| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
2004-03-16 Paolo Bonzini <bonzini@gnu.org>
* c-common.c (c_common_type_for_mode): Build vector types on
demand.
(handle_mode_attribute): Deprecate using the mode attribute
to create vector types. Fix indentation.
(vector_type_node_list): Remove.
(handle_vector_size_attribute): Create vector types on demand.
Strip a NON_LVALUE_EXPR from the attribute if there is one.
* c-typeck.c (comptypes): Make vector types compatible if they
have the same underlying mode.
(convert_for_assignment): Use comptypes to convert between
vector types.
* tree.c (build_common_tree_nodes_2): Do not create vector types.
* config/arm/arm.c (arm_init_iwmmxt_builtins): Create necessary
vector types.
* tree.h: Remove vector types.
* config/i386/i386.c (i386_init_mmx_sse_builtins): Likewise.
* config/rs6000/rs6000.c (rs6000_init_builtins): Likewise.
(V16QI_type_node, V2SI_type_node, V2SF_type_node, V4HI_type_node,
V4SI_type_node, V4SF_type_node, V8HI_type_node): New globals.
* doc/extend.texi (Vector Types): Document how to use the
vector_size attribute to create vectors, rather than mode.
* config/arm/mmintrin.h: Use vector_size attribute, not mode.
* config/i386/emmintrin.h: Likewise.
* config/i386/mmintrin.h: Likewise.
* config/i386/xmmintrin.h: Likewise.
* config/sh/ushmedia.h: Likwise.
testsuite/ChangeLog
2004-03-16 Paolo Bonzini <bonzini@gnu.org>
* g++.dg/eh/simd-1.C: Use vector_size attribute, not mode.
* g++.dg/eh/simd-2.C: Likewise.
* g++.dg/init/array10.C: Likewise.
* gcc.c-torture/compile/simd-1.c: Likewise.
* gcc.c-torture/compile/simd-2.c: Likewise.
* gcc.c-torture/compile/simd-3.c: Likewise.
* gcc.c-torture/compile/simd-4.c: Likewise.
* gcc.c-torture/compile/simd-6.c: Likewise.
* gcc.c-torture/execute/simd-1.c: Likewise.
* gcc.c-torture/execute/simd-2.c: Likewise.
* gcc.dg/compat/vector-defs.h: Likewise.
* gcc.dg/20020531-1.c: Likewise.
* gcc.dg/altivec-3.c: Likewise.
* gcc.dg/altivec-4.c: Likewise.
* gcc.dg/altivec-varargs-1.c: Likewise.
* testsuite/gcc.dg/compat/vector-defs.h: Likewise.
* gcc.dg/i386-mmx-3.c: Likewise.
* gcc.dg/i386-sse-4.c: Likewise.
* gcc.dg/i386-sse-5.c: Likewise.
* gcc.dg/i386-sse-8.c: Likewise.
* gcc.dg/simd-1.c: Likewise.
* gcc.dg/20030218-1.c: Likewise. Plus, do not declare
__ev64_opaque__ since the machine description provides it.
Index: c-common.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/c-common.c,v
retrieving revision 1.487
diff -u -r1.487 c-common.c
--- c-common.c 26 Feb 2004 01:24:37 -0000 1.487
+++ c-common.c 10 Mar 2004 10:25:28 -0000
@@ -1874,38 +1874,12 @@
if (mode == TYPE_MODE (build_pointer_type (integer_type_node)))
return unsignedp ? make_unsigned_type (mode) : make_signed_type (mode);
- switch (mode)
+ if (VECTOR_MODE_P (mode))
{
- case V16QImode:
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@79544 138bc75d-0d04-0410-961f-82ee72b054a4
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* optabs.c (expand_vector_binop): Don't store using a SUBREG smaller
than UNITS_PER_WORD, unless this is little endian and the first unit
in this word. Let extract_bit_field decide how to load an element.
Force arguments to matching mode.
(expand_vector_unop): Likewise.
* simplify-rtx.c (simplify_subreg): Don't assume that all vectors
consist of word_mode elements.
* c-typeck.c (build_binary_op): Allow vector types for BIT_AND_EXPR,
BIT_ANDTC_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
(build_unary_op): Allow vector types for BIT_NOT_EPR.
* emit-rtl.c (gen_lowpart_common): Use simplify_gen_subreg for
CONST_VECTOR.
* optabs.c (expand_vector_binop): Try to perform operation in
smaller vector modes with same inner size. Add handling of AND, IOR
and XOR. Reject expansion to inner-mode sized scalars when using
OPTAB_DIRECT. Use simplify_gen_subreg on constants.
(expand_vector_unop): Try to perform operation in smaller vector
modes with same inner size. Add handling of one's complement.
When there is no vector negate operation, try a vector subtract
operation. Use simplify_gen_subreg on constants.
* simplify-rtx.c (simplify_subreg): Add capability to convert vector
constants into smaller vectors with same inner mode, and to
integer CONST_DOUBLEs.
gcc/testsuite:
* gcc.c-torture/execute/simd-1.c (main): Also test &, |, ^, ~.
* gcc.c-torture/execute/simd-2.c (main): Likewise.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@55209 138bc75d-0d04-0410-961f-82ee72b054a4
|
| |
|
|
|
|
|
| |
avoid division by zero.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@55030 138bc75d-0d04-0410-961f-82ee72b054a4
|
|
|
* optabs.c (expand_vector_binop, expand_vector_unop): Don't assume
GET_MODE_UNIT_SIZE (mode) == UNITS_PER_WORD.
gcc/testsuite:
* gcc.c-torture/execute/simd-2.c: New test.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@54994 138bc75d-0d04-0410-961f-82ee72b054a4
|
