gcc: Upgrade to 4.5.3 and bring in latest linaro updates

We base now off release tarball of 4.5.3 instead of svn
Get relevant new arm linaro patches

Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
This commit is contained in:
Khem Raj
2011-05-01 03:26:50 +00:00
committed by Koen Kooi
parent 1e2b88cbc4
commit 0d04214334
9 changed files with 2754 additions and 199 deletions
+10 -15
View File
@@ -13,21 +13,12 @@ LIC_FILES_CHKSUM = "file://COPYING;md5=59530bdf33659b29e73d4adb9f9f6552 \
file://COPYING.RUNTIME;md5=fe60d87048567d4fe8c8a0ed2448bcc8" file://COPYING.RUNTIME;md5=fe60d87048567d4fe8c8a0ed2448bcc8"
SRCREV = "170880"
PV = "4.5" PV = "4.5"
INC_PR = "r35" INC_PR = "r36"
# BINV should point to minor release
# BINV should be incremented after updating to a revision
# after a minor gcc release (e.g. 4.5.1 or 4.5.2) has been made
# the value will be minor-release+1 e.g. if minor release was
# 4.5.1 then the value below will be 2 which will mean 4.5.2
# which will be next minor release and so on.
BINV = "${PV}.3" BINV = "${PV}.3"
BRANCH = "gcc-4_5-branch"
PR_append = "+svnr${SRCPV}"
SRC_URI = "svn://gcc.gnu.org/svn/gcc/branches;module=${BRANCH} \ SRC_URI = "${GNU_MIRROR}/gcc/gcc-${BINV}/gcc-${BINV}.tar.bz2 \
file://gcc-4.3.1-ARCH_FLAGS_FOR_TARGET.patch \ file://gcc-4.3.1-ARCH_FLAGS_FOR_TARGET.patch \
file://100-uclibc-conf.patch \ file://100-uclibc-conf.patch \
file://gcc-uclibc-locale-ctype_touplow_t.patch \ file://gcc-uclibc-locale-ctype_touplow_t.patch \
@@ -175,11 +166,15 @@ SRC_URI = "svn://gcc.gnu.org/svn/gcc/branches;module=${BRANCH} \
file://linaro/gcc-4.5-linaro-r99478.patch \ file://linaro/gcc-4.5-linaro-r99478.patch \
file://linaro/gcc-4.5-linaro-r99479.patch \ file://linaro/gcc-4.5-linaro-r99479.patch \
file://linaro/gcc-4.5-linaro-r99480.patch \ file://linaro/gcc-4.5-linaro-r99480.patch \
file://linaro/gcc-4.5-linaro-r99481.patch \
file://linaro/gcc-4.5-linaro-r99482.patch \
file://linaro/gcc-4.5-linaro-r99483.patch \ file://linaro/gcc-4.5-linaro-r99483.patch \
file://linaro/gcc-4.5-linaro-r99488.patch \ file://linaro/gcc-4.5-linaro-r99488.patch \
file://linaro/gcc-4.5-linaro-r99489.patch \ file://linaro/gcc-4.5-linaro-r99489.patch \
file://linaro/gcc-4.5-linaro-r99491.patch \
file://linaro/gcc-4.5-linaro-r99494.patch \
file://linaro/gcc-4.5-linaro-r99495.patch \
file://linaro/gcc-4.5-linaro-r99498.patch \
# file://linaro/gcc-4.5-linaro-r99499.patch \
file://linaro/gcc-4.5-linaro-r99502.patch \
file://more-epilogues.patch \ file://more-epilogues.patch \
file://gcc-scalar-widening-pr45847.patch \ file://gcc-scalar-widening-pr45847.patch \
file://gcc-arm-volatile-bitfield-fix.patch \ file://gcc-arm-volatile-bitfield-fix.patch \
@@ -207,7 +202,7 @@ SRC_URI = "svn://gcc.gnu.org/svn/gcc/branches;module=${BRANCH} \
FORTRAN = "" FORTRAN = ""
JAVA = "" JAVA = ""
S = "${WORKDIR}/${BRANCH}" S = "${WORKDIR}/gcc-${BINV}"
#EXTRA_OECONF_BASE = " --enable-cheaders=c_std \ #EXTRA_OECONF_BASE = " --enable-cheaders=c_std \
# --enable-libssp \ # --enable-libssp \
@@ -37,9 +37,11 @@
gcc/ gcc/
=== modified file 'gcc/config/arm/arm-modes.def' === modified file 'gcc/config/arm/arm-modes.def'
--- old/gcc/config/arm/arm-modes.def 2009-06-18 11:24:10 +0000 Index: gcc-4.5.3/gcc/config/arm/arm-modes.def
+++ new/gcc/config/arm/arm-modes.def 2010-07-29 16:58:56 +0000 ===================================================================
@@ -35,10 +35,16 @@ --- gcc-4.5.3.orig/gcc/config/arm/arm-modes.def
+++ gcc-4.5.3/gcc/config/arm/arm-modes.def
@@ -35,10 +35,16 @@ ADJUST_FLOAT_FORMAT (HF, ((arm_fp16_form
CC_NOOVmode should be used with SImode integer equalities. CC_NOOVmode should be used with SImode integer equalities.
CC_Zmode should be used if only the Z flag is set correctly CC_Zmode should be used if only the Z flag is set correctly
CC_Nmode should be used if only the N (sign) flag is set correctly CC_Nmode should be used if only the N (sign) flag is set correctly
@@ -56,11 +58,11 @@
CC_MODE (CC_SWP); CC_MODE (CC_SWP);
CC_MODE (CCFP); CC_MODE (CCFP);
CC_MODE (CCFPE); CC_MODE (CCFPE);
Index: gcc-4.5.3/gcc/config/arm/arm-protos.h
=== modified file 'gcc/config/arm/arm-protos.h' ===================================================================
--- old/gcc/config/arm/arm-protos.h 2009-11-11 14:23:03 +0000 --- gcc-4.5.3.orig/gcc/config/arm/arm-protos.h
+++ new/gcc/config/arm/arm-protos.h 2010-07-29 16:58:56 +0000 +++ gcc-4.5.3/gcc/config/arm/arm-protos.h
@@ -49,8 +49,7 @@ @@ -49,8 +49,7 @@ extern int arm_hard_regno_mode_ok (unsig
extern int const_ok_for_arm (HOST_WIDE_INT); extern int const_ok_for_arm (HOST_WIDE_INT);
extern int arm_split_constant (RTX_CODE, enum machine_mode, rtx, extern int arm_split_constant (RTX_CODE, enum machine_mode, rtx,
HOST_WIDE_INT, rtx, rtx, int); HOST_WIDE_INT, rtx, rtx, int);
@@ -70,7 +72,7 @@
extern int legitimate_pic_operand_p (rtx); extern int legitimate_pic_operand_p (rtx);
extern rtx legitimize_pic_address (rtx, enum machine_mode, rtx); extern rtx legitimize_pic_address (rtx, enum machine_mode, rtx);
extern rtx legitimize_tls_address (rtx, rtx); extern rtx legitimize_tls_address (rtx, rtx);
@@ -116,6 +115,7 @@ @@ -116,6 +115,7 @@ extern void arm_reload_in_hi (rtx *);
extern void arm_reload_out_hi (rtx *); extern void arm_reload_out_hi (rtx *);
extern int arm_const_double_inline_cost (rtx); extern int arm_const_double_inline_cost (rtx);
extern bool arm_const_double_by_parts (rtx); extern bool arm_const_double_by_parts (rtx);
@@ -78,11 +80,11 @@
extern const char *fp_immediate_constant (rtx); extern const char *fp_immediate_constant (rtx);
extern void arm_emit_call_insn (rtx, rtx); extern void arm_emit_call_insn (rtx, rtx);
extern const char *output_call (rtx *); extern const char *output_call (rtx *);
Index: gcc-4.5.3/gcc/config/arm/arm.c
=== modified file 'gcc/config/arm/arm.c' ===================================================================
--- old/gcc/config/arm/arm.c 2010-07-29 15:59:12 +0000 --- gcc-4.5.3.orig/gcc/config/arm/arm.c
+++ new/gcc/config/arm/arm.c 2010-07-29 16:58:56 +0000 +++ gcc-4.5.3/gcc/config/arm/arm.c
@@ -3190,13 +3190,82 @@ @@ -3191,13 +3191,82 @@ arm_gen_constant (enum rtx_code code, en
immediate value easier to load. */ immediate value easier to load. */
enum rtx_code enum rtx_code
@@ -169,7 +171,7 @@
switch (code) switch (code)
{ {
case EQ: case EQ:
@@ -9912,6 +9981,55 @@ @@ -9913,6 +9982,55 @@ arm_select_cc_mode (enum rtx_code op, rt
&& (rtx_equal_p (XEXP (x, 0), y) || rtx_equal_p (XEXP (x, 1), y))) && (rtx_equal_p (XEXP (x, 0), y) || rtx_equal_p (XEXP (x, 1), y)))
return CC_Cmode; return CC_Cmode;
@@ -225,18 +227,17 @@
return CCmode; return CCmode;
} }
@@ -9921,10 +10039,39 @@ @@ -9922,10 +10040,39 @@ arm_select_cc_mode (enum rtx_code op, rt
rtx rtx
arm_gen_compare_reg (enum rtx_code code, rtx x, rtx y) arm_gen_compare_reg (enum rtx_code code, rtx x, rtx y)
{ {
- enum machine_mode mode = SELECT_CC_MODE (code, x, y); - enum machine_mode mode = SELECT_CC_MODE (code, x, y);
- rtx cc_reg = gen_rtx_REG (mode, CC_REGNUM); - rtx cc_reg = gen_rtx_REG (mode, CC_REGNUM);
-
- emit_set_insn (cc_reg, gen_rtx_COMPARE (mode, x, y));
+ enum machine_mode mode; + enum machine_mode mode;
+ rtx cc_reg; + rtx cc_reg;
+ int dimode_comparison = GET_MODE (x) == DImode || GET_MODE (y) == DImode; + int dimode_comparison = GET_MODE (x) == DImode || GET_MODE (y) == DImode;
+
- emit_set_insn (cc_reg, gen_rtx_COMPARE (mode, x, y));
+ /* We might have X as a constant, Y as a register because of the predicates + /* We might have X as a constant, Y as a register because of the predicates
+ used for cmpdi. If so, force X to a register here. */ + used for cmpdi. If so, force X to a register here. */
+ if (dimode_comparison && !REG_P (x)) + if (dimode_comparison && !REG_P (x))
@@ -269,7 +270,7 @@
return cc_reg; return cc_reg;
} }
@@ -11253,6 +11400,34 @@ @@ -11254,6 +11401,34 @@ arm_const_double_by_parts (rtx val)
return false; return false;
} }
@@ -304,7 +305,7 @@
/* Scan INSN and note any of its operands that need fixing. /* Scan INSN and note any of its operands that need fixing.
