Comments on: Evolved compression in the x86 ISA I was doing the compilation on OSX (since it can easily compile for both) and you're right, OSX always uses SSE by default. Turning on -mfpmath=387 gets it down to just 116 bytes; smaller than PowerPC! Of course, then you get into the MMX register sharing problem which earns it a big asterisk. :-) I was doing the compilation on OSX (since it can easily compile for both) and you’re right, OSX always uses SSE by default. Turning on -mfpmath=387 gets it down to just 116 bytes; smaller than PowerPC! Of course, then you get into the MMX register sharing problem which earns it a big asterisk. :-)

]]>
By: Andreas Fredriksson/2011/05/15/evolved-compression-in-the-x86-isa/#comment-4094 Andreas Fredriksson Sun, 15 May 2011 10:17:32 +0000 You're right, Thumb2 is probably the closest we've got! I actually considered putting ARM and Thumb2 into the comparison list, but I decided that it wasn't an apples-to-apples comparison with x86 and PPC... ARM and Thumb2 are good ISAs and have been getting faster, but the chips are simply not yet in the same league with x86/PPC performance. I don't know if it's just their low-power heritage, or if there's some core assumption in the ISA (like the barrel shifter) which is difficult to scale to high clock rates, or what. You’re right, Thumb2 is probably the closest we’ve got!

I actually considered putting ARM and Thumb2 into the comparison list, but I decided that it wasn’t an apples-to-apples comparison with x86 and PPC… ARM and Thumb2 are good ISAs and have been getting faster, but the chips are simply not yet in the same league with x86/PPC performance. I don’t know if it’s just their low-power heritage, or if there’s some core assumption in the ISA (like the barrel shifter) which is difficult to scale to high clock rates, or what.

]]>
By: Richard Schreyer/2011/05/15/evolved-compression-in-the-x86-isa/#comment-4090 Richard Schreyer Sun, 15 May 2011 06:52:35 +0000