See: https://ci.freebsd.org/job/FreeBSD-head-amd64-test/3603/testReport/junit/lib.msun/trig_test/main/
A commit references this bug: Author: ngie Date: Tue Jul 25 03:48:25 UTC 2017 New revision: 321454 URL: https://svnweb.freebsd.org/changeset/base/321454 Log: Mark :reduction as an expected failure It fails with clang 5.0+. PR: 220989 MFC after: 2 months MFC with: r321369 Reported by: Jenkins Changes: projects/runtime-coverage/lib/msun/tests/trig_test.c
A commit references this bug: Author: ngie Date: Tue Jul 25 03:56:42 UTC 2017 New revision: 321457 URL: https://svnweb.freebsd.org/changeset/base/321457 Log: Mark :reduction as an expected failure It fails with clang 5.0+. PR: 220989 MFC after: 2 months MFC with: r321369 Reported by: Jenkins Changes: head/lib/msun/tests/trig_test.c
I tried adding a bit of debug info, and with a "good" libm, the first result is: DBG: ld_pi_odd[0] = 1.18918869603738415962e+101, sinl() = 2.47267281071058784804e-22, fabsl() = 2.47267281071058784804e-22, LDBL_EPSILON = 1.08420217248550443401e-19 With a "bad" libm (compiled by clang 5.0.0), it becomes: DBG: ld_pi_odd[0] = 1.18918869603738415962e+101, sinl() = 6.81892800593498228635e-01, fabsl() = 6.81892800593498228635e-01, LDBL_EPSILON = 1.08420217248550443401e-19 E.g. the sinl() result is completely off. I did some bisection, and it seems to have regressed with this upstream commit: https://reviews.llvm.org/rL307529 ("This patch completely replaces the scheduling information for the SandyBridge architecture target by modifying the file X86SchedSandyBridge.td located under the X86 Target") I am not entirely sure why such a commit would influence the operation of e.g. sinl() on generic amd64, since it does not use any Sandy Bridge specific assembly or the like. Maybe it exposes some other bug in the X86 code generation for long doubles? Or maybe there is some sort of undefined behavior in our sinl() implementation that is now being over-optimized? The challenge here is to come up with a good minimized test case for upstream, so we can submit this as a concrete bug. Maybe it is easiest to just copy paste all of the sinl() implementation plus the relevant part of trig_test.c into one big .c file, and make it print and/or check the end result.
This is the difference in assembly output for s_sinl.pico, which is the only file that seems to cause the problem: --- obj-r307528/s_sinl-min.s 2017-07-29 22:00:53.371688000 +0200 +++ obj-r307529/s_sinl-min.s 2017-07-29 22:01:43.929787000 +0200 @@ -179,7 +179,6 @@ jbe .LBB0_9 .LBB0_14: # %if.end80.i fstp %st(0) - shll $16, %ebx movq %rax, -96(%rbp) movw $16406, -88(%rbp) # imm = 0x4016 fldt -96(%rbp) @@ -194,18 +193,19 @@ movsd %xmm0, -48(%rbp) movsd %xmm0, -152(%rbp) fsubl -152(%rbp) - flds .LCPI0_1(%rip) - fmul %st(0), %st(1) fnstcw -66(%rbp) + flds .LCPI0_1(%rip) movzwl -66(%rbp), %eax movw $3199, -66(%rbp) # imm = 0xC7F fldcw -66(%rbp) + fmul %st(0), %st(1) movw %ax, -66(%rbp) fxch %st(1) fistl -76(%rbp) fldcw -66(%rbp) xorps %xmm0, %xmm0 cvtsi2sdl -76(%rbp), %xmm0 + shll $16, %ebx movsd %xmm0, -40(%rbp) movsd %xmm0, -144(%rbp) fsubl -144(%rbp) @@ -354,5 +354,5 @@ .size pio2_3tlo, 8 - .ident "clang version 5.0.0 (trunk 307528)" + .ident "clang version 5.0.0 (trunk 307529)" .section ".note.GNU-stack","",@progbits
