands[1], 0, (1 << 5));
+  return "<ldc>\\tp%c0, CR%c1, %2";
+}
+  [(set_attr "length" "4")
+   (set_attr "type" "coproc")])
+
+(define_insn "*stc"
+  [(unspec_volatile [(match_operand:SI 0 "immediate_operand" "n")
+		     (match_operand:SI 1 "immediate_operand" "n")
+		     (match_operand:SI 2 "memory_operand" "=Uz")] STCI)]
+  "arm_coproc_builtin_available (VUNSPEC_<STC>)"
+{
+  arm_const_bounds (operands[0], 0, 16);
+  arm_const_bounds (operands[1], 0, (1 << 5));
+  return "<stc>\\tp%c0, CR%c1, %2";
+}
+  [(set_attr "length" "4")
+   (set_attr "type" "coproc")])
+
+(define_expand "<ldc>"
+  [(unspec_volatile [(match_operand:SI 0 "immediate_operand")
+		     (match_operand:SI 1 "immediate_operand")
+		     (mem:SI (match_operand:SI 2 "s_register_operand"))] LDCI)]
+  "arm_coproc_builtin_available (VUNSPEC_<LDC>)")
+
+(define_expand "<stc>"
+  [(unspec_volatile [(match_operand:SI 0 "immediate_operand")
+		     (match_operand:SI 1 "immediate_operand")
+		     (mem:SI (match_operand:SI 2 "s_register_operand"))] STCI)]
+  "arm_coproc_builtin_available (VUNSPEC_<STC>)")
+
 ;; Vector bits common to IWMMXT and Neon
 (include "vec-common.md")
 ;; Load the Intel Wireless Multimedia Extension patterns

gcc/config/arm/arm_acle.h

@@ -41,6 +41,33 @@ __arm_cdp (const unsigned int __coproc, const unsigned int __opc1,
   return __builtin_arm_cdp (__coproc, __opc1, __CRd, __CRn, __CRm, __opc2);
 }
 
+__extension__ static __inline void __attribute__ ((__always_inline__))
+__arm_ldc (const unsigned int __coproc, const unsigned int __CRd,
+	   const void * __p)
+{
+  return __builtin_arm_ldc (__coproc, __CRd, __p);
+}
+
+__extension__ static __inline void __attribute__ ((__always_inline__))
+__arm_ldcl (const unsigned int __coproc, const unsigned int __CRd,
+	    const void * __p)
+{
+  return __builtin_arm_ldcl (__coproc, __CRd, __p);
+}
+
+__extension__ static __inline void __attribute__ ((__always_inline__))
+__arm_stc (const unsigned int __coproc, const unsigned int __CRd,
+	   void * __p)
+{
+  return __builtin_arm_stc (__coproc, __CRd, __p);
+}
+
+__extension__ static __inline void __attribute__ ((__always_inline__))
+__arm_stcl (const unsigned int __coproc, const unsigned int __CRd,
+	    void * __p)
+{
+  return __builtin_arm_stcl (__coproc, __CRd, __p);
+}
 #if __ARM_ARCH >= 5
 __extension__ static __inline void __attribute__ ((__always_inline__))
 __arm_cdp2 (const unsigned int __coproc, const unsigned int __opc1,
@@ -49,6 +76,34 @@ __arm_cdp2 (const unsigned int __coproc, const unsigned int __opc1,
 {
   return __builtin_arm_cdp2 (__coproc, __opc1, __CRd, __CRn, __CRm, __opc2);
 }
+
+__extension__ static __inline void __attribute__ ((__always_inline__))
+__arm_ldc2 (const unsigned int __coproc, const unsigned int __CRd,
+	    const void * __p)
+{
+  return __builtin_arm_ldc2 (__coproc, __CRd, __p);
+}
+
+__extension__ static __inline void __attribute__ ((__always_inline__))
+__arm_ldc2l (const unsigned int __coproc, const unsigned int __CRd,
+	     const void * __p)
+{
+  return __builtin_arm_ldc2l (__coproc, __CRd, __p);
+}
+
+__extension__ static __inline void __attribute__ ((__always_inline__))
+__arm_stc2 (const unsigned int __coproc, const unsigned int __CRd,
+	    void * __p)
+{
+  return __builtin_arm_stc2 (__coproc, __CRd, __p);
+}
+
+__extension__ static __inline void __attribute__ ((__always_inline__))
+__arm_stc2l (const unsigned int __coproc, const unsigned int __CRd,
+	     void * __p)
+{
+  return __builtin_arm_stc2l (__coproc, __CRd, __p);
+}
 #endif /*  __ARM_ARCH >= 5.  */
 #endif /* (!__thumb__ || __thumb2__) &&  __ARM_ARCH >= 4.  */
 

gcc/config/arm/arm_acle_builtins.def

@@ -26,3 +26,11 @@ VAR1 (UBINOP, crc32ch, si)
 VAR1 (UBINOP, crc32cw, si)
 VAR1 (CDP, cdp, void)
 VAR1 (CDP, cdp2, void)
+VAR1 (LDC, ldc, void)
+VAR1 (LDC, ldc2, void)
+VAR1 (LDC, ldcl, void)
+VAR1 (LDC, ldc2l, void)
+VAR1 (STC, stc, void)
+VAR1 (STC, stc2, void)
+VAR1 (STC, stcl, void)
+VAR1 (STC, stc2l, void)

gcc/config/arm/constraints.md

@@ -447,6 +447,12 @@
  (match_code "symbol_ref")
 )
 