If DO_PUSHES is false we do not actually push any of the fixups If DO_PUSHES is false we do not actually push any of the fixups
needed. The function returns TRUE if any fixups were needed/pushed. needed. The function returns TRUE if any fixups were needed/pushed.
@@ -15097,8 +15272,18 @@ @@ -15150,8 +15325,18 @@ arm_print_operand (FILE *stream, rtx x,
the value being loaded is big-wordian or little-wordian. The the value being loaded is big-wordian or little-wordian. The
order of the two register loads can matter however, if the address order of the two register loads can matter however, if the address
of the memory location is actually held in one of the registers of the memory location is actually held in one of the registers
@@ -324,7 +325,7 @@
if (GET_CODE (x) != REG || REGNO (x) > LAST_ARM_REGNUM) if (GET_CODE (x) != REG || REGNO (x) > LAST_ARM_REGNUM)
{ {
output_operand_lossage ("invalid operand for code '%c'", code); output_operand_lossage ("invalid operand for code '%c'", code);
@@ -15109,6 +15294,18 @@ @@ -15162,6 +15347,18 @@ arm_print_operand (FILE *stream, rtx x,
return; return;
case 'R': case 'R':
@@ -343,7 +344,7 @@
if (GET_CODE (x) != REG || REGNO (x) > LAST_ARM_REGNUM) if (GET_CODE (x) != REG || REGNO (x) > LAST_ARM_REGNUM)
{ {
output_operand_lossage ("invalid operand for code '%c'", code); output_operand_lossage ("invalid operand for code '%c'", code);
@@ -15801,6 +15998,28 @@ @@ -15854,6 +16051,28 @@ get_arm_condition_code (rtx comparison)
default: gcc_unreachable (); default: gcc_unreachable ();
} }
@@ -372,11 +373,11 @@
case CCmode: case CCmode:
switch (comp_code) switch (comp_code)
{ {
Index: gcc-4.5.3/gcc/config/arm/arm.h
=== modified file 'gcc/config/arm/arm.h' ===================================================================
--- old/gcc/config/arm/arm.h 2009-12-23 16:36:40 +0000 --- gcc-4.5.3.orig/gcc/config/arm/arm.h
+++ new/gcc/config/arm/arm.h 2010-07-29 16:58:56 +0000 +++ gcc-4.5.3/gcc/config/arm/arm.h
@@ -2253,19 +2253,7 @@ @@ -2253,19 +2253,7 @@ extern int making_const_table;
: reverse_condition (code)) : reverse_condition (code))
#define CANONICALIZE_COMPARISON(CODE, OP0, OP1) \ #define CANONICALIZE_COMPARISON(CODE, OP0, OP1) \
@@ -397,10 +398,10 @@
/* The arm5 clz instruction returns 32. */ /* The arm5 clz instruction returns 32. */
#define CLZ_DEFINED_VALUE_AT_ZERO(MODE, VALUE) ((VALUE) = 32, 1) #define CLZ_DEFINED_VALUE_AT_ZERO(MODE, VALUE) ((VALUE) = 32, 1)
Index: gcc-4.5.3/gcc/config/arm/arm.md
=== modified file 'gcc/config/arm/arm.md' ===================================================================
--- old/gcc/config/arm/arm.md 2010-07-29 15:59:12 +0000 --- gcc-4.5.3.orig/gcc/config/arm/arm.md
+++ new/gcc/config/arm/arm.md 2010-07-29 16:58:56 +0000 +++ gcc-4.5.3/gcc/config/arm/arm.md
@@ -6718,17 +6718,45 @@ @@ -6718,17 +6718,45 @@
operands[3])); DONE;" operands[3])); DONE;"
) )
@@ -506,7 +507,7 @@
;; Cirrus SF compare instruction ;; Cirrus SF compare instruction
(define_insn "*cirrus_cmpsf" (define_insn "*cirrus_cmpsf"
[(set (reg:CCFP CC_REGNUM) [(set (reg:CCFP CC_REGNUM)
@@ -8183,18 +8257,45 @@ @@ -8183,17 +8257,44 @@
operands[2], operands[3])); DONE;" operands[2], operands[3])); DONE;"
) )
@@ -519,13 +520,14 @@
- "TARGET_ARM && TARGET_HARD_FLOAT && TARGET_MAVERICK" - "TARGET_ARM && TARGET_HARD_FLOAT && TARGET_MAVERICK"
- "emit_insn (gen_cstore_cc (operands[0], operands[1], - "emit_insn (gen_cstore_cc (operands[0], operands[1],
- operands[2], operands[3])); DONE;" - operands[2], operands[3])); DONE;"
-)
+ [(match_operand:DI 2 "cmpdi_operand" "") + [(match_operand:DI 2 "cmpdi_operand" "")
+ (match_operand:DI 3 "cmpdi_operand" "")]))] + (match_operand:DI 3 "cmpdi_operand" "")]))]
+ "TARGET_32BIT" + "TARGET_32BIT"
+ "{ + "{
+ rtx swap = NULL_RTX; + rtx swap = NULL_RTX;
+ enum rtx_code code = GET_CODE (operands[1]); + enum rtx_code code = GET_CODE (operands[1]);
+
+ /* We should not have two constants. */ + /* We should not have two constants. */
+ gcc_assert (GET_MODE (operands[2]) == DImode + gcc_assert (GET_MODE (operands[2]) == DImode
+ || GET_MODE (operands[3]) == DImode); + || GET_MODE (operands[3]) == DImode);
@@ -553,16 +555,14 @@
+ operands[3])); + operands[3]));
+ DONE; + DONE;
+ }" + }"
) +)
-
(define_expand "cstoresi_eq0_thumb1" (define_expand "cstoresi_eq0_thumb1"
[(parallel [(parallel
[(set (match_operand:SI 0 "s_register_operand" "") Index: gcc-4.5.3/gcc/config/arm/constraints.md
===================================================================
=== modified file 'gcc/config/arm/constraints.md' --- gcc-4.5.3.orig/gcc/config/arm/constraints.md
--- old/gcc/config/arm/constraints.md 2009-12-07 20:34:53 +0000 +++ gcc-4.5.3/gcc/config/arm/constraints.md
+++ new/gcc/config/arm/constraints.md 2010-07-29 16:58:56 +0000
@@ -29,7 +29,7 @@ @@ -29,7 +29,7 @@
;; in Thumb-1 state: I, J, K, L, M, N, O ;; in Thumb-1 state: I, J, K, L, M, N, O
@@ -586,11 +586,11 @@
(define_constraint "Dn" (define_constraint "Dn"
"@internal "@internal
In ARM/Thumb-2 state a const_vector which can be loaded with a Neon vmov In ARM/Thumb-2 state a const_vector which can be loaded with a Neon vmov
Index: gcc-4.5.3/gcc/config/arm/predicates.md
=== modified file 'gcc/config/arm/predicates.md' ===================================================================
--- old/gcc/config/arm/predicates.md 2010-07-29 15:59:12 +0000 --- gcc-4.5.3.orig/gcc/config/arm/predicates.md
+++ new/gcc/config/arm/predicates.md 2010-07-29 16:58:56 +0000 +++ gcc-4.5.3/gcc/config/arm/predicates.md
@@ -86,6 +86,12 @@ @@ -101,6 +101,12 @@
(and (match_code "const_int") (and (match_code "const_int")
(match_test "const_ok_for_arm (INTVAL (op))"))) (match_test "const_ok_for_arm (INTVAL (op))")))
@@ -603,7 +603,7 @@
(define_predicate "arm_neg_immediate_operand" (define_predicate "arm_neg_immediate_operand"
(and (match_code "const_int") (and (match_code "const_int")
(match_test "const_ok_for_arm (-INTVAL (op))"))) (match_test "const_ok_for_arm (-INTVAL (op))")))
@@ -115,6 +121,10 @@ @@ -130,6 +136,10 @@
(ior (match_operand 0 "arm_rhs_operand") (ior (match_operand 0 "arm_rhs_operand")
(match_operand 0 "arm_not_immediate_operand"))) (match_operand 0 "arm_not_immediate_operand")))
@@ -614,23 +614,23 @@
;; True if the operand is a memory reference which contains an ;; True if the operand is a memory reference which contains an
;; offsettable address. ;; offsettable address.
(define_predicate "offsettable_memory_operand" (define_predicate "offsettable_memory_operand"
@@ -522,4 +532,12 @@ @@ -538,3 +548,12 @@
(define_predicate "neon_lane_number"
(and (match_code "const_int") (and (match_code "const_int")
(match_test "INTVAL (op) >= 0 && INTVAL (op) <= 7"))) (match_test "INTVAL (op) >= 0 && INTVAL (op) <= 15")))
+;; Predicates for named expanders that overlap multiple ISAs. +;; Predicates for named expanders that overlap multiple ISAs.
