Re: [HACKERS] [PATCHES] MemSet inline for newNode

2002-11-10 Thread Bruce Momjian
Further update: I ran a test and compared the sizes of the postgres binary: 2926925 with palloc0, 2931293 without, a 0.14% decrease. Not very much, considering the 1-14% speedup of MemSet with a constant. Tom, I really didn't think those conditional tests would be significant, but clearly they

Re: [HACKERS] [PATCHES] MemSet inline for newNode

2002-11-10 Thread Bruce Momjian
OK, my compiler, gcc 2.95 does the optimization at -O2, so I ran the tests here with the attached program, testing a constant of 256 and a variable 'j' equal to 256. I found a 1.7% slowdown with the variable compared to the constant. The 256 value is rather large. For a length of 32, the differe