Skip to content
Snippets Groups Projects
  1. Dec 10, 2019
  2. Dec 09, 2019
  3. Dec 06, 2019
    • Ömer Sinan Ağacan's avatar
      Fix compacting GC bug when chaining tagged and non-tagged fields together · da04db1a
      Ömer Sinan Ağacan authored
      Currently compacting GC has the invariant that in a chain all fields are tagged
      the same. However this does not really hold: root pointers are not tagged, so
      when we thread a root we initialize a chain without a tag. When the pointed
      objects is evaluated and we have more pointers to it from the heap, we then add
      *tagged* fields to the chain (because pointers to it from the heap are tagged),
      ending up chaining fields with different tags (pointers from roots are NOT
      tagged, pointers from heap are). This breaks the invariant and as a result
      compacting GC turns tagged pointers into non-tagged.
      
      This later causes problem in the generated code where we do reads assuming that
      the pointer is aligned, e.g.
      
          0x7(%rax) -- assumes that pointer is tagged 1
      
      which causes misaligned reads. This caused #17088.
      
      We fix this using the "pointer tagging for large families" patch (#14373,
      !1742):
      
      - With the pointer tagging patch the GC can know what the tagged pointer to a
        CONSTR should be (previously we'd need to know the family size -- large
        families are always tagged 1, small families are tagged depending on the
        constructor).
      
      - Since we now know what the tags should be we no longer need to store the
        pointer tag in the info table pointers when forming chains in the compacting
        GC.
      
      As a result we no longer need to tag pointers in chains with 1/2 depending on
      whether the field points to an info table pointer, or to another field: an info
      table pointer is always tagged 0, everything else in the chain is tagged 1. The
      lost tags in pointers can be retrieved by looking at the info table.
      
      Finally, instead of using tag 1 for fields and tag 0 for info table pointers, we
      use two different tags for fields:
      
      - 1 for fields that have untagged pointers
      - 2 for fields that have tagged pointers
      
      When unchaining we then look at the pointer to a field, and depending on its tag
      we either leave a tagged pointer or an untagged pointer in the field.
      
      This allows chaining untagged and tagged fields together in compacting GC.
      
      Fixes #17088
      da04db1a
    • Gabor Greif's avatar
      Implement pointer tagging for big families (#14373) · b911c532
      Gabor Greif authored
      Formerly we punted on these and evaluated constructors always got a tag
      of 1.
      
      We now cascade switches because we have to check the tag first and when
      it is MAX_PTR_TAG then get the precise tag from the info table and
      switch on that. The only technically tricky part is that the default
      case needs (logical) duplication. To do this we emit an extra label for
      it and branch to that from the second switch. This avoids duplicated
      codegen.
      
      Here's a simple example of the new code gen:
      
          data D = D1 | D2 | D3 | D4 | D5 | D6 | D7 | D8
      
      On a 64-bit system previously all constructors would be tagged 1. With
      the new code gen D7 and D8 are tagged 7:
      
          [Lib.D7_con_entry() {
               ...
               {offset
                 c1eu: // global
                     R1 = R1 + 7;
                     call (P64[Sp])(R1) args: 8, res: 0, upd: 8;
               }
           }]
      
          [Lib.D8_con_entry() {
               ...
               {offset
                 c1ez: // global
                     R1 = R1 + 7;
                     call (P64[Sp])(R1) args: 8, res: 0, upd: 8;
               }
           }]
      
      When switching we now look at the info table only when the tag is 7. For
      example, if we derive Enum for the type above, the Cmm looks like this:
      
          c2Le:
              _s2Js::P64 = R1;
              _c2Lq::P64 = _s2Js::P64 & 7;
              switch [1 .. 7] _c2Lq::P64 {
                  case 1 : goto c2Lk;
                  case 2 : goto c2Ll;
                  case 3 : goto c2Lm;
                  case 4 : goto c2Ln;
                  case 5 : goto c2Lo;
                  case 6 : goto c2Lp;
                  case 7 : goto c2Lj;
              }
      
          // Read info table for tag
          c2Lj:
              _c2Lv::I64 = %MO_UU_Conv_W32_W64(I32[I64[_s2Js::P64 & (-8)] - 4]);
              if (_c2Lv::I64 != 6) goto c2Lu; else goto c2Lt;
      
