3 ms·

Can you share the prefetching code you wrote? I tried to use __builtin_prefetch, but couldn't figure out how to make it faster.

by jvns 12y ago

nkurz 12y ago
My approach was somewhat ugly. I went this this: for (int i = 0; i < n; i += 8*16) { __builtin_prefetch(&a[i + 512]); for (int j = 0; j < 8; j++) { __m128i v = _mm_load_si128((__m128i *)&a[i + j*16]); __m128i vl = _mm_unpacklo_epi8(v, vk0); __m128i vh = _mm_unpackhi_epi8(v, vk0); vsum = _mm_add_epi32(vsum, _mm_madd_epi16(vl, vk1)); vsum = _mm_add_epi32(vsum, _mm_madd_epi16(vh, vk1)); } The goal is to issue one prefetch for each 128B block of data you read. There are probably better ways to do this than what I did. I'm hoping the compiler did something reasonable, and haven't really looked at the generated assembly. Also, if it indeed is that case that TLB misses are the major factor (and I think it is), I don't think you will have much success with by adding prefetch alone. Trying right now, I get a slight slowdown with just the prefetch. It may only in combination with hugepages that you get a positive effect.
- jvns 12y ago
  yeah, I tried prefetching like this: https://github.com/jvns/howcomputer/blob/master/bytesum_prefetch.c https://github.com/jvns/howcomputer/blob/master/bytesum_pref... (using MAP_POPULATE instead of hugepages) and got a slight slowdown
- jvns 12y ago
  I profiled my program looking for TLB misses, and got 0 dTLB misses (https://gist.github.com/jvns/a42ff6f48c659cfc4600 https://gist.github.com/jvns/a42ff6f48c659cfc4600).
  - nkurz 12y ago
    I think 'perf' is probably lying to you. Although maybe it's not a lie, as your Gist does contain that line '<not supported> dTLB-misses'. Perf tries very hard to be non-CPU specific, and thus doesn't do a great job of handling the CPU specific stuff. What processor are you running this on? If Intel, you might have luck with some of the more Intel specific wrappers here: https://github.com/andikleen/pmu-tools https://github.com/andikleen/pmu-tools You also might have better luck with 'likwid': https://code.google.com/p/likwid/ https://code.google.com/p/likwid/ Here's the arguments I was giving it to check: sudo likwid -C 1 -g \ INSTR_RETIRED_ANY:FIXC0, \ CPU_CLK_UNHALTED_CORE:FIXC1,\ CPU_CLK_UNHALTED_REF:FIXC2, \ DTLB_LOAD_MISSES_WALK_COMPLETED:PMC0 \ ./bytesum 1gb_file | INSTR_RETIRED_ANY | 7.38826e+08 | | CPU_CLK_UNHALTED_CORE | 5.42765e+08 | | CPU_CLK_UNHALTED_REF | 5.42753e+08 | | DTLB_LOAD_MISSES_WALK_COMPLETED | 1.04509e+06 | sudo likwid -C 1 -g \ INSTR_RETIRED_ANY:FIXC0, \ CPU_CLK_UNHALTED_CORE:FIXC1,\ CPU_CLK_UNHALTED_REF:FIXC2, \ DTLB_LOAD_MISSES_WALK_COMPLETED:PMC0 \ ./hugepage_prefetch 1gb_file | INSTR_RETIRED_ANY | 5.79098e+08 | | CPU_CLK_UNHALTED_CORE | 2.63809e+08 | | CPU_CLK_UNHALTED_REF | 2.63809e+08 | | DTLB_LOAD_MISSES_WALK_COMPLETED | 11970 | The other main advantage of 'likwid' is that it allows you to profile just a section of the code, rather than the program as a whole. For odd political reasons, 'perf' doesn't make this possible. ps. I think your 'argc' check is off by one. Since the name of the program is in argv[0], and argc is the length of argv, you want to check 'argc != 2' to confirm that a filename has been given.

