2003-08-19 Chris Demetriou <cgd@broadcom.com>
* config/mips/mips.md: Adjust SI-mode "trap_if" instruction
to use better predicates and constraints. Define new
instruction to handle "trap_if" with DI-mode arguments.
(conditional_trap): FAIL if trap code is not 0.
From-SVN: r70570
2003-08-19 Andrew Pinski <pinskia@physics.uc.edu>
* config/i386/i386.c (legitimate_pic_address_disp_p): Change the
strstr with $pb to a strcompare with "<pic base>"
(ix86_output_addr_diff_elt): Output the real pic base.
From-SVN: r70569
PR libgcj/11575
* java/io/natFileDescriptorWin32.cc (open): Set create
flag to OPEN_AWAYS when READ & WRITE regardless of APPEND flag.
Honor EXCL when openning with WRITE flag.
From-SVN: r70565
* include/jvm.h: New class _Jv_TempUTFString (helper class for
getting a temporary C string from a jstring)
New macro JV_TEMP_UTF_STRING, which leverages _Jv_TempUTFString
but uses a stack buffer if the string length is less than 256
bytes.
From-SVN: r70564
* config/mips/mips.c (mips_va_arg): If EABI_FLOAT_VARARGS_P,
expect SFmode and DFmode arguments to be passed in FPRs,
regardless of the underlying type.
From-SVN: r70557
PR libgcj/11951:
* java/lang/reflect/natMethod.cc (_Jv_CallAnyMethodA): Returns
void. Throw VirtualMachineError if ffi fails. Initialize return
value. Added is_jni_call argument; only wrap exception if not a
JNI call. Use descriptive message if operation not supported.
(_Jv_GetTypesFromSignature): Use declaring class' loader to find
array class.
* include/jvm.h (_Jv_CallAnyMethodA): Updated declaration.
* jni.cc (_Jv_JNI_CallAnyMethodV): Updated for new form of
_Jv_CallAnyMethodA.
(_Jv_JNI_CallAnyMethodA): Likewise.
(_Jv_JNI_CallAnyVoidMethodV): Likewise.
(_Jv_JNI_CallAnyVoidMethodA): Likewise.
From-SVN: r70544
* config/mips/mips.md: Quote C code in braces. Remove use of
fake const0_rtx operands. Remove double backslashes. Use \;.
Remove workarounds for bogus warnings.
From-SVN: r70535
* config/mips/mips.md (muldf3, mulsf3): Don't call a gen_* function.
(muldf3_internal, muldf3_r4300): Select based on TARGET_4300_MUL_FIX
rather than TARGET_MIPS4300.
(mulsf3_internal, mulsf3_r4300): Likewise.
From-SVN: r70534
2003-08-17 Phil Edwards <pme@gcc.gnu.org>
* configure.ac: GCC_NO_EXECUTABLES was supposed to be commented
in the patch from 3 minutes ago. Boy, is my face red.
* configure: At least I remembered to regenerate this.
From-SVN: r70523
2003-08-17 Phil Edwards <pme@gcc.gnu.org>
* acinclude.m4 (GLIBCXX_EXPORT_INCLUDES): Remove LIBMATH_INCLUDES
and LIBSUPCXX_INCLUDES. Re-purpose TOPLEVEL_INCLUDES to refer to
things from the top level.
* configure.ac (GLIBCXX_IS_NATIVE): Determine earlier and re-order.
Comment out the conditionals for CANADIAN and GLIBCXX_BUILD_LIBMATH
(currently unused). Strip the fake-VPATH shell fragment from
automake-generated rules, if present.
* linkage.m4: Add comment.
* fragment.am: New file, containing factored-out common settings.
(AM_CPPFLAGS): Absorb the deprecated INCLUDES variable contents.
* Makefile.am: Include fragment.am. Remove common variables.
* include/Makefile.am: Likewise.
* libmath/Makefile.am: Likewise.
* libsupc++/Makefile.am: Likewise.
* po/Makefile.am: Likewise. Print rules during check.
* src/Makefile.am: Likewise.
* testsuite/Makefile.am: Likewise.
* aclocal.m4, configure, Makefile.in, include/Makefile.in,
libmath/Makefile.in, libsupc++/Makefile.in, po/Makefile.in,
src/Makefile.in, testsuite/Makefile.in: Regenerate.
From-SVN: r70522
* simplify-rtx.c (associative_constant_p): New function to test
whether an RTX expression is an immediate constant.
(simplify_associative_operation): New function to perform some
reassociation optimizations of associative binary expressions.
(simplify_binary_operation): Use simplify_associative_operation
to simplify PLUS, MULT, AND, IOR, XOR, SMIN, SMAX, UMIN and UMAX.
Floating point expressions are only reassociated when unsafe
math optimizations are permitted.
From-SVN: r70521
* gcc_release (build_sources): Use two new variables EXPORTTAG and
EXPORTDATE to make the extraction of sources more transparent and
also allow snapshots (off mainline) without a tag.
From-SVN: r70507
PR c++/11512
* stmt.c (expand_expr_stmt_value): Don't warn about any void
typed expression.
cp:
PR c++/11512
* cvt.c (convert_to_void): Indicate which side of conditional has
no effects, and rhs of comma operator. Test for no sideeffect
expressions here and always build a convert expr.
* init.c (expand_default_init): Convert the init to void.
* typeck.c (build_x_compound_expr): Do not check for side effects
here.
(build_compound_expr): Do not convert lhs when building a
template.
testsuite:
PR C++/11512
* g++.dg/template/warn1.C: New.
From-SVN: r70505
* i386.c (ix86_fntype_regparm): Rename from ...
(ix86_function_regparm): ... this one; add fastcall and local
functions.
(ix86_function_ok_for_sibcall): Update.
(ix86_return_pops_args): Likewise.
(init_cumulative_args): Likewise.
(x86_can_output_mi_thunk): Likewise.
(function_arg): Fix formating.
(x86_this_parameter): Fix fastcall.
(x86_output_mi_thunk): Likewise.
* cgraph.c (cgraph_mark_needed_node): Do not mark functions without
body as reachable; mark nested functions as needed too.
(dump_cgraph): Do not output global.calls.
* cgraph.h (cgraph_global_info): Kill.
* cgraphunit.c (cgraph_finalize_function): Enqueue needed functions.
(record_call_1): Speedup.
(cgraph_analyze_function): Break out from ...; compute inlining
parameters.
(cgraph_finalize_compilation_unit): ... here.
(cgraph_mark_inline): Kill computation of calls.
(cgraph_decide_inlining): Do not compute most of initial values.
From-SVN: r70504
* fold-const.c (negate_expr_p): MULT_EXPRs and RDIV_EXPRs are easy
to negate if either operand is easy to negate, if we don't care
about sign-dependent rounding.
(negate_expr): Make the logic to negate a REAL_CST explicit.
Attempt to negate a MULT_EXPR or RDIV_EXPR by negating an operand
that's easy to negate, if we don't honor sign-dependent rounding.
(fold <MULT_EXPR>): Optimize -A * B as A * -B if B is easy to
negate, and the symmetric A * -B as -A * B if A is easy to negate.
(fold <RDIV_EXPR>): Likewise, optimize -A/B and C/-D as A/-B and
-C/D if B and C are cheap to negate. Add an explicit rule to
optimize X/-1.0 as -X when we don't care about signaling NaNs.
From-SVN: r70455