Summary: | [4.0 Regression] ICE with SSE code in extract_insn, at recog.c:2020 | ||
---|---|---|---|
Product: | gcc | Reporter: | Michael Cieslinski <micis> |
Component: | target | Assignee: | Not yet assigned to anyone <unassigned> |
Status: | RESOLVED WONTFIX | ||
Severity: | normal | CC: | gcc-bugs |
Priority: | P3 | Keywords: | ice-on-valid-code |
Version: | 4.0.3 | ||
Target Milestone: | 4.0.4 | ||
Host: | Target: | x86_64-*-linux-gnu | |
Build: | Known to work: | 4.2.0 3.4.0 | |
Known to fail: | 4.0.3 | Last reconfirmed: | 2006-09-03 06:28:09 |
Description
Michael Cieslinski
2006-07-18 14:31:17 UTC
(reg:V2SI 0 ax) HUH, how can that happen. Reload 1: reload_in (V2SI) = (const_vector:V2SI [ (const_int -1 [0xffffffffffffffff]) (const_int -1 [0xffffffffffffffff]) ]) GENERAL_REGS, RELOAD_FOR_INPUT (opnum = 1), can't combine reload_in_reg: (reg:V2SI 88) reload_reg_rtx: (reg:V2SI 0 ax) There is a problem first with overflowed integers that the back-end does not understand in 4.0.x which was fixed in 4.1.0. This is also a regression from 3.4.0 so confirmed. Won't fix in GCC-4.0.x. |