Can you share the prefetching code you wrote? I tried to use __builtin_prefetch, but couldn't figure out how to make it faster.

by jvns 12y ago

Can you share the prefetching code you wrote? I tried to use __builtin_prefetch, but couldn't figure out how to make it faster.

nkurz 12y ago

My approach was somewhat ugly. I went this this: for (int i = 0; i < n; i += 8*16) { __builtin_prefetch(&a[i + 512]); for (int j = 0; j < 8; j++) { __m128i v = _mm_load_si128((__m128i *)&a[i + j*16]); __m128i vl = _mm_unpacklo_epi8(v, vk0); __m128i vh = _mm_unpackhi_epi8(v, vk0); vsum = _mm_add_epi32(vsum, _mm_madd_epi16(vl, vk1)); vsum = _mm_add_epi32(vsum, _mm_madd_epi16(vh, vk1)); } The goal is to issue one prefetch for each 128B block of data you read. There are probably better ways to do this than what I did. I'm hoping the compiler did something reasonable, and haven't really looked at the generated assembly. Also, if it indeed is that case that TLB misses are the major factor (and I think it is), I don't think you will have much success with by adding prefetch alone. Trying right now, I get a slight slowdown with just the prefetch. It may only in combination with hugepages that you get a positive effect.

jvns 12y ago
yeah, I tried prefetching like this: https://github.com/jvns/howcomputer/blob/master/bytesum_prefetch.c https://github.com/jvns/howcomputer/blob/master/bytesum_pref... (using MAP_POPULATE instead of hugepages) and got a slight slowdown
jvns 12y ago
I profiled my program looking for TLB misses, and got 0 dTLB misses (https://gist.github.com/jvns/a42ff6f48c659cfc4600 https://gist.github.com/jvns/a42ff6f48c659cfc4600).
- nkurz 12y ago
  I think 'perf' is probably lying to you. Although maybe it's not a lie, as your Gist does contain that line '<not supported> dTLB-misses'. Perf tries very hard to be non-CPU specific, and thus doesn't do a great job of handling the CPU specific stuff. What processor are you running this on? If Intel, you might have luck with some of the more Intel specific wrappers here: https://github.com/andikleen/pmu-tools https://github.com/andikleen/pmu-tools You also might have better luck with 'likwid': https://code.google.com/p/likwid/ https://code.google.com/p/likwid/ Here's the arguments I was giving it to check: sudo likwid -C 1 -g \ INSTR_RETIRED_ANY:FIXC0, \ CPU_CLK_UNHALTED_CORE:FIXC1,\ CPU_CLK_UNHALTED_REF:FIXC2, \ DTLB_LOAD_MISSES_WALK_COMPLETED:PMC0 \ ./bytesum 1gb_file | INSTR_RETIRED_ANY | 7.38826e+08 | | CPU_CLK_UNHALTED_CORE | 5.42765e+08 | | CPU_CLK_UNHALTED_REF | 5.42753e+08 | | DTLB_LOAD_MISSES_WALK_COMPLETED | 1.04509e+06 | sudo likwid -C 1 -g \ INSTR_RETIRED_ANY:FIXC0, \ CPU_CLK_UNHALTED_CORE:FIXC1,\ CPU_CLK_UNHALTED_REF:FIXC2, \ DTLB_LOAD_MISSES_WALK_COMPLETED:PMC0 \ ./hugepage_prefetch 1gb_file | INSTR_RETIRED_ANY | 5.79098e+08 | | CPU_CLK_UNHALTED_CORE | 2.63809e+08 | | CPU_CLK_UNHALTED_REF | 2.63809e+08 | | DTLB_LOAD_MISSES_WALK_COMPLETED | 11970 | The other main advantage of 'likwid' is that it allows you to profile just a section of the code, rather than the program as a whole. For odd political reasons, 'perf' doesn't make this possible. ps. I think your 'argc' check is off by one. Since the name of the program is in argv[0], and argc is the length of argv, you want to check 'argc != 2' to confirm that a filename has been given.