Adam Nemet wrote:
Jeff Law writes:
Ian Lance Taylor wrote:
Adam Nemet <ane...@caviumnetworks.com> writes:
I am trying to understand the checkin by Jeff Law from about 11 years ago:
r19204 | law | 1998-04-14 01:04:21 -0700 (Tue, 14 Apr 1998) | 4 lines
* combine.c (simplify_rtx, case TRUNCATE): Respect value of
TRULY_NOOP_TRUNCATION.
Index: combine.c
===================================================================
--- combine.c (revision 19018)
+++ combine.c (revision 19204)
@@ -3736,7 +3736,9 @@ simplify_rtx (x, op0_mode, last, in_dest
if (GET_MODE_CLASS (mode) == MODE_PARTIAL_INT)
break;
- if (GET_MODE_BITSIZE (mode) <= HOST_BITS_PER_WIDE_INT)
+ if (GET_MODE_BITSIZE (mode) <= HOST_BITS_PER_WIDE_INT
+ && TRULY_NOOP_TRUNCATION (GET_MODE_BITSIZE (mode),
+ GET_MODE_BITSIZE (GET_MODE (XEXP (x, 0)))))
SUBST (XEXP (x, 0),
force_to_mode (XEXP (x, 0), GET_MODE (XEXP (x, 0)),
GET_MODE_MASK (mode), NULL_RTX, 0));
This optimization simplifies the input to a truncate by only computing bits
that won't be eliminated by the truncation. Normally these are the bits in
the output mode mask. Note that the optimization does not change the truncate
into a low-part subreg, which would pretty automatically warrant the
TRULY_NOOP_TRUNCATION check.
I agree that this patch looks wrong in todays compiler. There should be
no need to call TRULY_NOOP_TRUNCATION if you are in a TRUNCATE anyhow.
Based on reviewing my old notes, we do have to ensure that combine
doesn't replace a TRUNCATE with a SUBREG as that can result in having a
32bit value that isn't sign-extended, which clearly causes MIPS64 ports
grief.
Thanks for the long information in your other reply.
As I said in the original email, we are not turning a TRUNCATE into a SUBREG
in this transformation. We're just simplifying the input expression to
truncate with the knowledge that only the truncated mode bits are relevant
from the input. At the end we are still left with a truncate expression but
possibly with a simpler operand.
Agreed.
I think it's a distinct possibility that the patch in question just
papered over the problem and the Oct 1998 patch actually fixed the problem.
I've got no objections to removing the code you're referring to.
Jeff
Adam