+(define_memory_constraint "Uz"
+ "@internal
+  A memory access that is accessible as an LDC/STC operand"
+ (and (match_code "mem")
+      (match_test "arm_coproc_ldc_stc_legitimate_address (op)")))
+
 ;; We used to have constraint letters for S and R in ARM state, but
 ;; all uses of these now appear to have been removed.
 

gcc/config/arm/iterators.md

@@ -948,3 +948,19 @@
 (define_int_iterator CDPI [VUNSPEC_CDP VUNSPEC_CDP2])
 (define_int_attr cdp [(VUNSPEC_CDP "cdp") (VUNSPEC_CDP2 "cdp2")])
 (define_int_attr CDP [(VUNSPEC_CDP "CDP") (VUNSPEC_CDP2 "CDP2")])
+
+;; An iterator for the LDC coprocessor instruction
+(define_int_iterator LDCI [VUNSPEC_LDC VUNSPEC_LDC2
+			   VUNSPEC_LDCL VUNSPEC_LDC2L])
+(define_int_attr ldc [(VUNSPEC_LDC "ldc") (VUNSPEC_LDC2 "ldc2")
+		      (VUNSPEC_LDCL "ldcl") (VUNSPEC_LDC2L "ldc2l")])
+(define_int_attr LDC [(VUNSPEC_LDC "LDC") (VUNSPEC_LDC2 "LDC2")
+		      (VUNSPEC_LDCL "LDCL") (VUNSPEC_LDC2L "LDC2L")])
+
+;; An iterator for the STC coprocessor instructions
+(define_int_iterator STCI [VUNSPEC_STC VUNSPEC_STC2
+			   VUNSPEC_STCL VUNSPEC_STC2L])
+(define_int_attr stc [(VUNSPEC_STC "stc") (VUNSPEC_STC2 "stc2")
+		      (VUNSPEC_STCL "stcl") (VUNSPEC_STC2L "stc2l")])
+(define_int_attr STC [(VUNSPEC_STC "STC") (VUNSPEC_STC2 "STC2")
+		      (VUNSPEC_STCL "STCL") (VUNSPEC_STC2L "STC2L")])

gcc/config/arm/unspecs.md

@@ -152,6 +152,14 @@
   VUNSPEC_PROBE_STACK_RANGE ; Represent stack range probing.
   VUNSPEC_CDP		; Represent the coprocessor cdp instruction.
   VUNSPEC_CDP2		; Represent the coprocessor cdp2 instruction.
+  VUNSPEC_LDC		; Represent the coprocessor ldc instruction.
+  VUNSPEC_LDC2		; Represent the coprocessor ldc2 instruction.
+  VUNSPEC_LDCL		; Represent the coprocessor ldcl instruction.
+  VUNSPEC_LDC2L		; Represent the coprocessor ldc2l instruction.
+  VUNSPEC_STC		; Represent the coprocessor stc instruction.
+  VUNSPEC_STC2		; Represent the coprocessor stc2 instruction.
+  VUNSPEC_STCL		; Represent the coprocessor stcl instruction.
+  VUNSPEC_STC2L		; Represent the coprocessor stc2l instruction.
 ])
 
 ;; Enumerators for NEON unspecs.

gcc/testsuite/ChangeLog

 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
 
+	* gcc.target/arm/acle/ldc: New.
+	* gcc.target/arm/acle/ldc2: New.
+	* gcc.target/arm/acle/ldcl: New.
+	* gcc.target/arm/acle/ldc2l: New.
+	* gcc.target/arm/acle/stc: New.
+	* gcc.target/arm/acle/stc2: New.
+	* gcc.target/arm/acle/stcl: New.
+	* gcc.target/arm/acle/stc2l: New.
+
+2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
+
 	* gcc.target/arm/acle/acle.exp: Run tests for different options
 	and make sure fat-lto-objects is used such that we can still do
 	assemble scans.

gcc/testsuite/gcc.target/arm/acle/ldc.c

+/* Test the ldc ACLE intrinsic.  */
+
+/* { dg-do assemble } */
+/* { dg-options "-save-temps" } */
+/* { dg-require-effective-target arm_coproc1_ok } */
+
+#include "arm_acle.h"
+
+extern void * p;
+
+void test_ldc (void)
+{
+  __arm_ldc (10, 1, p + 4);
+  __arm_ldc (11, 1, p + 1024);
+}
+
+/* { dg-final { scan-assembler "ldc\tp10, CR1, \[r\[0-9\]+" } } */
+/* { dg-final { scan-assembler "ldc\tp11, CR1, \[r\[0-9\]+\]\n" } } */