+ +
+(define_predicate "cmpdi_operand" +(define_predicate "cmpdi_operand"
+ (if_then_else (match_test "TARGET_HARD_FLOAT && TARGET_MAVERICK") + (if_then_else (match_test "TARGET_HARD_FLOAT && TARGET_MAVERICK")
+ (and (match_test "TARGET_ARM") + (and (match_test "TARGET_ARM")
+ (match_operand 0 "cirrus_fp_register")) + (match_operand 0 "cirrus_fp_register"))
+ (and (match_test "TARGET_32BIT") + (and (match_test "TARGET_32BIT")
+ (match_operand 0 "arm_di_operand")))) + (match_operand 0 "arm_di_operand"))))
+
Index: gcc-4.5.3/gcc/testsuite/gcc.c-torture/execute/20100416-1.c
=== added file 'gcc/testsuite/gcc.c-torture/execute/20100416-1.c' ===================================================================
--- old/gcc/testsuite/gcc.c-torture/execute/20100416-1.c 1970-01-01 00:00:00 +0000 --- /dev/null
+++ new/gcc/testsuite/gcc.c-torture/execute/20100416-1.c 2010-07-29 16:58:56 +0000 +++ gcc-4.5.3/gcc/testsuite/gcc.c-torture/execute/20100416-1.c
@@ -0,0 +1,40 @@ @@ -0,0 +1,40 @@
+void abort(void); +void abort(void);
+ +
@@ -672,4 +672,3 @@
+ } + }
+ return 0; + return 0;
+} +}
@@ -1,40 +0,0 @@
2011-02-02 Richard Sandiford <richard.sandiford@linaro.org>
gcc/testsuite/
PR target/47553
* gcc.target/arm/neon-vld-1.c: New test.
gcc/
PR target/47553
* config/arm/predicates.md (neon_lane_number): Accept 0..15.
=== modified file 'gcc/config/arm/predicates.md'
--- old/gcc/config/arm/predicates.md 2011-02-08 12:07:29 +0000
+++ new/gcc/config/arm/predicates.md 2011-03-02 12:28:41 +0000
@@ -607,7 +607,7 @@
;; TODO: We could check lane numbers more precisely based on the mode.
(define_predicate "neon_lane_number"
(and (match_code "const_int")
- (match_test "INTVAL (op) >= 0 && INTVAL (op) <= 7")))
+ (match_test "INTVAL (op) >= 0 && INTVAL (op) <= 15")))
;; Predicates for named expanders that overlap multiple ISAs.
(define_predicate "cmpdi_operand"
=== added file 'gcc/testsuite/gcc.target/arm/neon-vld-1.c'
--- old/gcc/testsuite/gcc.target/arm/neon-vld-1.c 1970-01-01 00:00:00 +0000
+++ new/gcc/testsuite/gcc.target/arm/neon-vld-1.c 2011-02-02 13:57:54 +0000
@@ -0,0 +1,13 @@
+/* { dg-do compile } */
+/* { dg-require-effective-target arm_neon_ok } */
+/* { dg-options "-O1" } */
+/* { dg-add-options arm_neon } */
+
+#include <arm_neon.h>
+
+uint8x16_t
+foo (uint8_t *a, uint8x16_t b)
+{
+ vst1q_lane_u8 (a, b, 14);
+ return vld1q_lane_u8 (a + 0x100, b, 15);
+}
@@ -1,84 +0,0 @@
2011-02-02 Richard Sandiford <richard.sandiford@linaro.org>
gcc/
Backport from mainline:
2011-01-23 Bernd Schmidt <bernds@codesourcery.com>
Richard Sandiford <rdsandiford@googlemail.com>
PR rtl-optimization/47166
* reload1.c (emit_reload_insns): Disable the spill_reg_store
mechanism for PRE_MODIFY and POST_MODIFY.
(inc_for_reload): For PRE_MODIFY, return the insn that sets the
reloadreg.
gcc/testsuite/
* gcc.c-torture/execute/postmod-1.c: New test.
=== added file 'gcc/testsuite/gcc.c-torture/execute/postmod-1.c'
--- old/gcc/testsuite/gcc.c-torture/execute/postmod-1.c 1970-01-01 00:00:00 +0000
+++ new/gcc/testsuite/gcc.c-torture/execute/postmod-1.c 2011-02-02 14:23:10 +0000
@@ -0,0 +1,62 @@
+#define DECLARE_ARRAY(A) array##A[0x10]
+#define DECLARE_COUNTER(A) counter##A = 0
+#define DECLARE_POINTER(A) *pointer##A = array##A + x
+/* Create a loop that allows post-modification of pointerA, followed by
+ a use of the post-modified address. */
+#define BEFORE(A) counter##A += *pointer##A, pointer##A += 3
+#define AFTER(A) counter##A += pointer##A[x]
+
+/* Set up the arrays so that one iteration of the loop sets the counter
+ to 3.0f. */
+#define INIT_ARRAY(A) array##A[1] = 1.0f, array##A[5] = 2.0f
+
+/* Check that the loop worked correctly for all values. */
+#define CHECK_ARRAY(A) exit_code |= (counter##A != 3.0f)
+
+/* Having 6 copies triggered the bug for ARM and Thumb. */
+#define MANY(A) A (0), A (1), A (2), A (3), A (4), A (5)
+
+/* Each addendA should be allocated a register. */
+#define INIT_VOLATILE(A) addend##A = vol
+#define ADD_VOLATILE(A) vol += addend##A
+
+/* Having 5 copies triggered the bug for ARM and Thumb. */
+#define MANY2(A) A (0), A (1), A (2), A (3), A (4)
+
+float MANY (DECLARE_ARRAY);
+float MANY (DECLARE_COUNTER);
+
+volatile int stop = 1;
+volatile int vol;
+
+void __attribute__((noinline))
+foo (int x)
+{
+ float MANY (DECLARE_POINTER);
+ int i;
+
+ do
+ {
+ MANY (BEFORE);
+ MANY (AFTER);
+ /* Create an inner loop that should ensure the code above
+ has registers free for reload inheritance. */
+ {
+ int MANY2 (INIT_VOLATILE);
+ for (i = 0; i < 10; i++)
+ MANY2 (ADD_VOLATILE);
+ }
+ }
+ while (!stop);
+}
+
+int
+main (void)
+{
+ int exit_code = 0;
+
+ MANY (INIT_ARRAY);
+ foo (1);
+ MANY (CHECK_ARRAY);
+ return exit_code;
+}
@@ -0,0 +1,309 @@
2011-04-08 Richard Sandiford <richard.sandiford@linaro.org>
Backport from mainline:
gcc/
* dwarf2out.c (dw_loc_list_node): Add resolved_addr and replaced.
(cached_dw_loc_list_def): New structure.
(cached_dw_loc_list): New typedef.
(cached_dw_loc_list_table): New variable.
(cached_dw_loc_list_table_hash): New function.
(cached_dw_loc_list_table_eq): Likewise.
(add_location_or_const_value_attribute): Take a bool cache_p.
Cache the list when the parameter is true.
(gen_formal_parameter_die): Update caller.
(gen_variable_die): Likewise.
(dwarf2out_finish): Likewise.
(dwarf2out_abstract_function): Nullify cached_dw_loc_list_table
while generating debug info for the decl.
(dwarf2out_function_decl): Clear cached_dw_loc_list_table.
(dwarf2out_init): Initialize cached_dw_loc_list_table.
(resolve_addr): Cache the result of resolving a chain of
location lists.
From: 2010-10-12 Jakub Jelinek <jakub@redhat.com>
* dwarf2out.c (dw_loc_list_node): Add emitted field.
(output_loc_list): Return immediately if emitted is set, set it.
=== modified file 'gcc/dwarf2out.c'
--- old/gcc/dwarf2out.c 2011-03-10 14:55:10 +0000
+++ new/gcc/dwarf2out.c 2011-03-28 09:43:50 +0000
@@ -4256,6 +4256,12 @@
Only on head of list */
const char *section; /* Section this loclist is relative to */
dw_loc_descr_ref expr;
+ /* True if all addresses in this and subsequent lists are known to be
+ resolved. */
+ bool resolved_addr;
+ /* True if this list has been replaced by dw_loc_next. */
+ bool replaced;
+ bool emitted;
} dw_loc_list_node;
#if defined (DWARF2_DEBUGGING_INFO) || defined (DWARF2_UNWIND_INFO)
@@ -5759,6 +5765,19 @@
/* Table of decl location linked lists. */
static GTY ((param_is (var_loc_list))) htab_t decl_loc_table;
+/* A cached location list. */
+struct GTY (()) cached_dw_loc_list_def {
+ /* The DECL_UID of the decl that this entry describes. */
+ unsigned int decl_id;
+
+ /* The cached location list. */
+ dw_loc_list_ref loc_list;
+};
+typedef struct cached_dw_loc_list_def cached_dw_loc_list;
+
+/* Table of cached location lists. */
+static GTY ((param_is (cached_dw_loc_list))) htab_t cached_dw_loc_list_table;
+
/* A pointer to the base of a list of references to DIE's that
are uniquely identified by their tag, presence/absence of
children DIE's, and list of attribute/value pairs. */
@@ -6112,7 +6131,7 @@
static void insert_int (HOST_WIDE_INT, unsigned, unsigned char *);
static void insert_float (const_rtx, unsigned char *);
static rtx rtl_for_decl_location (tree);
-static bool add_location_or_const_value_attribute (dw_die_ref, tree,
+static bool add_location_or_const_value_attribute (dw_die_ref, tree, bool,
enum dwarf_attribute);
static bool tree_add_const_value_attribute (dw_die_ref, tree);
static bool tree_add_const_value_attribute_for_decl (dw_die_ref, tree);
@@ -7731,6 +7750,24 @@
htab_find_with_hash (decl_loc_table, decl, DECL_UID (decl));
}
+/* Returns a hash value for X (which really is a cached_dw_loc_list_list). */
+
+static hashval_t
+cached_dw_loc_list_table_hash (const void *x)
+{
+ return (hashval_t) ((const cached_dw_loc_list *) x)->decl_id;
+}
+
+/* Return nonzero if decl_id of cached_dw_loc_list X is the same as
+ UID of decl *Y. */
+
+static int
+cached_dw_loc_list_table_eq (const void *x, const void *y)
+{
+ return (((const cached_dw_loc_list *) x)->decl_id
+ == DECL_UID ((const_tree) y));
+}
+
/* Equate a DIE to a particular declaration. */
static void
@@ -10355,6 +10392,10 @@
{
dw_loc_list_ref curr = list_head;
+ if (list_head->emitted)
+ return;
+ list_head->emitted = true;
+
ASM_OUTPUT_LABEL (asm_out_file, list_head->ll_symbol);
/* Walk the location list, and output each range + expression. */
@@ -15920,15 +15961,22 @@
these things can crop up in other ways also.) Note that one type of
constant value which can be passed into an inlined function is a constant
pointer. This can happen for example if an actual argument in an inlined
- function call evaluates to a compile-time constant address. */
+ function call evaluates to a compile-time constant address.
