This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[PATCH, ARM 5/6] Add support for MOVT/MOVW to ARMv8-M Baseline


Hi,

This patch is part of a patch series to add support for ARMv8-M[1] to GCC. This specific patch makes the compiler start generating code with the new MOVT/MOVW instructions for ARMv8-M Baseline.

[1] For a quick overview of ARMv8-M please refer to the initial cover letter.

ChangeLog entry is as follows:


*** gcc/ChangeLog ***

2015-11-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>

        * config/arm/arm.h (TARGET_HAVE_MOVT): Include ARMv8-M as having MOVT.
        * config/arm/arm.c (arm_arch_name): (const_ok_for_op): Check MOVT/MOVW
        availability with TARGET_HAVE_MOVT.
        (thumb_legitimate_constant_p): Legalize high part of a label_ref as a
        constant.
        (thumb1_rtx_costs): Also return 0 if setting a half word constant and
        movw is available.
        (thumb1_size_rtx_costs): Make set of half word constant also cost 1
        extra instruction if MOVW is available.  Make constant with bottom half
        word zero cost 2 instruction if MOVW is available.
        * config/arm/arm.md (define_attr "arch"): Add v8mb.
        (define_attr "arch_enabled"): Set to yes if arch value is v8mb and
        target is ARMv8-M Baseline.
        * config/arm/thumb1.md (thumb1_movdi_insn): Add ARMv8-M Baseline only
        alternative for constants satisfying j constraint.
        (thumb1_movsi_insn): Likewise.
        (movsi splitter for K alternative): Tighten condition to not trigger
        if movt is available and j constraint is satisfied.
        (Pe immediate splitter): Likewise.
        (thumb1_movhi_insn): Add ARMv8-M Baseline only alternative for
        constant fitting in an halfword to use movw.
        * doc/sourcebuild.texi (arm_thumb1_movt_ko): Document new ARM
        effective target.


*** gcc/testsuite/ChangeLog ***

2015-11-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>

        * lib/target-supports.exp (check_effective_target_arm_thumb1_movt_ko):
        Define effective target.
        * gcc.target/arm/pr42574.c: Require arm_thumb1_movt_ko instead of
        arm_thumb1_ok as effective target to exclude ARMv8-M Baseline.


diff --git a/gcc/config/arm/arm.h b/gcc/config/arm/arm.h
index ff3cfcd..015df50 100644
--- a/gcc/config/arm/arm.h
+++ b/gcc/config/arm/arm.h
@@ -261,7 +261,7 @@ extern void (*arm_lang_output_object_attributes_hook)(void);
 #define TARGET_HAVE_LDACQ	(TARGET_ARM_ARCH >= 8 && arm_arch_notm)
 
 /* Nonzero if this chip provides the movw and movt instructions.  */
-#define TARGET_HAVE_MOVT	(arm_arch_thumb2)
+#define TARGET_HAVE_MOVT	(arm_arch_thumb2 || arm_arch8)
 
