GCC 4.8.0 Release Marks Completion of C++ Migration

hypnosec writes "GCC 4.8.0 has been released (download), and with it, the developers of the GNU Compiler Collection have switched to C++ as the implementation language, a project the developers have been working for years. Licensed under the GPLv3 or later, version 4.8.0 of the GCC not only brings with it performance improvements but also adds memory error detector AddressSanitizer, and race condition detection tool the ThreadSanitizer. Developers wanting to build their own version of GCC should have at their disposal a C++ compiler that understands C++ 2003."

Comments

Nice new feature

[EvanED]

From the changelog: "A new general optimization level, -Og, has been introduced. It addresses the need for fast compilation and a superior debugging experience while providing a reasonable level of runtime performance. Overall experience for development should be better than the default optimization level -O0."

This actually sounds really attractive to me... I'll have to try it out.

Re:chicken or egg?

[Greyfox]

Back in '87 in my assembly language class, which was on some ancient DEC machine (Think it was a PDP 11-03 or something,) our instructor gave us all a sheet with the bootstrap code we had to type in, in octal, to make the system jump to the OS instructions on our 8" floppy disks. He showed us how to use a table in our assembly language book to convert assembly code to machine code by hand. If all you're looking at is a slab of iron (Arrhn!) that's where you start! It's a little intimidating imagining having to start with nothing but a table of opcodes and a slab of raw iron that you can enter bits into somehow. I'd think you'd want a basic program loader, so you could save the assembler that you carve out of opcodes. Once you have your assembler stored on disk, you can write your initial C compiler (Not the standard library mind you, just the compiler!) At some point along the way, you compile EMACS.

Of course, all those pokes in the back of Compute! magazine were really just hand-converted machine instructions to be stuffed into specific areas of memory, but I didn't realize that at the time. I forget exactly when I finally made that association. I think it was during that assembly class; I was like "Oh, the number is just an opcode! Derp!" I probably could have squeezed a lot more interesting stuff out of my first computer (A TI 99/4A) if I'd know that at the time.

Re:chicken or egg?

[demonlapin]

I remember seeing a pretty cool 8080 program to zero memory. It went something like:
LXI BC, 0000
LXI HL, 0009
LXI SP, 0008
PUSH BC
PCHL
HL, BC, and SP (the stack pointer) are all 16-bit registers, and the first three instructions occupy 0x0000-0x0008 in memory. PUSH BC is one byte at 0x0009, and PCHL (another one-byte instruction) sets the program counter to the value of HL; essentially it allows you to jump to the address contained in the HL register. The PUSH operation decremented the stack pointer before storing each of the bytes, so you would set the stack pointer to one higher than the highest address you wanted to overwrite. Because the 8080 was little-endian, address 0x0008 was the upper byte of the third instruction's numeric value - it was already zero (if it were big-endian you could just throw in a NOP to get it to align properly). The final PUSH operation - after the program had cycled through all of memory - wiped out the PUSH BC and PCHL operations, and the computer would go to all NOPs.