+
+ CACHE_P is true if it is worth caching the location list for DECL,
+ so that future calls can reuse it rather than regenerate it from scratch.
+ This is true for BLOCK_NONLOCALIZED_VARS in inlined subroutines,
+ since we will need to refer to them each time the function is inlined. */
static bool
-add_location_or_const_value_attribute (dw_die_ref die, tree decl,
+add_location_or_const_value_attribute (dw_die_ref die, tree decl, bool cache_p,
enum dwarf_attribute attr)
{
rtx rtl;
dw_loc_list_ref list;
var_loc_list *loc_list;
+ cached_dw_loc_list *cache;
+ void **slot;
if (TREE_CODE (decl) == ERROR_MARK)
return false;
@@ -15964,7 +16012,34 @@
&& add_const_value_attribute (die, rtl))
return true;
}
- list = loc_list_from_tree (decl, decl_by_reference_p (decl) ? 0 : 2);
+ /* If this decl is from BLOCK_NONLOCALIZED_VARS, we might need its
+ list several times. See if we've already cached the contents. */
+ list = NULL;
+ if (loc_list == NULL || cached_dw_loc_list_table == NULL)
+ cache_p = false;
+ if (cache_p)
+ {
+ cache = (cached_dw_loc_list *)
+ htab_find_with_hash (cached_dw_loc_list_table, decl, DECL_UID (decl));
+ if (cache)
+ list = cache->loc_list;
+ }
+ if (list == NULL)
+ {
+ list = loc_list_from_tree (decl, decl_by_reference_p (decl) ? 0 : 2);
+ /* It is usually worth caching this result if the decl is from
+ BLOCK_NONLOCALIZED_VARS and if the list has at least two elements. */
+ if (cache_p && list && list->dw_loc_next)
+ {
+ slot = htab_find_slot_with_hash (cached_dw_loc_list_table, decl,
+ DECL_UID (decl), INSERT);
+ cache = (cached_dw_loc_list *)
+ ggc_alloc_cleared (sizeof (cached_dw_loc_list));
+ cache->decl_id = DECL_UID (decl);
+ cache->loc_list = list;
+ *slot = cache;
+ }
+ }
if (list)
{
add_AT_location_description (die, attr, list);
@@ -17532,7 +17607,7 @@
equate_decl_number_to_die (node, parm_die);
if (! DECL_ABSTRACT (node_or_origin))
add_location_or_const_value_attribute (parm_die, node_or_origin,
- DW_AT_location);
+ node == NULL, DW_AT_location);
break;
@@ -17712,6 +17787,7 @@
tree context;
int was_abstract;
htab_t old_decl_loc_table;
+ htab_t old_cached_dw_loc_list_table;
/* Make sure we have the actual abstract inline, not a clone. */
decl = DECL_ORIGIN (decl);
@@ -17725,7 +17801,9 @@
DIE. Be sure to not clobber the outer location table nor use it or we would
get locations in abstract instantces. */
old_decl_loc_table = decl_loc_table;
+ old_cached_dw_loc_list_table = cached_dw_loc_list_table;
decl_loc_table = NULL;
+ cached_dw_loc_list_table = NULL;
/* Be sure we've emitted the in-class declaration DIE (if any) first, so
we don't get confused by DECL_ABSTRACT. */
@@ -17750,6 +17828,7 @@
current_function_decl = save_fn;
decl_loc_table = old_decl_loc_table;
+ cached_dw_loc_list_table = old_cached_dw_loc_list_table;
pop_cfun ();
}
@@ -18431,9 +18510,8 @@
&& !TREE_SYMBOL_REFERENCED (DECL_ASSEMBLER_NAME (decl_or_origin)))
defer_location (decl_or_origin, var_die);
else
- add_location_or_const_value_attribute (var_die,
- decl_or_origin,
- DW_AT_location);
+ add_location_or_const_value_attribute (var_die, decl_or_origin,
+ decl == NULL, DW_AT_location);
add_pubname (decl_or_origin, var_die);
}
else
@@ -20049,6 +20127,7 @@
dwarf2out_decl (decl);
htab_empty (decl_loc_table);
+ htab_empty (cached_dw_loc_list_table);
}
/* Output a marker (i.e. a label) for the beginning of the generated code for
@@ -20649,6 +20728,11 @@
decl_loc_table = htab_create_ggc (10, decl_loc_table_hash,
decl_loc_table_eq, NULL);
+ /* Allocate the cached_dw_loc_list_table. */
+ cached_dw_loc_list_table
+ = htab_create_ggc (10, cached_dw_loc_list_table_hash,
+ cached_dw_loc_list_table_eq, NULL);
+
/* Allocate the initial hunk of the decl_scope_table. */
decl_scope_table = VEC_alloc (tree, gc, 256);
@@ -21278,30 +21362,53 @@
{
dw_die_ref c;
dw_attr_ref a;
- dw_loc_list_ref *curr;
+ dw_loc_list_ref *curr, *start, loc;
unsigned ix;
for (ix = 0; VEC_iterate (dw_attr_node, die->die_attr, ix, a); ix++)
switch (AT_class (a))
{
case dw_val_class_loc_list:
- curr = AT_loc_list_ptr (a);
- while (*curr)
+ start = curr = AT_loc_list_ptr (a);
+ loc = *curr;
+ gcc_assert (loc);
+ /* The same list can be referenced more than once. See if we have
+ already recorded the result from a previous pass. */
+ if (loc->replaced)
+ *curr = loc->dw_loc_next;
+ else if (!loc->resolved_addr)
{
- if (!resolve_addr_in_expr ((*curr)->expr))
+ /* As things stand, we do not expect or allow one die to
+ reference a suffix of another die's location list chain.
+ References must be identical or completely separate.
+ There is therefore no need to cache the result of this
+ pass on any list other than the first; doing so
+ would lead to unnecessary writes. */
+ while (*curr)
{
- dw_loc_list_ref next = (*curr)->dw_loc_next;
- if (next && (*curr)->ll_symbol)
+ gcc_assert (!(*curr)->replaced && !(*curr)->resolved_addr);
+ if (!resolve_addr_in_expr ((*curr)->expr))
{
- gcc_assert (!next->ll_symbol);
- next->ll_symbol = (*curr)->ll_symbol;
+ dw_loc_list_ref next = (*curr)->dw_loc_next;
+ if (next && (*curr)->ll_symbol)
+ {
+ gcc_assert (!next->ll_symbol);
+ next->ll_symbol = (*curr)->ll_symbol;
+ }
+ *curr = next;
}
- *curr = next;
+ else
+ curr = &(*curr)->dw_loc_next;
}
+ if (loc == *start)
+ loc->resolved_addr = 1;
else
- curr = &(*curr)->dw_loc_next;
+ {
+ loc->replaced = 1;
+ loc->dw_loc_next = *start;
+ }
}
- if (!AT_loc_list (a))
+ if (!*start)
{
remove_AT (die, a->dw_attr);
ix--;
@@ -21361,6 +21468,7 @@
add_location_or_const_value_attribute (
VEC_index (deferred_locations, deferred_locations_list, i)->die,
VEC_index (deferred_locations, deferred_locations_list, i)->variable,
+ false,
DW_AT_location);
}
File diff suppressed because it is too large Load Diff
@@ -0,0 +1,784 @@
2011-03-24 Revital Eres <revital.eres@linaro.org>
gcc/
* loop-doloop.c (doloop_condition_get): Support new form of
doloop pattern and use prev_nondebug_insn instead of PREV_INSN.
* config/arm/thumb2.md (*thumb2_addsi3_compare0): Remove "*".
(doloop_end): New.
* config/arm/arm.md (*addsi3_compare0): Remove "*".
* ddg.c (check_closing_branch_deps, get_node_of_insn_uid):
New functions.
(create_ddg): Pass sbitmap containing do-loop related
instructions instead of closing_branch_deps parameter and call
check_closing_branch_deps function.
* ddg.h (create_ddg): Adjust the function declaration.
* modulo-sched.c (PS_STAGE_COUNT): Rename to CALC_STAGE_COUNT
and redefine.
(doloop_register_get): Handle NONDEBUG_INSN_P.
(stage_count): New field in struct partial_schedule.
(mark_doloop_insns, calculate_stage_count): New functions.
(normalize_sched_times): Rename to reset_sched_times and handle
incrementing the sched time of the nodes by a constant value
passed as parameter.
(duplicate_insns_of_cycles): Skip closing branch.
(sms_schedule_by_order): Schedule closing branch when
closing_branch_deps is true.
(ps_insn_find_column): Handle closing branch.
(sms_schedule): Call reset_sched_times and handle case where
do-loop pattern is not decoupled from the other loop instructions.
Support new form of doloop pattern.
(ps_insert_empty_row): Update calls to normalize_sched_times
and rotate_partial_schedule functions.
=== modified file 'gcc/config/arm/arm.md'
--- old/gcc/config/arm/arm.md 2011-03-11 14:26:34 +0000
+++ new/gcc/config/arm/arm.md 2011-03-24 07:45:38 +0000
@@ -734,7 +734,7 @@
""
)
-(define_insn "*addsi3_compare0"
+(define_insn "addsi3_compare0"
[(set (reg:CC_NOOV CC_REGNUM)
(compare:CC_NOOV
(plus:SI (match_operand:SI 1 "s_register_operand" "r, r")
=== modified file 'gcc/config/arm/thumb2.md'
--- old/gcc/config/arm/thumb2.md 2011-02-08 10:51:58 +0000
+++ new/gcc/config/arm/thumb2.md 2011-03-24 07:45:38 +0000
@@ -1194,7 +1194,7 @@
(set_attr "length" "2")]
)
-(define_insn "*thumb2_addsi3_compare0"
+(define_insn "thumb2_addsi3_compare0"
[(set (reg:CC_NOOV CC_REGNUM)
(compare:CC_NOOV
(plus:SI (match_operand:SI 1 "s_register_operand" "l, 0, r")
@@ -1445,3 +1445,56 @@
[(set_attr "length" "4,4,16")
(set_attr "predicable" "yes")]
)
+
+
+;; Define the subtract-one-and-jump insns so loop.c
+;; knows what to generate.