 /* Nonzero if integer division instructions supported.  */
 #define TARGET_IDIV	((TARGET_ARM && arm_arch_arm_hwdiv)	\
diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c
index 51d501e..d832309 100644
--- a/gcc/config/arm/arm.c
+++ b/gcc/config/arm/arm.c
@@ -8158,6 +8158,12 @@ arm_legitimate_constant_p_1 (machine_mode, rtx x)
 static bool
 thumb_legitimate_constant_p (machine_mode mode ATTRIBUTE_UNUSED, rtx x)
 {
+  /* Splitters for TARGET_USE_MOVT call arm_emit_movpair which creates high
+     RTX.  These RTX must therefore be allowed for Thumb-1 so that when run
+     for ARMv8-M baseline or later the result is valid.  */
+  if (TARGET_HAVE_MOVT && GET_CODE (x) == HIGH)
+    x = XEXP (x, 0);
+
   return (CONST_INT_P (x)
 	  || CONST_DOUBLE_P (x)
 	  || CONSTANT_ADDRESS_P (x)
@@ -8244,7 +8250,8 @@ thumb1_rtx_costs (rtx x, enum rtx_code code, enum rtx_code outer)
     case CONST_INT:
       if (outer == SET)
 	{
-	  if ((unsigned HOST_WIDE_INT) INTVAL (x) < 256)
+	  if ((unsigned HOST_WIDE_INT) INTVAL (x) < 256
+	      || (TARGET_HAVE_MOVT && !(INTVAL (x) & 0xffff0000)))
 	    return 0;
 	  if (thumb_shiftable_const (INTVAL (x)))
 	    return COSTS_N_INSNS (2);
@@ -8994,16 +9001,24 @@ thumb1_size_rtx_costs (rtx x, enum rtx_code code, enum rtx_code outer)
 	 the mode.  */
       words = ARM_NUM_INTS (GET_MODE_SIZE (GET_MODE (SET_DEST (x))));
       return COSTS_N_INSNS (words)
-	     + COSTS_N_INSNS (1) * (satisfies_constraint_J (SET_SRC (x))
-				    || satisfies_constraint_K (SET_SRC (x))
-				       /* thumb1_movdi_insn.  */
-				    || ((words > 1) && MEM_P (SET_SRC (x))));
+	     + COSTS_N_INSNS (1)
+	       * (satisfies_constraint_J (SET_SRC (x))
+		  || satisfies_constraint_K (SET_SRC (x))
+		     /* Too big immediate for 2byte mov, using movt.  */
+		  || ((unsigned HOST_WIDE_INT) INTVAL (SET_SRC (x)) >= 256
+		      && TARGET_HAVE_MOVT
+		      && satisfies_constraint_j (SET_SRC (x)))
+		     /* thumb1_movdi_insn.  */
+		  || ((words > 1) && MEM_P (SET_SRC (x))));
 
     case CONST_INT:
       if (outer == SET)
         {
           if ((unsigned HOST_WIDE_INT) INTVAL (x) < 256)
             return COSTS_N_INSNS (1);
+	  /* movw is 4byte long.  */
+	  if (TARGET_HAVE_MOVT && !(INTVAL (x) & 0xffff0000))
+	    return COSTS_N_INSNS (2);
 	  /* See split "TARGET_THUMB1 && satisfies_constraint_J".  */
 	  if (INTVAL (x) >= -255 && INTVAL (x) <= -1)
             return COSTS_N_INSNS (2);
diff --git a/gcc/config/arm/arm.md b/gcc/config/arm/arm.md
index 78dafa0..ce5ea83 100644
--- a/gcc/config/arm/arm.md
+++ b/gcc/config/arm/arm.md
@@ -115,10 +115,10 @@
 ; This can be "a" for ARM, "t" for either of the Thumbs, "32" for
 ; TARGET_32BIT, "t1" or "t2" to specify a specific Thumb mode.  "v6"
 ; for ARM or Thumb-2 with arm_arch6, and nov6 for ARM without
-; arm_arch6.  "v6t2" for Thumb-2 with arm_arch6.  This attribute is
-; used to compute attribute "enabled", use type "any" to enable an
-; alternative in all cases.
-(define_attr "arch" "any,a,t,32,t1,t2,v6,nov6,v6t2,neon_for_64bits,avoid_neon_for_64bits,iwmmxt,iwmmxt2,armv6_or_vfpv3"
+; arm_arch6.  "v6t2" for Thumb-2 with arm_arch6 and v8mb for ARMv8-M
+; baseline.  This attribute is used to compute attribute "enabled",
+; use type "any" to enable an alternative in all cases.
+(define_attr "arch" "any,a,t,32,t1,t2,v6,nov6,v6t2,v8mb,neon_for_64bits,avoid_neon_for_64bits,iwmmxt,iwmmxt2,armv6_or_vfpv3"
   (const_string "any"))
 
 (define_attr "arch_enabled" "no,yes"
@@ -157,6 +157,10 @@
 	      (match_test "TARGET_32BIT && arm_arch6 && arm_arch_thumb2"))
 	 (const_string "yes")
 
