[Bug gas/6957] i386 NOPs must be derived from march not mtune
--- Additional Comments From dsd at laptop dot org 2010-06-08 21:49 --- Created an attachment (id=4835) --> (http://sourceware.org/bugzilla/attachment.cgi?id=4835&action=view) testcase.s -- http://sourceware.org/bugzilla/show_bug.cgi?id=6957 --- You are receiving this mail because: --- You are on the CC list for the bug, or are watching someone who is. ___ bug-binutils mailing list bug-binutils@gnu.org http://lists.gnu.org/mailman/listinfo/bug-binutils
[Bug gas/6957] i386 NOPs must be derived from march not mtune
--- Additional Comments From dsd at laptop dot org 2010-06-08 21:53 --- (In reply to comment #8) > I agree. If it is true, assembler is broken. If you can show me > a small testcase, I will fix it. I can confirm the bug and I'd like to take you up on that offer, but I may have fallen short of the *small* test case :) I attached testcase.s which is some assembly caught from a gcc invokation inside glibc's compile process. It's 1300 lines (sorry!) but does confirm the usage of the nopl instruction when -mtune=i686 is passed : $ as -Qy -o foo testcase.s $ objdump -d foo | grep nopl $ as -Qy -mtune=i686 -o foo testcase.s $ objdump -d foo | grep nopl 3d: 0f 1f 00nopl (%eax) Reproduced on both: GNU assembler version 2.19.51.0.14-37.fc12 20090722 GNU assembler version 2.20.51.0.2-17.fc14 20091009 Someone more savvy with assembly code could probably cut down my test case into something more trivial with a small amount of effort - I'm a bit out of my comfort zone. -- What|Removed |Added Status|RESOLVED|REOPENED Resolution|WORKSFORME | http://sourceware.org/bugzilla/show_bug.cgi?id=6957 --- You are receiving this mail because: --- You are on the CC list for the bug, or are watching someone who is. ___ bug-binutils mailing list bug-binutils@gnu.org http://lists.gnu.org/mailman/listinfo/bug-binutils
[Bug gas/6957] i386 NOPs must be derived from march not mtune
--- Additional Comments From dsd at laptop dot org 2010-06-08 22:04 --- Created an attachment (id=4836) --> (http://sourceware.org/bugzilla/attachment.cgi?id=4836&action=view) testcase2.s Duh. The previous testcase.s was generated using "gcc -g" so its naturally huge. Taking off the -g results in a much smaller (95 line) test case which shows the issue in an identical manner. -- http://sourceware.org/bugzilla/show_bug.cgi?id=6957 --- You are receiving this mail because: --- You are on the CC list for the bug, or are watching someone who is. ___ bug-binutils mailing list bug-binutils@gnu.org http://lists.gnu.org/mailman/listinfo/bug-binutils
[Bug gas/6957] i386 NOPs must be derived from march not mtune
--- Additional Comments From dsd at laptop dot org 2010-06-08 22:07 --- Created an attachment (id=4837) --> (http://sourceware.org/bugzilla/attachment.cgi?id=4837&action=view) testcase3.s One more. By deleting (more or less) random lines from testcase2.s I came up with a 9-line test case which produces 2 nopl instructions when assembled with -mtune=i686. $ as -mtune=i686 -o foo testcase3.s $ objdump -d foo Disassembly of section .text: <.text>: 0: 55 push %ebp 1: 5d pop%ebp 2: c3 ret 3: 0f 1f 44 00 00 nopl 0x0(%eax,%eax,1) 8: c3 ret 9: 0f 1f 80 00 00 00 00nopl 0x0(%eax) -- http://sourceware.org/bugzilla/show_bug.cgi?id=6957 --- You are receiving this mail because: --- You are on the CC list for the bug, or are watching someone who is. ___ bug-binutils mailing list bug-binutils@gnu.org http://lists.gnu.org/mailman/listinfo/bug-binutils
[Bug gas/6957] i386 NOPs must be derived from march not mtune
--- Additional Comments From dsd at laptop dot org 2010-06-08 22:17 --- Oops, sorry for the test case spam. I did not see your comment. I understand now. "as -mtune=i686" basically works out to be "as -march=i686 -mtune=i686" Hence generating i686-specific instructions is acceptable/expected. My confusion stemmed from the fact that this is a minor behavioural difference from the equivalently named parameters to gcc. (maybe this is why gcc does not pass those parameters down to the assembler by default) -- http://sourceware.org/bugzilla/show_bug.cgi?id=6957 --- You are receiving this mail because: --- You are on the CC list for the bug, or are watching someone who is. ___ bug-binutils mailing list bug-binutils@gnu.org http://lists.gnu.org/mailman/listinfo/bug-binutils