      Generated Cmm sizes do not change too much, but binaries are very
      slightly larger, due to the fact that the new instructions are longer in
      encoded form. E.g. previously entry code for D8 above would be
      
          00000000000001c0 <Lib_D8_con_info>:
           1c0:	48 ff c3             	inc    %rbx
           1c3:	ff 65 00             	jmpq   *0x0(%rbp)
      
      With this patch
      
          00000000000001d0 <Lib_D8_con_info>:
           1d0:	48 83 c3 07          	add    $0x7,%rbx
           1d4:	ff 65 00             	jmpq   *0x0(%rbp)
      
      This is one byte longer.
      
      Secondly, reading info table directly and then switching is shorter
      
          _c1co:
                  movq -1(%rbx),%rax
                  movl -4(%rax),%eax
                  // Switch on info table tag
                  jmp *_n1d5(,%rax,8)
      
      than doing the same switch, and then for the tag 7 doing another switch:
      
          // When tag is 7
          _c1ct:
                  andq $-8,%rbx
                  movq (%rbx),%rax
                  movl -4(%rax),%eax
                  // Switch on info table tag
                  ...
      
      Some changes of binary sizes in actual programs:
      
      - In NoFib the worst case is 0.1% increase in benchmark "parser" (see
        NoFib results below). All programs get slightly larger.
      
      - Stage 2 compiler size does not change.
      
      - In "containers" (the library) size of all object files increases
        0.0005%. Size of the test program "bitqueue-properties" increases
        0.03%.
      
      nofib benchmarks kindly provided by Ömer (@osa1):
      