+	 (and (eq_attr "arch" "v8mb")
+	      (match_test "TARGET_THUMB1 && arm_arch8"))
+	 (const_string "yes")
+
 	 (and (eq_attr "arch" "avoid_neon_for_64bits")
 	      (match_test "TARGET_NEON")
 	      (not (match_test "TARGET_PREFER_NEON_64BITS")))
diff --git a/gcc/config/arm/thumb1.md b/gcc/config/arm/thumb1.md
index 3847b33b..7e3bcb4 100644
--- a/gcc/config/arm/thumb1.md
+++ b/gcc/config/arm/thumb1.md
@@ -590,8 +590,8 @@
 ;;; ??? The 'i' constraint looks funny, but it should always be replaced by
 ;;; thumb_reorg with a memory reference.
 (define_insn "*thumb1_movdi_insn"
-  [(set (match_operand:DI 0 "nonimmediate_operand" "=l,l,l,l,>,l, m,*r")
-	(match_operand:DI 1 "general_operand"      "l, I,J,>,l,mi,l,*r"))]
+  [(set (match_operand:DI 0 "nonimmediate_operand" "=l,l,l,r,l,>,l, m,*r")
+	(match_operand:DI 1 "general_operand"      "l, I,J,j,>,l,mi,l,*r"))]
   "TARGET_THUMB1
    && (   register_operand (operands[0], DImode)
        || register_operand (operands[1], DImode))"
@@ -609,37 +609,41 @@
     case 2:
       operands[1] = GEN_INT (- INTVAL (operands[1]));
       return \"movs\\t%Q0, %1\;rsbs\\t%Q0, %Q0, #0\;asrs\\t%R0, %Q0, #31\";
-    case 3:
-      return \"ldmia\\t%1, {%0, %H0}\";
+    case 3: if (!TARGET_HAVE_MOVT) gcc_unreachable ();
+	    return \"movw\\t%Q0, %L1\;movs\\tR0, #0\";
     case 4:
-      return \"stmia\\t%0, {%1, %H1}\";
+      return \"ldmia\\t%1, {%0, %H0}\";
     case 5:
-      return thumb_load_double_from_address (operands);
+      return \"stmia\\t%0, {%1, %H1}\";
     case 6:
+      return thumb_load_double_from_address (operands);
+    case 7:
       operands[2] = gen_rtx_MEM (SImode,
 			     plus_constant (Pmode, XEXP (operands[0], 0), 4));
       output_asm_insn (\"str\\t%1, %0\;str\\t%H1, %2\", operands);
       return \"\";
-    case 7:
+    case 8:
       if (REGNO (operands[1]) == REGNO (operands[0]) + 1)
 	return \"mov\\t%0, %1\;mov\\t%H0, %H1\";
       return \"mov\\t%H0, %H1\;mov\\t%0, %1\";
     }
   }"
-  [(set_attr "length" "4,4,6,2,2,6,4,4")
-   (set_attr "type" "multiple,multiple,multiple,load2,store2,load2,store2,multiple")
-   (set_attr "pool_range" "*,*,*,*,*,1018,*,*")]
+  [(set_attr "length" "4,4,6,6,2,2,6,4,4")
+   (set_attr "type" "multiple,multiple,multiple,multiple,load2,store2,load2,store2,multiple")
+   (set_attr "arch" "t1,t1,t1,v8mb,t1,t1,t1,t1,t1")
+   (set_attr "pool_range" "*,*,*,*,*,*,1018,*,*")]
 )
 
