http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56114
Bug #: 56114 Summary: x86_64-linux-gnu-gcc generate wrong asm instruction MOVABS for intel syntax Classification: Unclassified Product: gcc Version: 4.7.2 Status: UNCONFIRMED Severity: normal Priority: P3 Component: target AssignedTo: unassig...@gcc.gnu.org ReportedBy: akob...@mail.ru For example file l.c long foo2 (void) { return *(volatile long*)0xFEE00000; } x86_64-linux-gnu-gcc -c -save-temps l.c -O2 -masm=intel cat l.s .file "l.c" .intel_syntax noprefix .text .p2align 4,,15 .globl foo2 .type foo2, @function foo2: .LFB0: .cfi_startproc movabs rax, 4276092928 ret This is erroneous instruction, because movabs rax, 4276092928 loads immediate data. This code must be movabs rax, [4276092928] with square brackets. When we try to read 32-bit data from memory, then we get error message long foo2 (void) { return *(volatile int*)0xFEE00000; } x86_64-linux-gnu-gcc -c -save-temps l.c -O2 -masm=intel cat l.s .file "l.c" .intel_syntax noprefix .text .p2align 4,,15 .globl foo2 .type foo2, @function foo2: .LFB0: .cfi_startproc movabs eax, 4276092928 cdqe ret Because movabs allows load 64-bit only immediate data. Here is GCC loose square brackets.