+(define_expand "doloop_end"
+ [(use (match_operand 0 "" "")) ; loop pseudo
+ (use (match_operand 1 "" "")) ; iterations; zero if unknown
+ (use (match_operand 2 "" "")) ; max iterations
+ (use (match_operand 3 "" "")) ; loop level
+ (use (match_operand 4 "" ""))] ; label
+ "TARGET_32BIT"
+ "
+ {
+ /* Currently SMS relies on the do-loop pattern to recognize loops
+ where (1) the control part consists of all insns defining and/or
+ using a certain 'count' register and (2) the loop count can be
+ adjusted by modifying this register prior to the loop.
+ ??? The possible introduction of a new block to initialize the
+ new IV can potentially affect branch optimizations. */
+ if (optimize > 0 && flag_modulo_sched)
+ {
+ rtx s0;
+ rtx bcomp;
+ rtx loc_ref;
+ rtx cc_reg;
+ rtx insn;
+ rtx cmp;
+
+ /* Only use this on innermost loops. */
+ if (INTVAL (operands[3]) > 1)
+ FAIL;
+
+ if (GET_MODE (operands[0]) != SImode)
+ FAIL;
+
+ s0 = operands [0];
+ if (TARGET_THUMB2)
+ insn = emit_insn (gen_thumb2_addsi3_compare0 (s0, s0, GEN_INT (-1)));
+ else
+ insn = emit_insn (gen_addsi3_compare0 (s0, s0, GEN_INT (-1)));
+
+ cmp = XVECEXP (PATTERN (insn), 0, 0);
+ cc_reg = SET_DEST (cmp);
+ bcomp = gen_rtx_NE (VOIDmode, cc_reg, const0_rtx);
+ loc_ref = gen_rtx_LABEL_REF (VOIDmode, operands [4]);
+ emit_jump_insn (gen_rtx_SET (VOIDmode, pc_rtx,
+ gen_rtx_IF_THEN_ELSE (VOIDmode, bcomp,
+ loc_ref, pc_rtx)));
+ DONE;
+ }else
+ FAIL;
+ }")
+
=== modified file 'gcc/ddg.c'
--- old/gcc/ddg.c 2010-07-19 08:58:53 +0000
+++ new/gcc/ddg.c 2011-03-24 07:45:38 +0000
@@ -60,6 +60,8 @@
static ddg_edge_ptr create_ddg_edge (ddg_node_ptr, ddg_node_ptr, dep_type,
dep_data_type, int, int);
static void add_edge_to_ddg (ddg_ptr g, ddg_edge_ptr);
+static ddg_node_ptr get_node_of_insn_uid (ddg_ptr, int);
+
/* Auxiliary variable for mem_read_insn_p/mem_write_insn_p. */
static bool mem_ref_p;
@@ -450,12 +452,65 @@
sched_free_deps (head, tail, false);
}
+/* Given DOLOOP_INSNS which holds the instructions that
+ belong to the do-loop part; mark closing_branch_deps field in ddg G
+ as TRUE if the do-loop part's instructions are dependent on the other
+ loop instructions. Otherwise mark it as FALSE. */
+static void
+check_closing_branch_deps (ddg_ptr g, sbitmap doloop_insns)
+{
+ sbitmap_iterator sbi;
+ unsigned int u = 0;
+
+ EXECUTE_IF_SET_IN_SBITMAP (doloop_insns, 0, u, sbi)
+ {
+ ddg_edge_ptr e;
+ ddg_node_ptr u_node = get_node_of_insn_uid (g, u);
+
+ gcc_assert (u_node);
+
+ for (e = u_node->in; e != 0; e = e->next_in)
+ {
+ ddg_node_ptr v_node = e->src;
+
+ if (((unsigned int) INSN_UID (v_node->insn) == u)
+ || DEBUG_INSN_P (v_node->insn))
+ continue;
+
+ /* Ignore dependencies between memory writes and the
+ jump. */
+ if (JUMP_P (u_node->insn)
+ && e->type == OUTPUT_DEP
+ && mem_write_insn_p (v_node->insn))
+ continue;
+ if (!TEST_BIT (doloop_insns, INSN_UID (v_node->insn)))
+ {
+ g->closing_branch_deps = 1;
+ return;
+ }
+ }
+ for (e = u_node->out; e != 0; e = e->next_out)
+ {
+ ddg_node_ptr v_node = e->dest;
+
+ if (((unsigned int) INSN_UID (v_node->insn) == u)
+ || DEBUG_INSN_P (v_node->insn))
+ continue;
+ if (!TEST_BIT (doloop_insns, INSN_UID (v_node->insn)))
+ {
+ g->closing_branch_deps = 1;
+ return;
+ }
+ }
+ }
+ g->closing_branch_deps = 0;
+}
/* Given a basic block, create its DDG and return a pointer to a variable
of ddg type that represents it.
Initialize the ddg structure fields to the appropriate values. */
ddg_ptr
-create_ddg (basic_block bb, int closing_branch_deps)
+create_ddg (basic_block bb, sbitmap doloop_insns)
{
ddg_ptr g;
rtx insn, first_note;
@@ -465,7 +520,6 @@
g = (ddg_ptr) xcalloc (1, sizeof (struct ddg));
g->bb = bb;
- g->closing_branch_deps = closing_branch_deps;
/* Count the number of insns in the BB. */
for (insn = BB_HEAD (bb); insn != NEXT_INSN (BB_END (bb));
@@ -538,6 +592,11 @@
/* Build the data dependency graph. */
build_intra_loop_deps (g);
build_inter_loop_deps (g);
+
+ /* Check whether the do-loop part is decoupled from the other loop
+ instructions. */
+ check_closing_branch_deps (g, doloop_insns);
+
return g;
}
@@ -831,6 +890,18 @@
return NULL;
}
+/* Given the uid of an instruction UID return the node that represents it. */
+static ddg_node_ptr
+get_node_of_insn_uid (ddg_ptr g, int uid)
+{
+ int i;
+
+ for (i = 0; i < g->num_nodes; i++)
+ if (uid == INSN_UID (g->nodes[i].insn))
+ return &g->nodes[i];
+ return NULL;
+}
+
/* Given a set OPS of nodes in the DDG, find the set of their successors
which are not in OPS, and set their bits in SUCC. Bits corresponding to
OPS are cleared from SUCC. Leaves the other bits in SUCC unchanged. */
=== modified file 'gcc/ddg.h'
--- old/gcc/ddg.h 2009-11-25 10:55:54 +0000
+++ new/gcc/ddg.h 2011-03-24 07:45:38 +0000
@@ -167,7 +167,7 @@
};
-ddg_ptr create_ddg (basic_block, int closing_branch_deps);
+ddg_ptr create_ddg (basic_block, sbitmap);
void free_ddg (ddg_ptr);
void print_ddg (FILE *, ddg_ptr);
=== modified file 'gcc/loop-doloop.c'
--- old/gcc/loop-doloop.c 2010-07-19 08:58:53 +0000
+++ new/gcc/loop-doloop.c 2011-03-24 07:45:38 +0000
@@ -78,6 +78,8 @@
rtx inc_src;
rtx condition;
rtx pattern;
+ rtx cc_reg = NULL_RTX;
+ rtx reg_orig = NULL_RTX;
/* The canonical doloop pattern we expect has one of the following
forms:
@@ -96,7 +98,16 @@
2) (set (reg) (plus (reg) (const_int -1))
(set (pc) (if_then_else (reg != 0)
(label_ref (label))
- (pc))). */
+ (pc))).
+
+ Some targets (ARM) do the comparison before the branch, as in the
+ following form:
+
+ 3) (parallel [(set (cc) (compare ((plus (reg) (const_int -1), 0)))
+ (set (reg) (plus (reg) (const_int -1)))])
+ (set (pc) (if_then_else (cc == NE)
+ (label_ref (label))
+ (pc))) */
pattern = PATTERN (doloop_pat);
@@ -104,19 +115,47 @@
{
rtx cond;
rtx prev_insn = prev_nondebug_insn (doloop_pat);
+ rtx cmp_arg1, cmp_arg2;
+ rtx cmp_orig;
- /* We expect the decrement to immediately precede the branch. */
+ /* In case the pattern is not PARALLEL we expect two forms
+ of doloop which are cases 2) and 3) above: in case 2) the
+ decrement immediately precedes the branch, while in case 3)
+ the compare and decrement instructions immediately precede
+ the branch. */
if (prev_insn == NULL_RTX || !INSN_P (prev_insn))
return 0;
cmp = pattern;
- inc = PATTERN (PREV_INSN (doloop_pat));
+ if (GET_CODE (PATTERN (prev_insn)) == PARALLEL)
+ {
+ /* The third case: the compare and decrement instructions
+ immediately precede the branch. */
+ cmp_orig = XVECEXP (PATTERN (prev_insn), 0, 0);
+ if (GET_CODE (cmp_orig) != SET)
+ return 0;
+ if (GET_CODE (SET_SRC (cmp_orig)) != COMPARE)
+ return 0;
+ cmp_arg1 = XEXP (SET_SRC (cmp_orig), 0);
+ cmp_arg2 = XEXP (SET_SRC (cmp_orig), 1);
+ if (cmp_arg2 != const0_rtx
+ || GET_CODE (cmp_arg1) != PLUS)
+ return 0;
+ reg_orig = XEXP (cmp_arg1, 0);
+ if (XEXP (cmp_arg1, 1) != GEN_INT (-1)
+ || !REG_P (reg_orig))
+ return 0;
+ cc_reg = SET_DEST (cmp_orig);
+
+ inc = XVECEXP (PATTERN (prev_insn), 0, 1);
+ }
+ else
+ inc = PATTERN (prev_insn);
/* We expect the condition to be of the form (reg != 0) */
cond = XEXP (SET_SRC (cmp), 0);
if (GET_CODE (cond) != NE || XEXP (cond, 1) != const0_rtx)
return 0;
-
}
else
{
@@ -162,11 +201,15 @@
return 0;
if ((XEXP (condition, 0) == reg)
+ /* For the third case: */
+ || ((cc_reg != NULL_RTX)
+ && (XEXP (condition, 0) == cc_reg)
+ && (reg_orig == reg))
|| (GET_CODE (XEXP (condition, 0)) == PLUS
- && XEXP (XEXP (condition, 0), 0) == reg))
+ && XEXP (XEXP (condition, 0), 0) == reg))
{
if (GET_CODE (pattern) != PARALLEL)
- /* The second form we expect:
+ /* For the second form we expect:
(set (reg) (plus (reg) (const_int -1))
(set (pc) (if_then_else (reg != 0)
@@ -181,7 +224,24 @@
(set (reg) (plus (reg) (const_int -1)))
(additional clobbers and uses)])
- So we return that form instead.