(In reply to Dimitry Andric from comment #4) > This is the difference in assembly output for s_sinl.pico In particular, the moving of shll $16, %ebx does not matter for the end result, but the following change does: > @@ -194,18 +193,19 @@ > movsd %xmm0, -48(%rbp) > movsd %xmm0, -152(%rbp) > fsubl -152(%rbp) > - flds .LCPI0_1(%rip) > - fmul %st(0), %st(1) > fnstcw -66(%rbp) > + flds .LCPI0_1(%rip) > movzwl -66(%rbp), %eax > movw $3199, -66(%rbp) # imm = 0xC7F > fldcw -66(%rbp) > + fmul %st(0), %st(1) > movw %ax, -66(%rbp) > fxch %st(1) > fistl -76(%rbp) > fldcw -66(%rbp) > xorps %xmm0, %xmm0 > cvtsi2sdl -76(%rbp), %xmm0 E.g. the fld and fmul are moved around, and that botches things up. This particular part corresponds to the following loop in lib/msun/ld80/e_rem_pio2l.h: for(i=0;i<2;i++) { tx[i] = (double)((int32_t)(z)); z = (z-tx[i])*two24; }
(In reply to Dimitry Andric from comment #5) The fmul is moved to after the fldcw, which is setting the FPU control word to 0xc7f, which may change the rounding and precision modes. (I guess the FPU control word is being set for the purpose of the cast to int32_t?) Is clang missing barriers around fldcw?
(In reply to Ryan Libby from comment #6) > (In reply to Dimitry Andric from comment #5) > > The fmul is moved to after the fldcw, which is setting the FPU control > word to 0xc7f, which may change the rounding and precision modes. (I > guess the FPU control word is being set for the purpose of the cast to > int32_t?) Is clang missing barriers around fldcw? At some point, upstream did an update of (part of) the instruction scheduler, and this messed up this particular test case. For the full details, see: https://bugs.llvm.org/show_bug.cgi?id=34080 Sorry that I didn't link to that upstream bug report earlier here. The jury is still out on what to do with it for the 5.0.0 release.
A commit references this bug: Author: dim Date: Thu Aug 24 20:19:29 UTC 2017 New revision: 322855 URL: https://svnweb.freebsd.org/changeset/base/322855 Log: Upgrade our copies of clang, llvm, lldb and compiler-rt to r311606 from the upstream release_50 branch. As of this version, lib/msun's trig test should also work correctly again (see bug 220989 for more information). PR: 220989 MFC after: 2 months X-MFC-with: r321369 Changes: _U head/contrib/compiler-rt/ head/contrib/compiler-rt/lib/builtins/arm/aeabi_dcmp.S head/contrib/compiler-rt/lib/builtins/arm/aeabi_fcmp.S head/contrib/compiler-rt/lib/esan/esan_sideline_linux.cpp head/contrib/compiler-rt/lib/profile/InstrProfilingNameVar.c _U head/contrib/libc++/ _U head/contrib/llvm/ head/contrib/llvm/include/llvm/CodeGen/SelectionDAGNodes.h head/contrib/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h