 (define_insn "*thumb1_movsi_insn"
-  [(set (match_operand:SI 0 "nonimmediate_operand" "=l,l,l,l,l,>,l, m,*l*h*k")
-	(match_operand:SI 1 "general_operand"      "l, I,J,K,>,l,mi,l,*l*h*k"))]
+  [(set (match_operand:SI 0 "nonimmediate_operand" "=l,l,r,l,l,l,>,l, m,*l*h*k")
+	(match_operand:SI 1 "general_operand"      "l, I,j,J,K,>,l,mi,l,*l*h*k"))]
   "TARGET_THUMB1
    && (   register_operand (operands[0], SImode)
        || register_operand (operands[1], SImode))"
   "@
    movs	%0, %1
    movs	%0, %1
+   movw	%0, %1
    #
    #
    ldmia\\t%1, {%0}
@@ -647,10 +651,11 @@
    ldr\\t%0, %1
    str\\t%1, %0
    mov\\t%0, %1"
-  [(set_attr "length" "2,2,4,4,2,2,2,2,2")
-   (set_attr "type" "mov_reg,mov_imm,multiple,multiple,load1,store1,load1,store1,mov_reg")
-   (set_attr "pool_range" "*,*,*,*,*,*,1018,*,*")
-   (set_attr "conds" "set,clob,*,*,nocond,nocond,nocond,nocond,nocond")])
+  [(set_attr "length" "2,2,4,4,4,2,2,2,2,2")
+   (set_attr "type" "mov_reg,mov_imm,mov_imm,multiple,multiple,load1,store1,load1,store1,mov_reg")
+   (set_attr "pool_range" "*,*,*,*,*,*,*,1018,*,*")
+   (set_attr "arch" "t1,t1,v8mb,t1,t1,t1,t1,t1,t1,t1")
+   (set_attr "conds" "set,clob,nocond,*,*,nocond,nocond,nocond,nocond,nocond")])
 
 ; Split the load of 64-bit constant into two loads for high and low 32-bit parts respectively
 ; to see if we can load them in fewer instructions or fewer cycles.
@@ -687,7 +692,8 @@
 (define_split
   [(set (match_operand:SI 0 "register_operand" "")
 	(match_operand:SI 1 "const_int_operand" ""))]
-  "TARGET_THUMB1 && satisfies_constraint_K (operands[1])"
+  "TARGET_THUMB1 && satisfies_constraint_K (operands[1])
+   && !(TARGET_HAVE_MOVT && satisfies_constraint_j (operands[1]))"
   [(set (match_dup 2) (match_dup 1))
    (set (match_dup 0) (ashift:SI (match_dup 2) (match_dup 3)))]
   "
@@ -714,7 +720,8 @@
 (define_split
   [(set (match_operand:SI 0 "register_operand" "")
 	(match_operand:SI 1 "const_int_operand" ""))]
-  "TARGET_THUMB1 && satisfies_constraint_Pe (operands[1])"
+  "TARGET_THUMB1 && satisfies_constraint_Pe (operands[1])
+   && !(TARGET_HAVE_MOVT && satisfies_constraint_j (operands[1]))"
   [(set (match_dup 2) (match_dup 1))
    (set (match_dup 0) (plus:SI (match_dup 2) (match_dup 3)))]
   "
@@ -726,8 +733,8 @@
 )
 
 (define_insn "*thumb1_movhi_insn"
-  [(set (match_operand:HI 0 "nonimmediate_operand" "=l,l,m,l*r,*h,l")
-	(match_operand:HI 1 "general_operand"       "l,m,l,k*h,*r,I"))]
+  [(set (match_operand:HI 0 "nonimmediate_operand" "=l,l,m,l*r,*h,l,r")
+	(match_operand:HI 1 "general_operand"       "l,m,l,k*h,*r,I,n"))]
   "TARGET_THUMB1
    && (   register_operand (operands[0], HImode)
        || register_operand (operands[1], HImode))"
@@ -739,6 +746,8 @@
     case 3: return \"mov	%0, %1\";
     case 4: return \"mov	%0, %1\";
     case 5: return \"movs	%0, %1\";
+    case 6: if (!TARGET_HAVE_MOVT) gcc_unreachable ();
+	    return \"movw	%0, %L1\";
     default: gcc_unreachable ();
     case 1:
       /* The stack pointer can end up being taken as an index register.
@@ -758,9 +767,10 @@
 	}
       return \"ldrh	%0, %1\";
     }"
-  [(set_attr "length" "2,4,2,2,2,2")
-   (set_attr "type" "alus_imm,load1,store1,mov_reg,mov_reg,mov_imm")
-   (set_attr "conds" "clob,nocond,nocond,nocond,nocond,clob")])
+  [(set_attr "length" "2,4,2,2,2,2,4")
+   (set_attr "type" "alus_imm,load1,store1,mov_reg,mov_reg,mov_imm,mov_imm")
+   (set_attr "arch" "t1,t1,t1,t1,t1,t1,v8mb")
+   (set_attr "conds" "clob,nocond,nocond,nocond,nocond,clob,nocond")])
 