+ For the third form we expect:
+
+ (parallel [(set (cc) (compare ((plus (reg) (const_int -1)), 0))
+ (set (reg) (plus (reg) (const_int -1)))])
+ (set (pc) (if_then_else (cc == NE)
+ (label_ref (label))
+ (pc)))
+
+ which is equivalent to the following:
+
+ (parallel [(set (cc) (compare (reg, 1))
+ (set (reg) (plus (reg) (const_int -1)))
+ (set (pc) (if_then_else (NE == cc)
+ (label_ref (label))
+ (pc))))])
+
+ So we return the second form instead for the two cases.
+
*/
condition = gen_rtx_fmt_ee (NE, VOIDmode, inc_src, const1_rtx);
=== modified file 'gcc/modulo-sched.c'
--- old/gcc/modulo-sched.c 2009-11-25 10:55:54 +0000
+++ new/gcc/modulo-sched.c 2011-03-24 07:45:38 +0000
@@ -116,8 +116,10 @@
/* The number of different iterations the nodes in ps span, assuming
the stage boundaries are placed efficiently. */
-#define PS_STAGE_COUNT(ps) ((PS_MAX_CYCLE (ps) - PS_MIN_CYCLE (ps) \
- + 1 + (ps)->ii - 1) / (ps)->ii)
+#define CALC_STAGE_COUNT(min_cycle,max_cycle,ii) ((max_cycle - min_cycle \
+ + 1 + ii - 1) / ii)
+/* The stage count of ps. */
+#define PS_STAGE_COUNT(ps) (((partial_schedule_ptr)(ps))->stage_count)
/* A single instruction in the partial schedule. */
struct ps_insn
@@ -155,6 +157,8 @@
int max_cycle;
ddg_ptr g; /* The DDG of the insns in the partial schedule. */
+
+ int stage_count; /* The stage count of the partial schedule. */
};
/* We use this to record all the register replacements we do in
@@ -195,6 +199,7 @@
rtx, rtx);
static void duplicate_insns_of_cycles (partial_schedule_ptr,
int, int, int, rtx);
+static int calculate_stage_count (partial_schedule_ptr ps);
#define SCHED_ASAP(x) (((node_sched_params_ptr)(x)->aux.info)->asap)
#define SCHED_TIME(x) (((node_sched_params_ptr)(x)->aux.info)->time)
@@ -310,10 +315,10 @@
either a single (parallel) branch-on-count or a (non-parallel)
branch immediately preceded by a single (decrement) insn. */
first_insn_not_to_check = (GET_CODE (PATTERN (tail)) == PARALLEL ? tail
- : PREV_INSN (tail));
+ : prev_nondebug_insn (tail));
for (insn = head; insn != first_insn_not_to_check; insn = NEXT_INSN (insn))
- if (reg_mentioned_p (reg, insn))
+ if (reg_mentioned_p (reg, insn) && NONDEBUG_INSN_P (insn))
{
if (dump_file)
{
@@ -332,6 +337,24 @@
#endif
}
+/* Mark in DOLOOP_INSNS the instructions that belong to the do-loop part.
+ Use TAIL to recognize that part. */
+static void
+mark_doloop_insns (sbitmap doloop_insns, rtx tail)
+{
+ rtx first_insn_not_to_check, insn;
+
+ /* This is the first instruction which belongs the doloop part. */
+ first_insn_not_to_check = (GET_CODE (PATTERN (tail)) == PARALLEL ? tail
+ : prev_nondebug_insn (tail));
+
+ sbitmap_zero (doloop_insns);
+ for (insn = first_insn_not_to_check; insn != NEXT_INSN (tail);
+ insn = NEXT_INSN (insn))
+ if (NONDEBUG_INSN_P (insn))
+ SET_BIT (doloop_insns, INSN_UID (insn));
+}
+
/* Check if COUNT_REG is set to a constant in the PRE_HEADER block, so
that the number of iterations is a compile-time constant. If so,
return the rtx that sets COUNT_REG to a constant, and set COUNT to
@@ -569,13 +592,12 @@
}
}
-/* Bump the SCHED_TIMEs of all nodes to start from zero. Set the values
- of SCHED_ROW and SCHED_STAGE. */
+/* Bump the SCHED_TIMEs of all nodes by AMOUNT. Set the values of
+ SCHED_ROW and SCHED_STAGE. */
static void
-normalize_sched_times (partial_schedule_ptr ps)
+reset_sched_times (partial_schedule_ptr ps, int amount)
{
int row;
- int amount = PS_MIN_CYCLE (ps);
int ii = ps->ii;
ps_insn_ptr crr_insn;
@@ -584,6 +606,10 @@
{
ddg_node_ptr u = crr_insn->node;
int normalized_time = SCHED_TIME (u) - amount;
+ int new_min_cycle = PS_MIN_CYCLE (ps) - amount;
+ /* The first cycle in row zero after the rotation. */
+ int new_first_cycle_in_row_zero =
+ new_min_cycle + ii - SMODULO (new_min_cycle, ii);
if (dump_file)
fprintf (dump_file, "crr_insn->node=%d, crr_insn->cycle=%d,\
@@ -592,8 +618,30 @@
gcc_assert (SCHED_TIME (u) >= ps->min_cycle);
gcc_assert (SCHED_TIME (u) <= ps->max_cycle);
SCHED_TIME (u) = normalized_time;
- SCHED_ROW (u) = normalized_time % ii;
- SCHED_STAGE (u) = normalized_time / ii;
+ crr_insn->cycle = normalized_time;
+ SCHED_ROW (u) = SMODULO (normalized_time, ii);
+
+ /* If min_cycle is in row zero after the rotation then
+ the stage count can be calculated by dividing the cycle
+ with ii. Otherwise, the calculation is done by dividing the
+ SMSed kernel into two intervals:
+
+ 1) min_cycle <= interval 0 < first_cycle_in_row_zero
+ 2) first_cycle_in_row_zero <= interval 1 < max_cycle
+
+ Cycles in interval 0 are in stage 0. The stage of cycles
+ in interval 1 should be added by 1 to take interval 0 into
+ account. */
+ if (SMODULO (new_min_cycle, ii) == 0)
+ SCHED_STAGE (u) = normalized_time / ii;
+ else
+ {
+ if (crr_insn->cycle < new_first_cycle_in_row_zero)
+ SCHED_STAGE (u) = 0;
+ else
+ SCHED_STAGE (u) =
+ ((SCHED_TIME (u) - new_first_cycle_in_row_zero) / ii) + 1;
+ }
}
}
@@ -646,9 +694,12 @@
/* Do not duplicate any insn which refers to count_reg as it
belongs to the control part.
+ If closing_branch_deps is true the closing branch is scheduled
+ as well and thus should be ignored.
TODO: This should be done by analyzing the control part of
the loop. */
- if (reg_mentioned_p (count_reg, u_node->insn))
+ if (reg_mentioned_p (count_reg, u_node->insn)
+ || JUMP_P (ps_ij->node->insn))
continue;
if (for_prolog)
@@ -894,7 +945,8 @@
basic_block condition_bb = NULL;
edge latch_edge;
gcov_type trip_count = 0;
-
+ sbitmap doloop_insns;
+
loop_optimizer_init (LOOPS_HAVE_PREHEADERS
| LOOPS_HAVE_RECORDED_EXITS);
if (number_of_loops () <= 1)
@@ -919,6 +971,7 @@
setup_sched_infos ();
haifa_sched_init ();
+ doloop_insns = sbitmap_alloc (get_max_uid () + 1);
/* Allocate memory to hold the DDG array one entry for each loop.
We use loop->num as index into this array. */
g_arr = XCNEWVEC (ddg_ptr, number_of_loops ());
@@ -1009,9 +1062,11 @@
continue;
}
- /* Don't handle BBs with calls or barriers, or !single_set insns,
- or auto-increment insns (to avoid creating invalid reg-moves
- for the auto-increment insns).
+ /* Don't handle BBs with calls or barriers or auto-increment insns
+ (to avoid creating invalid reg-moves for the auto-increment insns),
+ or !single_set with the exception of instructions that include
+ count_reg---these instructions are part of the control part
+ that do-loop recognizes.
??? Should handle auto-increment insns.