head/contrib/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp head/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp head/contrib/llvm/lib/IR/AutoUpgrade.cpp head/contrib/llvm/lib/Object/COFFModuleDefinition.cpp head/contrib/llvm/lib/Target/ARM/ARMISelLowering.cpp head/contrib/llvm/lib/Target/X86/X86ISelLowering.cpp head/contrib/llvm/lib/Target/X86/X86InstrAVX512.td head/contrib/llvm/lib/Target/X86/X86SchedSandyBridge.td head/contrib/llvm/lib/ToolDrivers/llvm-dlltool/DlltoolDriver.cpp head/contrib/llvm/lib/ToolDrivers/llvm-dlltool/Options.td head/contrib/llvm/lib/Transforms/Scalar/LowerAtomic.cpp head/contrib/llvm/lib/Transforms/Scalar/Reassociate.cpp head/contrib/llvm/lib/Transforms/Utils/CloneFunction.cpp _U head/contrib/llvm/tools/clang/ head/contrib/llvm/tools/clang/include/clang/AST/DeclCXX.h head/contrib/llvm/tools/clang/include/clang/Lex/Preprocessor.h head/contrib/llvm/tools/clang/lib/AST/ASTImporter.cpp head/contrib/llvm/tools/clang/lib/AST/DeclCXX.cpp head/contrib/llvm/tools/clang/lib/CodeGen/CGCXXABI.cpp head/contrib/llvm/tools/clang/lib/CodeGen/ItaniumCXXABI.cpp head/contrib/llvm/tools/clang/lib/CodeGen/MicrosoftCXXABI.cpp head/contrib/llvm/tools/clang/lib/Driver/ToolChains/Darwin.cpp head/contrib/llvm/tools/clang/lib/Driver/ToolChains/MSVC.cpp head/contrib/llvm/tools/clang/lib/Driver/ToolChains/MSVC.h head/contrib/llvm/tools/clang/lib/Format/WhitespaceManager.cpp head/contrib/llvm/tools/clang/lib/Headers/unwind.h head/contrib/llvm/tools/clang/lib/Lex/PPLexerChange.cpp head/contrib/llvm/tools/clang/lib/Lex/Preprocessor.cpp head/contrib/llvm/tools/clang/lib/Parse/Parser.cpp head/contrib/llvm/tools/clang/lib/Sema/SemaDeclCXX.cpp head/contrib/llvm/tools/clang/lib/Sema/SemaObjCProperty.cpp head/contrib/llvm/tools/clang/lib/Serialization/ASTReaderDecl.cpp head/contrib/llvm/tools/clang/lib/Serialization/ASTWriter.cpp head/contrib/llvm/tools/clang/lib/StaticAnalyzer/Core/RegionStore.cpp _U head/contrib/llvm/tools/lld/ _U head/contrib/llvm/tools/lldb/ head/contrib/llvm/tools/lldb/source/Plugins/Process/Utility/RegisterContextLinux_i386.cpp head/contrib/llvm/tools/lldb/source/Plugins/Process/Utility/RegisterContext_x86.h head/lib/clang/include/clang/Basic/Version.inc head/lib/clang/include/lld/Config/Version.inc head/lib/clang/include/llvm/Support/VCSRevision.h
Ngie, if you could verify that reverting r321457 now makes trig_test work OK, can we close this bug?
A commit references this bug: Author: ngie Date: Sat Aug 26 06:44:50 UTC 2017 New revision: 322921 URL: https://svnweb.freebsd.org/changeset/base/322921 Log: Revert r321457 It doesn't fail after ^/head@r322855 (the releng_50 clang merge). PR: 220989 MFC after: 2 months MFC with: r321369, r322855 Changes: head/lib/msun/tests/trig_test.c
(In reply to commit-hook from comment #10) I confirmed this was the case per ci.freebsd.org: https://ci.freebsd.org/job/FreeBSD-head-amd64-test/4046/testReport/junit/lib.msun/trig_test/reduction/ .