      NoFib Results
      =============
      
      --------------------------------------------------------------------------------
              Program           Size    Allocs    Instrs     Reads    Writes
      --------------------------------------------------------------------------------
                   CS          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
                  CSD          +0.0%      0.0%      0.0%     +0.0%     +0.0%
                   FS          +0.0%      0.0%      0.0%     +0.0%      0.0%
                    S          +0.0%      0.0%     -0.0%      0.0%      0.0%
                   VS          +0.0%      0.0%     -0.0%     +0.0%     +0.0%
                  VSD          +0.0%      0.0%     -0.0%     +0.0%     -0.0%
                  VSM          +0.0%      0.0%      0.0%      0.0%      0.0%
                 anna          +0.0%      0.0%     +0.1%     -0.9%     -0.0%
                 ansi          +0.0%      0.0%     -0.0%     +0.0%     +0.0%
                 atom          +0.0%      0.0%      0.0%      0.0%      0.0%
               awards          +0.0%      0.0%     -0.0%     +0.0%      0.0%
               banner          +0.0%      0.0%     -0.0%     +0.0%      0.0%
           bernouilli          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
         binary-trees          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
                boyer          +0.0%      0.0%     +0.0%      0.0%     -0.0%
               boyer2          +0.0%      0.0%     +0.0%      0.0%     -0.0%
                 bspt          +0.0%      0.0%     +0.0%     +0.0%      0.0%
            cacheprof          +0.0%      0.0%     +0.1%     -0.8%      0.0%
             calendar          +0.0%      0.0%     -0.0%     +0.0%     -0.0%
             cichelli          +0.0%      0.0%     +0.0%      0.0%      0.0%
              circsim          +0.0%      0.0%     -0.0%     -0.1%     -0.0%
             clausify          +0.0%      0.0%     +0.0%     +0.0%      0.0%
        comp_lab_zift          +0.0%      0.0%     +0.0%      0.0%     -0.0%
             compress          +0.0%      0.0%     +0.0%     +0.0%      0.0%
            compress2          +0.0%      0.0%      0.0%      0.0%      0.0%
          constraints          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
         cryptarithm1          +0.0%      0.0%     +0.0%      0.0%      0.0%
         cryptarithm2          +0.0%      0.0%     +0.0%     -0.0%      0.0%
                  cse          +0.0%      0.0%     +0.0%     +0.0%      0.0%
         digits-of-e1          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
         digits-of-e2          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
               dom-lt          +0.0%      0.0%     +0.0%     +0.0%      0.0%
                eliza          +0.0%      0.0%     -0.0%     +0.0%      0.0%
                event          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
          exact-reals          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
               exp3_8          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
               expert          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
       fannkuch-redux          +0.0%      0.0%     +0.0%      0.0%      0.0%
                fasta          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
                  fem          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
                  fft          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
                 fft2          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
             fibheaps          +0.0%      0.0%     +0.0%     +0.0%      0.0%
                 fish          +0.0%      0.0%     +0.0%     +0.0%      0.0%
                fluid          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
               fulsom          +0.0%      0.0%     +0.0%     -0.0%     +0.0%
               gamteb          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
                  gcd          +0.0%      0.0%     +0.0%     +0.0%      0.0%
          gen_regexps          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
               genfft          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
                   gg          +0.0%      0.0%      0.0%     -0.0%      0.0%
                 grep          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
               hidden          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
                  hpg          +0.0%      0.0%     +0.0%     -0.1%     -0.0%
                  ida          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
                infer          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
              integer          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
            integrate          +0.0%      0.0%      0.0%     +0.0%      0.0%
         k-nucleotide          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
                kahan          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
              knights          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
               lambda          +0.0%      0.0%     +1.2%     -6.1%     -0.0%
           last-piece          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
                 lcss          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
                 life          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
                 lift          +0.0%      0.0%     +0.0%     +0.0%      0.0%
               linear          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
            listcompr          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
             listcopy          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
             maillist          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
               mandel          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
              mandel2          +0.0%      0.0%     +0.0%     +0.0%     -0.0%
                 mate          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
              minimax          +0.0%      0.0%     -0.0%     +0.0%     -0.0%
              mkhprog          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
           multiplier          +0.0%      0.0%      0.0%     +0.0%     -0.0%
               n-body          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
             nucleic2          +0.0%      0.0%     +0.0%     +0.0%     -0.0%
                 para          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
            paraffins          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
               parser          +0.1%      0.0%     +0.4%     -1.7%     -0.0%
              parstof          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
                  pic          +0.0%      0.0%     +0.0%      0.0%     -0.0%
             pidigits          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
                power          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
               pretty          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
               primes          +0.0%      0.0%     +0.0%      0.0%      0.0%
            primetest          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
               prolog          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
               puzzle          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
               queens          +0.0%      0.0%      0.0%     +0.0%     +0.0%
              reptile          +0.0%      0.0%     +0.0%     +0.0%      0.0%
      reverse-complem          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
              rewrite          +0.0%      0.0%     +0.0%      0.0%     -0.0%
                 rfib          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
                  rsa          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
                  scc          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
                sched          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
                  scs          +0.0%      0.0%     +0.0%     +0.0%      0.0%
               simple          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
                solid          +0.0%      0.0%     +0.0%     +0.0%      0.0%
              sorting          +0.0%      0.0%     +0.0%     -0.0%      0.0%
        spectral-norm          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
               sphere          +0.0%      0.0%     +0.0%     -1.0%      0.0%
               symalg          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
                  tak          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
            transform          +0.0%      0.0%     +0.4%     -1.3%     +0.0%
             treejoin          +0.0%      0.0%     +0.0%     -0.0%      0.0%
            typecheck          +0.0%      0.0%     -0.0%     +0.0%      0.0%
              veritas          +0.0%      0.0%     +0.0%     -0.1%     +0.0%
                 wang          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
            wave4main          +0.0%      0.0%     +0.0%      0.0%     -0.0%
         wheel-sieve1          +0.0%      0.0%     +0.0%     +0.0%     +0.0%
         wheel-sieve2          +0.0%      0.0%     +0.0%     +0.0%      0.0%
                 x2n1          +0.0%      0.0%     +0.0%     +0.0%      0.0%
      --------------------------------------------------------------------------------
                  Min          +0.0%      0.0%     -0.0%     -6.1%     -0.0%
                  Max          +0.1%      0.0%     +1.2%     +0.0%     +0.0%
       Geometric Mean          +0.0%     -0.0%     +0.0%     -0.1%     -0.0%
      