??? Should handle insns defining subregs. */
for (insn = head; insn != NEXT_INSN (tail); insn = NEXT_INSN (insn))
@@ -1021,7 +1076,8 @@
if (CALL_P (insn)
|| BARRIER_P (insn)
|| (NONDEBUG_INSN_P (insn) && !JUMP_P (insn)
- && !single_set (insn) && GET_CODE (PATTERN (insn)) != USE)
+ && !single_set (insn) && GET_CODE (PATTERN (insn)) != USE
+ && !reg_mentioned_p (count_reg, insn))
|| (FIND_REG_INC_NOTE (insn, NULL_RTX) != 0)
|| (INSN_P (insn) && (set = single_set (insn))
&& GET_CODE (SET_DEST (set)) == SUBREG))
@@ -1048,14 +1104,16 @@
continue;
}
-
- if (! (g = create_ddg (bb, 0)))
+ mark_doloop_insns (doloop_insns, tail);
+ if (! (g = create_ddg (bb, doloop_insns)))
{
if (dump_file)
fprintf (dump_file, "SMS create_ddg failed\n");
continue;
}
-
+ if (dump_file)
+ fprintf (dump_file, "SMS closing_branch_deps: %d\n",
+ g->closing_branch_deps);
g_arr[loop->num] = g;
if (dump_file)
fprintf (dump_file, "...OK\n");
@@ -1157,11 +1215,13 @@
ps = sms_schedule_by_order (g, mii, maxii, node_order);
- if (ps){
- stage_count = PS_STAGE_COUNT (ps);
- gcc_assert(stage_count >= 1);
- }
-
+ if (ps)
+ {
+ stage_count = calculate_stage_count (ps);
+ gcc_assert(stage_count >= 1);
+ PS_STAGE_COUNT(ps) = stage_count;
+ }
+
/* Stage count of 1 means that there is no interleaving between
iterations, let the scheduling passes do the job. */
if (stage_count <= 1
@@ -1182,17 +1242,7 @@
else
{
struct undo_replace_buff_elem *reg_move_replaces;
-
- if (dump_file)
- {
- fprintf (dump_file,
- "SMS succeeded %d %d (with ii, sc)\n", ps->ii,
- stage_count);
- print_partial_schedule (ps, dump_file);
- fprintf (dump_file,
- "SMS Branch (%d) will later be scheduled at cycle %d.\n",
- g->closing_branch->cuid, PS_MIN_CYCLE (ps) - 1);
- }
+ int amount;
/* Set the stage boundaries. If the DDG is built with closing_branch_deps,
the closing_branch was scheduled and should appear in the last (ii-1)
@@ -1202,12 +1252,28 @@
TODO: Revisit the issue of scheduling the insns of the
control part relative to the branch when the control part
has more than one insn. */
- normalize_sched_times (ps);
- rotate_partial_schedule (ps, PS_MIN_CYCLE (ps));
+ amount = (g->closing_branch_deps)? SCHED_TIME (g->closing_branch) + 1:
+ PS_MIN_CYCLE (ps);
+ reset_sched_times (ps, amount);
+ rotate_partial_schedule (ps, amount);
+
set_columns_for_ps (ps);
canon_loop (loop);
+ if (dump_file)
+ {
+ fprintf (dump_file,
+ "SMS succeeded %d %d (with ii, sc)\n", ps->ii,
+ stage_count);
+ print_partial_schedule (ps, dump_file);
+ if (!g->closing_branch_deps)
+ fprintf (dump_file,
+ "SMS Branch (%d) will later be scheduled at \
+ cycle %d.\n",
+ g->closing_branch->cuid, PS_MIN_CYCLE (ps) - 1);
+ }
+
/* case the BCT count is not known , Do loop-versioning */
if (count_reg && ! count_init)
{
@@ -1252,6 +1318,7 @@
}
free (g_arr);
+ sbitmap_free (doloop_insns);
/* Release scheduler data, needed until now because of DFA. */
haifa_sched_finish ();
@@ -1759,8 +1826,9 @@
RESET_BIT (tobe_scheduled, u);
continue;
}
-
- if (JUMP_P (insn)) /* Closing branch handled later. */
+ /* Closing branch handled later unless closing_branch_deps
+ is true. */
+ if (JUMP_P (insn) && !g->closing_branch_deps)
{
RESET_BIT (tobe_scheduled, u);
continue;
@@ -1893,8 +1961,8 @@
if (dump_file)
fprintf (dump_file, "split_row=%d\n", split_row);
- normalize_sched_times (ps);
- rotate_partial_schedule (ps, ps->min_cycle);
+ reset_sched_times (ps, PS_MIN_CYCLE (ps));
+ rotate_partial_schedule (ps, PS_MIN_CYCLE (ps));
rows_new = (ps_insn_ptr *) xcalloc (new_ii, sizeof (ps_insn_ptr));
for (row = 0; row < split_row; row++)
@@ -2571,6 +2639,7 @@
ps_insn_ptr next_ps_i;
ps_insn_ptr first_must_follow = NULL;
ps_insn_ptr last_must_precede = NULL;
+ ps_insn_ptr last_in_row = NULL;
int row;
if (! ps_i)
@@ -2597,8 +2666,37 @@
else
last_must_precede = next_ps_i;
}
+ /* The closing branch must be the last in the row. */
+ if (must_precede
+ && TEST_BIT (must_precede, next_ps_i->node->cuid)
+ && JUMP_P (next_ps_i->node->insn))
+ return false;
+
+ last_in_row = next_ps_i;
}
+ /* If closing_branch_deps is true we are scheduling the closing
+ branch as well. Make sure there is no dependent instruction after
+ it as the branch should be the last instruction. */
+ if (JUMP_P (ps_i->node->insn))
+ {
+ if (first_must_follow)
+ return false;
+ if (last_in_row)
+ {
+ /* Make the branch the last in the row. New instructions
+ will be inserted at the beginning of the row or after the
+ last must_precede instruction thus the branch is guaranteed
+ to remain the last instruction in the row. */
+ last_in_row->next_in_row = ps_i;
+ ps_i->prev_in_row = last_in_row;
+ ps_i->next_in_row = NULL;
+ }
+ else
+ ps->rows[row] = ps_i;
+ return true;
+ }
+
/* Now insert the node after INSERT_AFTER_PSI. */
if (! last_must_precede)
@@ -2820,6 +2918,54 @@
return ps_i;
}
+/* Calculate the stage count of the partial schedule PS. */
+int
+calculate_stage_count (partial_schedule_ptr ps)
+{
+ int stage_count;
+
+ /* If closing_branch_deps is false then the stage
+ boundaries are placed efficiently, meaning that min_cycle will be
+ placed at row 0. Otherwise, the closing branch will be placed in
+ row ii-1. For the later case we assume the final SMSed kernel can
+ be divided into two intervals. This assumption is used for the
+ stage count calculation:
+
+ 1) min_cycle <= interval 0 < first_cycle_in_row_zero
+ 2) first_cycle_in_row_zero <= interval 1 < max_cycle
+ */
+ stage_count =
+ CALC_STAGE_COUNT (PS_MIN_CYCLE (ps), PS_MAX_CYCLE (ps), ps->ii);
+ if (ps->g->closing_branch_deps)
+ {
+ int new_min_cycle;
+ int new_min_cycle_row;
+ int rotation_amount = SCHED_TIME (ps->g->closing_branch) + 1;
+
+ /* This is the new value of min_cycle after the final rotation to
+ bring closing branch into row ii-1. */
+ new_min_cycle = PS_MIN_CYCLE (ps) - rotation_amount;
+ /* This is the row which the the new min_cycle will be placed in. */
+ new_min_cycle_row = SMODULO (new_min_cycle, ps->ii);
+ /* If the row of min_cycle is zero then interval 0 is empty.
+ Otherwise, we need to calculate interval 1 and add it by one
+ to take interval 0 into account. */
+ if (new_min_cycle_row != 0)
+ {
+ int new_max_cycle, first_cycle_in_row_zero;
+
+ new_max_cycle = PS_MAX_CYCLE (ps) - rotation_amount;
+ first_cycle_in_row_zero =
+ new_min_cycle + ps->ii - new_min_cycle_row;
+
+ stage_count =
+ CALC_STAGE_COUNT (first_cycle_in_row_zero, new_max_cycle,
+ ps->ii) + 1;
+ }
+ }
+ return stage_count;
+}
+
/* Rotate the rows of PS such that insns scheduled at time
START_CYCLE will appear in row 0. Updates max/min_cycles. */
void
@@ -0,0 +1,186 @@
2011-02-16 Nathan Sidwell <nathan@codesourcery.com>
Issue #10439
gcc/
* config/arm/unwind-arm.c (enum __cxa_type_match_result): New.
(cxa_type_match): Correct declaration.
(__gnu_unwind_pr_common): Reconstruct
additional indirection when __cxa_type_match returns
succeeded_with_ptr_to_base.
libstdc++/
* libsupc++/eh_arm.c (__cxa_type_match): Construct address of
thrown object here. Return succeded_with_ptr_to_base for all
pointer cases.
=== modified file 'gcc/config/arm/unwind-arm.c'
--- old/gcc/config/arm/unwind-arm.c 2010-08-12 12:39:35 +0000
+++ new/gcc/config/arm/unwind-arm.c 2011-04-08 10:41:46 +0000
@@ -32,13 +32,18 @@
typedef unsigned char bool;
typedef struct _ZSt9type_info type_info; /* This names C++ type_info type */
+enum __cxa_type_match_result
+ {
+ ctm_failed = 0,
+ ctm_succeeded = 1,
+ ctm_succeeded_with_ptr_to_base = 2
+ };
void __attribute__((weak)) __cxa_call_unexpected(_Unwind_Control_Block *ucbp);
bool __attribute__((weak)) __cxa_begin_cleanup(_Unwind_Control_Block *ucbp);
-bool __attribute__((weak)) __cxa_type_match(_Unwind_Control_Block *ucbp,
- const type_info *rttip,
- bool is_reference,
- void **matched_object);
+enum __cxa_type_match_result __attribute__((weak)) __cxa_type_match
+ (_Unwind_Control_Block *ucbp, const type_info *rttip,
+ bool is_reference, void **matched_object);
_Unwind_Ptr __attribute__((weak))
__gnu_Unwind_Find_exidx (_Unwind_Ptr, int *);
@@ -1107,6 +1112,7 @@
_uw rtti;
bool is_reference = (data[0] & uint32_highbit) != 0;
void *matched;
+ enum __cxa_type_match_result match_type;
/* Check for no-throw areas. */
if (data[1] == (_uw) -2)
@@ -1118,17 +1124,31 @@
{
/* Match a catch specification. */
rtti = _Unwind_decode_target2 ((_uw) &data[1]);
- if (!__cxa_type_match (ucbp, (type_info *) rtti,
- is_reference,
- &matched))
- matched = (void *)0;
+ match_type = __cxa_type_match (ucbp,
+ (type_info *) rtti,
+ is_reference,
+ &matched);
}
+ else
+ match_type = ctm_succeeded;
- if (matched)
+ if (match_type)
{
ucbp->barrier_cache.sp =
_Unwind_GetGR (context, R_SP);
- ucbp->barrier_cache.bitpattern[0] = (_uw) matched;
+ // ctm_succeeded_with_ptr_to_base really
+ // means _c_t_m indirected the pointer
+ // object. We have to reconstruct the
+ // additional pointer layer by using a temporary.
