On 11/20/13 09:47, Andrew MacLeod wrote:
* gimplify.h (gimplify_hasher : typed_free_remove, struct gimplify_ctx):
Move to gimplify.c.
* gimplify.c (gimplify_hasher:typed_free_remove): Relocate here.
(struct gimplify_ctx): Relocate here and add 'malloced' field.
(gimplify_ctxp): Make static.
(ctx_pool, ctx_alloc, ctx_free): Manage a list of struct gimplify_ctx.
(push_gimplify_context): Add default parameters and allocate a
struct from the pool.
(pop_gimplify_context): Free a struct back to the pool.
(gimplify_scan_omp_clauses, gimplify_omp_parallel, gimplify_omp_task,
gimplify_omp_workshare, gimplify_transaction, gimplify_body): Don't
use a local 'struct gimplify_ctx'.
* gimplify-me.c (force_gimple_operand_1, gimple_regimplify_operands):
Likewise.
* omp-low.c (lower_omp_sections, lower_omp_single, lower_omp_master,
lower_omp_ordered, lower_omp_critical, lower_omp_for,
create_task_copyfn, lower_omp_taskreg, lower_omp_target,
lower_omp_teams, execute_lower_omp): Likewise.
* gimple-fold.c (gimplify_and_update_call_from_tree): Likewise.
* tree-inline.c (optimize_inline_calls): Likewise.
I don't see the malloced field in gimplify_ctx. ChangeLog from prior
version?
Any reason not to use xcalloc to allocate & clear the memory in
ctx_alloc. Oh, I see, you want to clear the cached one too. Nevermind.
Should we ever release the list of ctx pointers?
Do our coding standards allow using default arguments:
extern void push_gimplify_context (bool in_ssa = false,
bool rhs_cond_ok = false);
Jeff
