I just glanced at the patch out of curiosity and the first hunk
caught my eye. There's nothing wrong with the change and I like
how you make the APIs const-correct! Just a note that it looks
like the const in on the basic_block declaration might be missing
an underscore (it should be const_basic_block). Otherwise it's
superfluous. This is repeated in a bunch of places in the file
and its header.
Martin
diff --git a/gcc/gimple-range-cache.cc b/gcc/gimple-range-cache.cc
index 4347485cf98..bdecd212691 100644
--- a/gcc/gimple-range-cache.cc
+++ b/gcc/gimple-range-cache.cc
@@ -132,7 +132,7 @@ non_null_ref::process_name (tree name)
class ssa_block_ranges
{
public:
- virtual void set_bb_range (const basic_block bb, const irange &r) = 0;
+ virtual bool set_bb_range (const basic_block bb, const irange &r) = 0;
virtual bool get_bb_range (irange &r, const basic_block bb) = 0;
virtual bool bb_range_p (const basic_block bb) = 0;
On 6/23/21 8:27 AM, Andrew MacLeod via Gcc-patches wrote:
The introduction of the sparse bitmap on-entry cache for large BB
functions also introduced the concept that the value we ask to be
written to the cache may not be properly represented. It is limited to
15 unique ranges for any given ssa-name, then it reverts to varying for
any additional values to be safe.
This patch adds a boolean return value to the cache set routines,
allowing us to check if the value was properly written.
Other than that, no functional change.
Bootstrap on x86_64-pc-linux-gnu with no regressions. pushed.
Andrew