      NoFib GC Results
      ================
      
      --------------------------------------------------------------------------------
              Program           Size    Allocs    Instrs     Reads    Writes
      --------------------------------------------------------------------------------
              circsim          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
          constraints          +0.0%      0.0%     -0.0%      0.0%     -0.0%
             fibheaps          +0.0%      0.0%      0.0%     -0.0%     -0.0%
               fulsom          +0.0%      0.0%      0.0%     -0.6%     -0.0%
             gc_bench          +0.0%      0.0%      0.0%      0.0%     -0.0%
                 hash          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
                 lcss          +0.0%      0.0%      0.0%     -0.0%      0.0%
            mutstore1          +0.0%      0.0%      0.0%     -0.0%     -0.0%
            mutstore2          +0.0%      0.0%     +0.0%     -0.0%     -0.0%
                power          +0.0%      0.0%     -0.0%      0.0%     -0.0%
           spellcheck          +0.0%      0.0%     -0.0%     -0.0%     -0.0%
      --------------------------------------------------------------------------------
                  Min          +0.0%      0.0%     -0.0%     -0.6%     -0.0%
                  Max          +0.0%      0.0%     +0.0%      0.0%      0.0%
       Geometric Mean          +0.0%     +0.0%     +0.0%     -0.1%     +0.0%
      
      Fixes #14373
      
      These performance regressions appear to be a fluke in CI. See the
      discussion in !1742 for details.
      
      Metric Increase:
          T6048
          T12234
          T12425
          Naperian
          T12150
          T5837
          T13035
      b911c532
    • Ben Gamari's avatar
      gitlab-ci: Add Debian 10 builds · 99e1816a
      Ben Gamari authored
      (cherry picked from commit c0c77bda)
      99e1816a
    • Ben Gamari's avatar
      13dab275
    • Ben Gamari's avatar
      gitlab-ci: Set LANG on CentOS 7 · 113bf129
      Ben Gamari authored
      It otherwise seems to default to ascii
      
      (cherry picked from commit b6f7e407)
      113bf129
    • Ben Gamari's avatar
      gitlab-ci: pxz is unavailable on CentOS 7 · 61820b24
      Ben Gamari authored
      Fall back to xz
      
      (cherry picked from commit 8565f808)
      61820b24
    • Ben Gamari's avatar
      Release notes fixes · 599a5916
      Ben Gamari authored
      599a5916
    • Ben Gamari's avatar
      gitlab-ci: Always build source tarball · a41bd7ac
      Ben Gamari authored
      a41bd7ac
    • Ben Gamari's avatar
      gitlab-ci: Add release-x86_64-linux-deb9 job · 1a70568b
      Ben Gamari authored
      1a70568b
  4. Nov 23, 2019
    • vdukhovni's avatar
      On FreeBSD 12 sys/sysctl.h requires sys/types.h · 273d802b
      vdukhovni authored
      Else build fails with:
      
          In file included from ExecutablePath.hsc:42:
          /usr/include/sys/sysctl.h:1062:25: error: unknown type name 'u_int'; did you mean 'int'?
           int sysctl(const int *, u_int, void *, size_t *, const void *, size_t);
      			     ^~~~~
      			     int
          compiling libraries/base/dist-install/build/System/Environment/ExecutablePath_hsc_make.c failed (exit code 1)
      
      Perhaps also also other FreeBSD releases, but additional include
      will no harm even if not needed.
      273d802b
  5. Nov 22, 2019
  6. Nov 21, 2019
  7. Nov 19, 2019
    • Ben Gamari's avatar
      nonmoving: Drop redundant write barrier on stack underflow · 098d5017
      Ben Gamari authored and Marge Bot's avatar Marge Bot committed
      Previously we would push stack-carried return values to the new stack on
      a stack overflow. While the precise reasoning for this barrier is
      unfortunately lost to history, in hindsight I suspect it was prompted by
      a missing barrier elsewhere (that has been since fixed).
      
