#INTEL BURN TEST HASWELL UPDATE#

Perhaps a microcode update changed something.) ( update: trying my test code again now, even one vxorps seems to clean the register. But perhaps the difference is just in creating a gap between issue of the vpaddb into the back-end and the first vmulpd in my test loop I 'dirty' / pollute the register right before the loop.) AFAIK, xor-zeroing is always eliminated, unlike mov-elimination which sometimes fails. (On Skylake, it seems 3 or more vxorpd zeroing instructions in a row before the loop happen to work, removing the extra bypass latency. I'm surprised that 'infection' survived across a zeroing idiom, especially an FP zeroing instruction like vxorpd, but I can reproduce that effect on SKL (i7-6700k, counting clock cycles directly in a test loop with perf on Linux instead of messing around with time and frequency.) īypass-delay latency is 'sticky': an integer SIMD instruction can 'infect' all future instructions that read that value, even long after the instruction is done. I've been meaning to write something up about this for a few years now, since noticing it on Skylake. It seems related to upper half state, but I don't know which one is cleaner: vxorpd ymm1, ymm1, ymm1 (move from old value doesn't work)Īs for VEX xmm, the condition seems a little blur. In the case of ymm instructions, I guess the conditions to avoid 1 extra cycle are: There are other instructions to achieve the same effect: vmovupd ymm1, ymm0īut these instructions cannot: vmovupd ymm1, ymm2 suppose ymm2 is zeroed Vaddpd ymm1, ymm1, ymm1 added this lineįinally I get the result of 5 cycle. MORE RESULTS results on Core i5 4590 ymm0, ymm0, ymm1 6.056094 Is there an explanation? Is my test code problematic? Strange enough, if I change vmulpd ymm0, ymm0, ymm1 to vmulpd ymm0, ymm0, ymm0, then the output become: latency = 5.093745. My CPU is Core i5 4590, I locked its frequency at 3.3GHz. SetThreadAffinityMask(GetCurrentThread(), 1) // Avoid context switch I write some test code, but all of the results are 1 cycle slower. In Intel Intrinsics Guide, vmulpd and vfmadd213pd has latency of 5, vaddpd has latency of 3.

Intel burn test haswell