gcc/testsuite/gcc.target/arm/acle/ldc2.c

+/* Test the ldc2 ACLE intrinsic.  */
+
+/* { dg-do assemble } */
+/* { dg-options "-save-temps" } */
+/* { dg-require-effective-target arm_coproc2_ok } */
+
+#include "arm_acle.h"
+
+extern void * p;
+
+void test_ldc2 (void)
+{
+  __arm_ldc2 (10, 1, p - 120);
+  __arm_ldc2 (11, 1, p - 122);
+}
+
+/* { dg-final { scan-assembler "ldc2\tp10, CR1, \[r\[0-9\]+" } } */
+/* { dg-final { scan-assembler "ldc2\tp11, CR1, \[r\[0-9\]+\]\n" } } */

gcc/testsuite/gcc.target/arm/acle/ldc2l.c

+/* Test the ldc2l ACLE intrinsic.  */
+
+/* { dg-do assemble } */
+/* { dg-options "-save-temps" } */
+/* { dg-require-effective-target arm_coproc2_ok } */
+
+#include "arm_acle.h"
+
+extern void * p;
+
+void test_ldc2l (void)
+{
+  __arm_ldc2l (10, 1, p - 120);
+  __arm_ldc2l (11, 1, p - 122);
+}
+
+/* { dg-final { scan-assembler "ldc2l\tp10, CR1, \[r\[0-9\]+" } } */
+/* { dg-final { scan-assembler "ldc2l\tp11, CR1, \[r\[0-9\]+\]\n" } } */

gcc/testsuite/gcc.target/arm/acle/ldcl.c

+/* Test the ldcl ACLE intrinsic.  */
+
+/* { dg-do assemble } */
+/* { dg-options "-save-temps" } */
+/* { dg-require-effective-target arm_coproc1_ok } */
+
+#include "arm_acle.h"
+
+extern void * p;
+
+void test_ldcl (void)
+{
+  __arm_ldcl (10, 1, p + 4);
+  __arm_ldcl (11, 1, p + 1024);
+}
+
+/* { dg-final { scan-assembler "ldcl\tp10, CR1, \[r\[0-9\]+" } } */
+/* { dg-final { scan-assembler "ldcl\tp11, CR1, \[r\[0-9\]+\]\n" } } */

gcc/testsuite/gcc.target/arm/acle/stc.c

+/* Test the stc ACLE intrinsic.  */
+
+/* { dg-do assemble } */
+/* { dg-options "-save-temps" } */
+/* { dg-require-effective-target arm_coproc1_ok } */
+
+#include "arm_acle.h"
+
+extern void * p;
+
+void test_stc (void)
+{
+  __arm_stc (10, 1, p + 4);
+  __arm_stc (11, 1, p + 1024);
+}
+
+/* { dg-final { scan-assembler "stc\tp10, CR1, \[r\[0-9\]+" } } */
+/* { dg-final { scan-assembler "stc\tp11, CR1, \[r\[0-9\]+\]\n" } } */

gcc/testsuite/gcc.target/arm/acle/stc2.c

+/* Test the stc2 ACLE intrinsic.  */
+
+/* { dg-do assemble } */
+/* { dg-options "-save-temps" } */
+/* { dg-require-effective-target arm_coproc2_ok } */
+
+#include "arm_acle.h"
+
+extern void * p;
+
+void test_stc2 (void)
+{
+  __arm_stc2 (10, 1, p - 120);
+  __arm_stc2 (11, 1, p - 122);
+}
+
+/* { dg-final { scan-assembler "stc2\tp10, CR1, \[r\[0-9\]+" } } */
+/* { dg-final { scan-assembler "stc2\tp11, CR1, \[r\[0-9\]+\]\n" } } */

gcc/testsuite/gcc.target/arm/acle/stc2l.c

+/* Test the stc2l ACLE intrinsic.  */
+
+/* { dg-do assemble } */
+/* { dg-options "-save-temps" } */
+/* { dg-require-effective-target arm_coproc2_ok } */
+
+#include "arm_acle.h"
+
+extern void * p;
+
+void test_stc2l (void)
+{
+  __arm_stc2l (10, 1, p - 120);
+  __arm_stc2l (11, 1, p - 122);
+}
+
+/* { dg-final { scan-assembler "stc2l\tp10, CR1, \[r\[0-9\]+" } } */
+/* { dg-final { scan-assembler "stc2l\tp11, CR1, \[r\[0-9\]+\]\n" } } */

gcc/testsuite/gcc.target/arm/acle/stcl.c

+/* Test the stcl ACLE intrinsic.  */
+
+/* { dg-do assemble } */
+/* { dg-options "-save-temps" } */
+/* { dg-require-effective-target arm_coproc1_ok } */
+
+#include "arm_acle.h"
+
+extern void * p;
+
+void test_stcl (void)
+{
+  __arm_stcl (14, 10, p + 4);
+  __arm_stcl (10, 10, p + 1024);
+}
+
+/* { dg-final { scan-assembler "stcl\tp14, CR10, \[r\[0-9\]+" } } */
+/* { dg-final { scan-assembler "stcl\tp10, CR10, \[r\[0-9\]+\]\n" } } */