+ if (match_type == ctm_succeeded_with_ptr_to_base)
+ {
+ ucbp->barrier_cache.bitpattern[2]
+ = (_uw) matched;
+ ucbp->barrier_cache.bitpattern[0]
+ = (_uw) &ucbp->barrier_cache.bitpattern[2];
+ }
+ else
+ ucbp->barrier_cache.bitpattern[0] = (_uw) matched;
ucbp->barrier_cache.bitpattern[1] = (_uw) data;
return _URC_HANDLER_FOUND;
}
=== modified file 'libstdc++-v3/libsupc++/eh_arm.cc'
--- old/libstdc++-v3/libsupc++/eh_arm.cc 2009-04-09 14:00:19 +0000
+++ new/libstdc++-v3/libsupc++/eh_arm.cc 2011-04-08 10:41:46 +0000
@@ -30,10 +30,11 @@
using namespace __cxxabiv1;
-// Given the thrown type THROW_TYPE, pointer to a variable containing a
-// pointer to the exception object THROWN_PTR_P and a type CATCH_TYPE to
-// compare against, return whether or not there is a match and if so,
-// update *THROWN_PTR_P.
+// Given the thrown type THROW_TYPE, exception object UE_HEADER and a
+// type CATCH_TYPE to compare against, return whether or not there is
+// a match and if so, update *THROWN_PTR_P to point to either the
+// type-matched object, or in the case of a pointer type, the object
+// pointed to by the pointer.
extern "C" __cxa_type_match_result
__cxa_type_match(_Unwind_Exception* ue_header,
@@ -41,51 +42,51 @@
bool is_reference __attribute__((__unused__)),
void** thrown_ptr_p)
{
- bool forced_unwind = __is_gxx_forced_unwind_class(ue_header->exception_class);
- bool foreign_exception = !forced_unwind && !__is_gxx_exception_class(ue_header->exception_class);
- bool dependent_exception =
- __is_dependent_exception(ue_header->exception_class);
+ bool forced_unwind
+ = __is_gxx_forced_unwind_class(ue_header->exception_class);
+ bool foreign_exception
+ = !forced_unwind && !__is_gxx_exception_class(ue_header->exception_class);
+ bool dependent_exception
+ = __is_dependent_exception(ue_header->exception_class);
__cxa_exception* xh = __get_exception_header_from_ue(ue_header);
__cxa_dependent_exception *dx = __get_dependent_exception_from_ue(ue_header);
const std::type_info* throw_type;
+ void *thrown_ptr = 0;
if (forced_unwind)
throw_type = &typeid(abi::__forced_unwind);
else if (foreign_exception)
throw_type = &typeid(abi::__foreign_exception);
- else if (dependent_exception)
- throw_type = __get_exception_header_from_obj
- (dx->primaryException)->exceptionType;
else
- throw_type = xh->exceptionType;
-
- void* thrown_ptr = *thrown_ptr_p;
+ {
+ if (dependent_exception)
+ xh = __get_exception_header_from_obj (dx->primaryException);
+ throw_type = xh->exceptionType;
+ // We used to require the caller set the target of thrown_ptr_p,
+ // but that's incorrect -- the EHABI makes no such requirement
+ // -- and not all callers will set it. Fortunately callers that
+ // do initialize will always pass us the value we calculate
+ // here, so there's no backwards compatibility problem.
+ thrown_ptr = __get_object_from_ue (ue_header);
+ }
+
+ __cxa_type_match_result result = ctm_succeeded;
// Pointer types need to adjust the actual pointer, not
// the pointer to pointer that is the exception object.
// This also has the effect of passing pointer types
// "by value" through the __cxa_begin_catch return value.
if (throw_type->__is_pointer_p())
- thrown_ptr = *(void**) thrown_ptr;
+ {
+ thrown_ptr = *(void**) thrown_ptr;
+ // We need to indicate the indirection to our caller.
+ result = ctm_succeeded_with_ptr_to_base;
+ }
if (catch_type->__do_catch(throw_type, &thrown_ptr, 1))
{
*thrown_ptr_p = thrown_ptr;
-
- if (typeid(*catch_type) == typeid (typeid(void*)))
- {
- const __pointer_type_info *catch_pointer_type =
- static_cast<const __pointer_type_info *> (catch_type);
- const __pointer_type_info *throw_pointer_type =
- static_cast<const __pointer_type_info *> (throw_type);
-
- if (typeid (*catch_pointer_type->__pointee) != typeid (void)
- && (*catch_pointer_type->__pointee !=
- *throw_pointer_type->__pointee))
- return ctm_succeeded_with_ptr_to_base;
- }
-
- return ctm_succeeded;
+ return result;
}
return ctm_failed;
@@ -0,0 +1,134 @@
2011-04-26 Chung-Lin Tang <cltang@codesourcery.com>
Backport from mainline:
2011-03-21 Chung-Lin Tang <cltang@codesourcery.com>
gcc/
* simplify-rtx.c (simplify_binary_operation_1): Handle
(xor (and A B) C) case when B and C are both constants.
gcc/testsuite/
* gcc.target/arm/xor-and.c: New.
2011-03-18 Chung-Lin Tang <cltang@codesourcery.com>
gcc/
* combine.c (try_combine): Do simplification only call of
subst() on i2 even when i1 is present. Update comments.
gcc/testsuite/
* gcc.target/arm/unsigned-extend-1.c: New.
=== modified file 'gcc/combine.c'
--- old/gcc/combine.c 2011-01-06 11:02:44 +0000
+++ new/gcc/combine.c 2011-04-14 13:58:12 +0000
@@ -2939,7 +2939,7 @@
/* It is possible that the source of I2 or I1 may be performing
an unneeded operation, such as a ZERO_EXTEND of something
that is known to have the high part zero. Handle that case
- by letting subst look at the innermost one of them.
+ by letting subst look at the inner insns.
Another way to do this would be to have a function that tries
to simplify a single insn instead of merging two or more
@@ -2964,11 +2964,9 @@
subst_low_luid = DF_INSN_LUID (i1);
i1src = subst (i1src, pc_rtx, pc_rtx, 0, 0, 0);
}
- else
- {
- subst_low_luid = DF_INSN_LUID (i2);
- i2src = subst (i2src, pc_rtx, pc_rtx, 0, 0, 0);
- }
+
+ subst_low_luid = DF_INSN_LUID (i2);
+ i2src = subst (i2src, pc_rtx, pc_rtx, 0, 0, 0);
}
n_occurrences = 0; /* `subst' counts here */
=== modified file 'gcc/simplify-rtx.c'
--- old/gcc/simplify-rtx.c 2010-06-25 20:11:56 +0000
+++ new/gcc/simplify-rtx.c 2011-04-14 13:58:12 +0000
@@ -2413,6 +2413,46 @@
XEXP (op0, 1), mode),
op1);
+ /* Given (xor (and A B) C), using P^Q == (~P&Q) | (~Q&P),
+ we can transform like this:
+ (A&B)^C == ~(A&B)&C | ~C&(A&B)
+ == (~A|~B)&C | ~C&(A&B) * DeMorgan's Law
+ == ~A&C | ~B&C | A&(~C&B) * Distribute and re-order
+ Attempt a few simplifications when B and C are both constants. */
+ if (GET_CODE (op0) == AND
+ && CONST_INT_P (op1)
+ && CONST_INT_P (XEXP (op0, 1)))
+ {
+ rtx a = XEXP (op0, 0);
+ rtx b = XEXP (op0, 1);
+ rtx c = op1;
+ HOST_WIDE_INT bval = INTVAL (b);
+ HOST_WIDE_INT cval = INTVAL (c);
+
+ rtx na_c
+ = simplify_binary_operation (AND, mode,
+ simplify_gen_unary (NOT, mode, a, mode),
+ c);
+ if ((~cval & bval) == 0)
+ {
+ /* Try to simplify ~A&C | ~B&C. */
+ if (na_c != NULL_RTX)
+ return simplify_gen_binary (IOR, mode, na_c,
+ GEN_INT (~bval & cval));
+ }
+ else
+ {
+ /* If ~A&C is zero, simplify A&(~C&B) | ~B&C. */
+ if (na_c == const0_rtx)
+ {
+ rtx a_nc_b = simplify_gen_binary (AND, mode, a,
+ GEN_INT (~cval & bval));
+ return simplify_gen_binary (IOR, mode, a_nc_b,
+ GEN_INT (~bval & cval));
+ }
+ }
+ }
+
/* (xor (comparison foo bar) (const_int 1)) can become the reversed
comparison if STORE_FLAG_VALUE is 1. */
if (STORE_FLAG_VALUE == 1
=== added file 'gcc/testsuite/gcc.target/arm/unsigned-extend-1.c'
--- old/gcc/testsuite/gcc.target/arm/unsigned-extend-1.c 1970-01-01 00:00:00 +0000
+++ new/gcc/testsuite/gcc.target/arm/unsigned-extend-1.c 2011-04-14 13:58:12 +0000
@@ -0,0 +1,9 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -march=armv6" } */
+
+unsigned char foo (unsigned char c)
+{
+ return (c >= '0') && (c <= '9');
+}
+
+/* { dg-final { scan-assembler-not "uxtb" } } */
=== added file 'gcc/testsuite/gcc.target/arm/xor-and.c'
--- old/gcc/testsuite/gcc.target/arm/xor-and.c 1970-01-01 00:00:00 +0000
+++ new/gcc/testsuite/gcc.target/arm/xor-and.c 2011-04-14 13:58:12 +0000
@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-options "-O -march=armv6" } */
+
+unsigned short foo (unsigned short x)
+{
+ x ^= 0x4002;
+ x >>= 1;
+ x |= 0x8000;
+ return x;
+}
+
+/* { dg-final { scan-assembler "orr" } } */
+/* { dg-final { scan-assembler-not "mvn" } } */
+/* { dg-final { scan-assembler-not "uxth" } } */