A commit references this bug: Author: dim Date: Tue Sep 26 09:02:00 UTC 2017 New revision: 324006 URL: https://svnweb.freebsd.org/changeset/base/324006 Log: Synchronize most of libm with head as of r323004. This excludes a few arch-specific updates for powerpcspe, mips and riscv, for which support has not been merged yet. Bump __FreeBSD_version for the addition of cacoshl, cacosl, casinhl, casinl, catanl, catanhl, sincos, sincosf, and sincosl. MFC r305382 (by bde): Add asm versions of fmod(), fmodf() and fmodl() on amd64. Add asm versions of fmodf() amd fmodl() on i387. fmod is similar to remainder, and the C versions are 3 to 9 times slower than the asm versions on x86 for both, but we had the strange mixture of all 6 variants of remainder in asm and only 1 of 6 variants of fmod in asm. MFC r305384 (by bde): Disconnect the "optimized" asm variants of cos(), sin() and tan() from the build on i386. Leave them in the source tree for regression tests. The asm functions were always much less accurate (by a factor of more than 10**18 in the worst case). They were faster on old CPUs. But with each new generation of CPUs they get relatively slower. The double precision C version's average advantage is about a factor of 2 on Haswell. The asm functions were already intentionally avoided in float and long double precision on i386 and in all precisions on amd64. Float precision and amd64 give larger advantages to the C version. The long double precision C code and compilers' understanding of long double precision are not so good, so the i387 is still slightly faster for long double precision, except for the unimportant subcase of huge args where the sub-optimal C code now somehow beats the i387 by about a factor of 2. MFC r305385 (by bde): Oops, the previous i386 version of e_fmodf.S and e_fmodl.S was actually the amd64 version. MFC r306409 (by emaste): libm: fix some unused variable (rcsid) and dangling else warnings s_{fabs,fmax,logb,scalb}{,f,l}.c may be built elsewhere with a higher WARNS setting. Reviewed by: ed Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D8061 MFC r306410 (by emaste): libm: simplify i387 subdir logic with make's :S substitution MFC r306527 (by emaste): libm: remove unused variables for LDBL_MANT_DIG != 113 Sponsored by: The FreeBSD Foundation MFC r306709 (by emaste): libm: remove unused variables Sponsored by: The FreeBSD Foundation MFC r307066 (by br): Don't use fmaxl/fminl on platforms with no long double support, use fmax/fmin instead. This fixes fmaxmin test failure on MIPS64. Reviewed by: emaste Sponsored by: DARPA, AFRL Sponsored by: HEIF5 Differential Revision: https://reviews.freebsd.org/D8216 MFC r308172 (by emaste): libm: add braces around initialization of subobjects This cleans up a warning when building libm at higher WARNS levels and makes the intent more clear. By the C standard the values are assigned to subobject members in order so this change introduces no functional change. (6.7.9 20) Reviewed by: kib Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D8333 MFC r313761 (by mmokhi): Add casinl() cacosl() catanl() casinhl() cacoshl() catanhl() APIs to msun to improve C11 conformance. PR: 216850 216851 216852 216856 216857 216858 Submitted by: mmokhi Reported by: sgk@troutmask.apl.washington.edu Reviewed by: bde, mat, theraven Approved by: bde (src committer), mat (mentor) Differential Revision: https://reviews.freebsd.org/D9491 MFC r313863 (by mmokhi): Fix building of r313761 on platforms that `long double` is alias of `double` (MIPS, etc) PR: 216850 216851 216852 216856 216857 216858 Reported by: emsate Reviewed by: bde emaste hselasky Approved by: bde emaste hselasky Differential Revision: https://reviews.freebsd.org/D9491 MFC r313864 (by mmokhi): Add documentations related to new APIs of r313761 PR: 216850 216851 216852 216856 216857 216858 Submitted by: sgk@troutmask.apl.washington.edu Reported by: sgk@troutmask.apl.washington.edu Reviewed by: bde emaste hselasky Approved by: bde emaste hselasky Differential Revision: https://reviews.freebsd.org/D9491 MFC r314950 (by ngie): Don't expect :test_large_inputs to fail with i386 anymore Recent changes (maybe a side-effect of the ATF-ification in r314649) invalidate the failure expectation. PR: 205446 Sponsored by: