From: Cesar Philippidis <ce...@codesourcery.com>
This patch introduces a new struct offload_attrs, which contains the
details regarding the offload function launch geometry. In addition to
its current usage to neuter worker and vector threads, it will
eventually by used to validate the compile-time launch geometry
requested by the user.
2018-XX-YY Cesar Philippidis <ce...@codesourcery.com>
gcc/
* config/nvptx/nvptx.c (struct offload_attrs): New.
(populate_offload_attrs): New function.
(nvptx_reorg): Use it to extract partitioning mask.
diff --git a/gcc/config/nvptx/nvptx.c b/gcc/config/nvptx/nvptx.c
index fb3e0c7..1b83b3c 100644
--- a/gcc/config/nvptx/nvptx.c
+++ b/gcc/config/nvptx/nvptx.c
@@ -2871,6 +2871,17 @@ nvptx_reorg_uniform_simt ()
}
}
+/* Offloading function attributes. */
+
+struct offload_attrs
+{
+ unsigned mask;
+ int num_gangs;
+ int num_workers;
+ int vector_length;
+ int max_workers;
+};
+
/* Loop structure of the function. The entire function is described as
a NULL loop. */
@@ -4568,6 +4579,56 @@ nvptx_neuter_pars (parallel *par, unsigned modes,
unsigned outer)
nvptx_neuter_pars (par->next, modes, outer);
}
+static void
+populate_offload_attrs (offload_attrs *oa)
+{
+ tree attr = oacc_get_fn_attrib (current_function_decl);
+ tree dims = TREE_VALUE (attr);
+ unsigned ix;
+
+ oa->mask = 0;
+
+ for (ix = 0; ix != GOMP_DIM_MAX; ix++, dims = TREE_CHAIN (dims))
+ {
+ tree t = TREE_VALUE (dims);
+ int size = (t == NULL_TREE) ? 0 : TREE_INT_CST_LOW (t);
+ tree allowed = TREE_PURPOSE (dims);
+
+ if (size != 1 && !(allowed && integer_zerop (allowed)))
+ oa->mask |= GOMP_DIM_MASK (ix);
+
+ switch (ix)
+ {
+ case GOMP_DIM_GANG:
+ oa->num_gangs = size;
+ break;
+
+ case GOMP_DIM_WORKER:
+ oa->num_workers = size;
+ break;
+
+ case GOMP_DIM_VECTOR:
+ oa->vector_length = size;
+ break;
+ }
+ }
+
+ if (oa->vector_length == 0)
+ {
+ /* FIXME: Need a more graceful way to handle large vector
+ lengths in OpenACC routines. */
+ if (!lookup_attribute ("omp target entrypoint",
+ DECL_ATTRIBUTES (current_function_decl)))
+ oa->vector_length = PTX_WARP_SIZE;
+ else
+ oa->vector_length = PTX_VECTOR_LENGTH;
+ }
+ if (oa->num_workers == 0)
+ oa->max_workers = PTX_CTA_SIZE / oa->vector_length;
+ else
+ oa->max_workers = oa->num_workers;
+}
+
#if WORKAROUND_PTXJIT_BUG_2
/* Variant of pc_set that only requires JUMP_P (INSN) if STRICT. This variant
is needed in the nvptx target because the branches generated for
@@ -4749,27 +4810,19 @@ nvptx_reorg (void)
{
/* If we determined this mask before RTL expansion, we could
elide emission of some levels of forks and joins. */
- unsigned mask = 0;
- tree dims = TREE_VALUE (attr);
- unsigned ix;
+ offload_attrs oa;
- for (ix = 0; ix != GOMP_DIM_MAX; ix++, dims = TREE_CHAIN (dims))
- {
- int size = TREE_INT_CST_LOW (TREE_VALUE (dims));
- tree allowed = TREE_PURPOSE (dims);
+ populate_offload_attrs (&oa);
- if (size != 1 && !(allowed && integer_zerop (allowed)))
- mask |= GOMP_DIM_MASK (ix);
- }
/* If there is worker neutering, there must be vector
neutering. Otherwise the hardware will fail. */
- gcc_assert (!(mask & GOMP_DIM_MASK (GOMP_DIM_WORKER))
- || (mask & GOMP_DIM_MASK (GOMP_DIM_VECTOR)));
+ gcc_assert (!(oa.mask & GOMP_DIM_MASK (GOMP_DIM_WORKER))
+ || (oa.mask & GOMP_DIM_MASK (GOMP_DIM_VECTOR)));
/* Discover & process partitioned regions. */
parallel *pars = nvptx_discover_pars (&bb_insn_map);
nvptx_process_pars (pars);
- nvptx_neuter_pars (pars, mask, 0);
+ nvptx_neuter_pars (pars, oa.mask, 0);
delete pars;
}
--
2.7.4