      Moreover, there the redundant barrier is actively harmful: the stack may
      contain non-pointer values; blindly pushing these to the mark queue will
      result in a crash. This is precisely what happened in the `stack003`
      test. However, because of a (now fixed) deficiency in the test this
      crash did not trigger on amd64.
      098d5017
    • Ben Gamari's avatar
      testsuite: Increase width of stack003 test · a7571a74
      Ben Gamari authored and Marge Bot's avatar Marge Bot committed
      Previously the returned tuple seemed to fit in registers on amd64. This
      meant that non-moving collector bug would cause the test to fail on i386
      yet not amd64.
      a7571a74
    • Ben Gamari's avatar
      nonmoving: Fix handling on large object marking on 32-bit · eb7b233a
      Ben Gamari authored and Marge Bot's avatar Marge Bot committed
      Previously we would reset the pointer pointing to the object to be
      marked to the beginning of the block when marking a large object. This
      did no harm on 64-bit but on 32-bit it broke, e.g. `arr020`, since we
      align pinned ByteArray allocations such that the payload is 8
      byte-aligned. This means that the object might not begin at the
      beginning of the block.,
      eb7b233a
    • Ben Gamari's avatar
      nonmoving: Rework mark queue representation · 097f8072
      Ben Gamari authored and Marge Bot's avatar Marge Bot committed
      The previous representation needlessly limited the array length to
      16-bits on 32-bit platforms.
      097f8072
    • Ben Gamari's avatar
      nonmoving: Fix incorrect masking in mark queue type test · deed8e31
      Ben Gamari authored and Marge Bot's avatar Marge Bot committed
      We were using TAG_BITS instead of TAG_MASK. This happened to work on
      64-bit platforms where TAG_BITS==3 since we only use tag values 0 and
      3. However, this broken on 32-bit platforms where TAG_BITS==2.
      deed8e31
    • Ben Gamari's avatar
      nonmoving: Use correct info table pointer accessor · c819c0e4
      Ben Gamari authored and Marge Bot's avatar Marge Bot committed
      Previously we used INFO_PTR_TO_STRUCT instead of
      THUNK_INFO_PTR_TO_STRUCT when looking at a thunk. These two happen to be
      equivalent on 64-bit architectures due to alignment considerations
      however they are different on 32-bit platforms. This lead to #17487.
      
      To fix this we also employ a small optimization: there is only one thunk
      of type WHITEHOLE (namely stg_WHITEHOLE_info). Consequently, we can just
      use a plain pointer comparison instead of testing against info->type.
      c819c0e4
    • Ben Gamari's avatar
      rts: Add missing include of SymbolExtras.h · 0418c38d
      Ben Gamari authored and Marge Bot's avatar Marge Bot committed
      This broke the Windows build.
      0418c38d
    • Ben Gamari's avatar
      Properly account for libdw paths in make build system · 2b27cc16
      Ben Gamari authored and Marge Bot's avatar Marge Bot committed
      Should finally fix #17255.
      2b27cc16
    • Ömer Sinan Ağacan's avatar
      Packages.hs: use O(n*log(n)) ordNub instead of O(n*n) nub · cd40e12a
      Ömer Sinan Ağacan authored and Marge Bot's avatar Marge Bot committed
      As reported in #8173 in some environments package lists can get quite
      long, so we use more efficient ordNub instead of nub on package lists.
      cd40e12a
    • vdukhovni's avatar
      Enable USE_PTHREAD_FOR_ITIMER also on FreeBSD · ec8a463d
      vdukhovni authored and Marge Bot's avatar Marge Bot committed
      If using a pthread instead of a timer signal is more reliable, and
      has no known drawbacks, then FreeBSD is also capable of supporting
      this mode of operation (tested on FreeBSD 12 with GHC 8.8.1, but
      no reason why it would not also work on FreeBSD 11 or GHC 8.6).
      
      Proposed by Kevin Zhang in:
      
          https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=241849
      ec8a463d
    • Ben Gamari's avatar
      Give seq a more precise type and remove magic · 08d595c0
      Ben Gamari authored and Marge Bot's avatar Marge Bot committed
      `GHC.Prim.seq` previously had the rather plain type:
      
          seq :: forall a b. a -> b -> b
      
      However, it also had a special typing rule to applications
      where `b` is not of kind `Type`.
      
      Issue #17440 noted that levity polymorphism allows us to rather give
      it the more precise type:
      
          seq :: forall (r :: RuntimeRep) a (b :: TYPE r). a -> b -> b
      
      This allows us to remove the special typing rule that we previously
      required to allow applications on unlifted arguments. T9404 contains a
      non-Type application of `seq` which should verify that this works as
      expected.
      
      Closes #17440.
      08d595c0
Loading