Dell EMC Isilon MFC r317349 (by pfg): msun: Remove trailing space in Sunsoft copyright statement. Submittedby: kargl MFC r319047 (by mmel): Implement sincos, sincosf, and sincosl. The primary benefit of these functions is that argument reduction is done once instead of twice in independent calls to sin() and cos(). * lib/msun/Makefile: . Add s_sincos[fl].c to the build. . Add sincos.3 documentation. . Add appropriate MLINKS. * lib/msun/Symbol.map: . Expose sincos[fl] symbols in dynamic libm.so. * lib/msun/man/sincos.3: . Documentation for sincos[fl]. * lib/msun/src/k_sincos.h: . Kernel for sincos() function. This merges the individual kernels for sin() and cos(). The merger offered an opportunity to re-arrange the individual kernels for better performance. * lib/msun/src/k_sincosf.h: . Kernel for sincosf() function. This merges the individual kernels for sinf() and cosf(). The merger offered an opportunity to re-arrange the individual kernels for better performance. * lib/msun/src/k_sincosl.h: . Kernel for sincosl() function. This merges the individual kernels for sinl() and cosl(). The merger offered an opportunity to re-arrange the individual kernels for better performance. * lib/msun/src/math.h: . Add prototytpes for sincos[fl](). * lib/msun/src/math_private.h: . Add RETURNV macros. This is needed to reset fpsetprec on I386 hardware for a function with type void. * lib/msun/src/s_sincos.c: . Implementation of sincos() where sin() and cos() were merged into one routine and possibly re-arranged for better performance. * lib/msun/src/s_sincosf.c: . Implementation of sincosf() where sinf() and cosf() were merged into one routine and possibly re-arranged for better performance. * lib/msun/src/s_sincosl.c: . Implementation of sincosl() where sinl() and cosl() were merged into one routine and possibly re-arranged for better performance. PR: 215977, 218300 Submitted by: Steven G. Kargl <sgk@troutmask.apl.washington.edu> Differential Revision: https://reviews.freebsd.org/D10765 MFC r321457 (by ngie): Mark :reduction as an expected failure It fails with clang 5.0+. PR: 220989 Reported by: Jenkins MFC r322418 (by rlibby): lib/msun: avoid referring to broken LDBL_MAX LDBL_MAX is broken on i386: https://lists.freebsd.org/pipermail/freebsd-numerics/2012-September/000288.html Gcc has produced +Infinity for LDBL_MAX on i386 and amd64 with -m32 for some time, and newer versions of gcc are now warning that the "floating constant exceeds range of 'long double'". Avoid this by referring to half the value of LDBL_MAX instead. Reviewed by: bde Approved by: markj (mentor) Sponsored by: Dell EMC Isilon MFC r322435 (by rlibby): Revert r322418, LDBL_MAX_EXP unsuitable for macro pasting on some arches Either need a different way to spell HALF_LDBL_MAX, or a different way to spell LDBL_MAX_EXP, or a different approach. Reported by: ian MFC r322921 (by ngie): Revert r321457 It doesn't fail after ^/head@r322855 (the releng_50 clang merge). PR: 220989 Changes: _U stable/11/ stable/11/lib/msun/Makefile stable/11/lib/msun/Symbol.map stable/11/lib/msun/amd64/Makefile.inc stable/11/lib/msun/amd64/e_fmod.S stable/11/lib/msun/amd64/e_fmodf.S stable/11/lib/msun/amd64/e_fmodl.S stable/11/lib/msun/i387/Makefile.inc stable/11/lib/msun/i387/e_fmodf.S stable/11/lib/msun/i387/e_fmodl.S stable/11/lib/msun/ld80/e_lgammal_r.c stable/11/lib/msun/ld80/k_expl.h stable/11/lib/msun/ld80/s_logl.c stable/11/lib/msun/man/cacos.3 stable/11/lib/msun/man/sincos.3 stable/11/lib/msun/src/catrig.c stable/11/lib/msun/src/catrigl.c stable/11/lib/msun/src/e_asin.c stable/11/lib/msun/src/e_coshl.c stable/11/lib/msun/src/e_lgammaf_r.c stable/11/lib/msun/src/e_sinhl.c stable/11/lib/msun/src/k_sincos.h stable/11/lib/msun/src/k_sincosf.h stable/11/lib/msun/src/k_sincosl.h stable/11/lib/msun/src/math.h stable/11/lib/msun/src/math_private.h stable/11/lib/msun/src/s_fabs.c stable/11/lib/msun/src/s_fmax.c stable/11/lib/msun/src/s_fmin.c stable/11/lib/msun/src/s_logbl.c stable/11/lib/msun/src/s_scalbn.c stable/11/lib/msun/src/s_scalbnf.c stable/11/lib/msun/src/s_scalbnl.c stable/11/lib/msun/src/s_sincos.c stable/11/lib/msun/src/s_sincosf.c stable/11/lib/msun/src/s_sincosl.c stable/11/lib/msun/src/s_tanhl.c stable/11/lib/msun/tests/ctrig_test.c stable/11/sys/sys/param.h