 (define_expand "thumb_movhi_clobber"
   [(set (match_operand:HI     0 "memory_operand"   "")
diff --git a/gcc/doc/sourcebuild.texi b/gcc/doc/sourcebuild.texi
index 61de4a56..6c6f0f1 100644
--- a/gcc/doc/sourcebuild.texi
+++ b/gcc/doc/sourcebuild.texi
@@ -1586,6 +1586,10 @@ Some multilibs may be incompatible with these options.
 ARM target prefers @code{LDRD} and @code{STRD} instructions over
 @code{LDM} and @code{STM} instructions.
 
+@item arm_thumb1_movt_ko
+ARM target generates Thumb-1 code for @code{-mthumb} with no
+@code{MOVT} instruction available.
+
 @end table
 
 @subsubsection AArch64-specific attributes
diff --git a/gcc/testsuite/gcc.target/arm/pr42574.c b/gcc/testsuite/gcc.target/arm/pr42574.c
index 0ccd05f..53aea62 100644
--- a/gcc/testsuite/gcc.target/arm/pr42574.c
+++ b/gcc/testsuite/gcc.target/arm/pr42574.c
@@ -1,5 +1,5 @@
 /* { dg-options "-mthumb -Os -fpic" }  */
-/* { dg-require-effective-target arm_thumb1_ok } */
+/* { dg-require-effective-target arm_thumb1_movt_ko } */
 /* { dg-require-effective-target fpic } */
 /* Make sure the address of glob.c is calculated only once and using
    a logical shift for the offset (200<<1).  */
diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp
index 37c273a..44ebbe3 100644
--- a/gcc/testsuite/lib/target-supports.exp
+++ b/gcc/testsuite/lib/target-supports.exp
@@ -3211,6 +3211,23 @@ proc check_effective_target_arm_cortex_m { } {
     } "-mthumb"]
 }
 
+# Return 1 if this is an ARM target where -mthumb causes Thumb-1 to be
+# used and no movt/movw instructions to be available.
+
+proc check_effective_target_arm_thumb1_movt_ko {} {
+    if [check_effective_target_arm_thumb1_ok] {
+	return [expr ![check_no_compiler_messages arm_movt object {
+	    int
+	    foo (void)
+	    {
+	      asm ("movt r0, #42");
+	    }
+	} "-mthumb"]]
+    } else {
+	return 0
+    }
+}
+
 # Return 1 if this compilation turns on string_ops_prefer_neon on.
 
 proc check_effective_target_arm_tune_string_ops_prefer_neon { } {


Testing:

* Toolchain was built successfully with and without the ARMv8-M support patches[1] with the following multilib list: armv6-m,armv7-m,armv7e-m,cortex-m7. The code generation for crtbegin.o, crtend.o, crti.o, crtn.o, libgcc.a, libgcov.a, libc.a, libg.a, libgloss-linux.a, libm.a, libnosys.a, librdimon.a, librdpmon.a, libstdc++.a and libsupc++.a is unchanged for all these targets. 

[1] the binutils one, the GCC one, the newlib one and the 4 ld ones (adding support to ARMv8-M security extension)

* GCC also showed no testsuite regression when targeting ARMv8-M Baseline compared to ARMv6-M on ARM Fast Models and when targeting ARMv6-M and ARMv7-M (compared to without the patch)
* GCC was bootstrapped successfully targeting Thumb-1 and targeting Thumb-2

Is this ok for stage3?

Best regards,

Thomas


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]