[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 --- Comment #13 from uros at gcc dot gnu.org --- Author: uros Date: Wed Jun 18 20:01:37 2014 New Revision: 211803 URL: https://gcc.gnu.org/viewcvs?rev=211803root=gccview=rev Log: Backport from mainline 2014-06-06 Uros Bizjak ubiz...@gmail.com PR target/61423 * config/i386/i386.md (*floatunssimode2_i387_with_xmm): New define_insn_and_split pattern, merged from *floatunssimode2_1 and corresponding splitters. Zero extend general register or memory input operand to XMM temporary. Enable for TARGET_SSE2 and TARGET_INTER_UNIT_MOVES_TO_VEC only. (floatunssimode2): Update expander predicate. testsuite/ChangeLog: Backport from mainline 2014-06-13 Ilya Enkovich ilya.enkov...@intel.com PR rtl-optimization/61094 PR rtl-optimization/61446 * gcc.target/i386/pr61446.c : New. Backport from mainline 2014-06-06 Uros Bizjak ubiz...@gmail.com PR target/61423 * gcc.target/i386/pr61423.c: New test. Added: branches/gcc-4_8-branch/gcc/testsuite/gcc.target/i386/pr61423.c branches/gcc-4_8-branch/gcc/testsuite/gcc.target/i386/pr61446.c Modified: branches/gcc-4_8-branch/gcc/ChangeLog branches/gcc-4_8-branch/gcc/config/i386/i386.md branches/gcc-4_8-branch/gcc/testsuite/ChangeLog
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 Uroš Bizjak ubizjak at gmail dot com changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED Target Milestone|--- |4.8.4 --- Comment #14 from Uroš Bizjak ubizjak at gmail dot com --- Fixed in all release branches.
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 --- Comment #12 from uros at gcc dot gnu.org --- Author: uros Date: Tue Jun 17 05:00:52 2014 New Revision: 211723 URL: https://gcc.gnu.org/viewcvs?rev=211723root=gccview=rev Log: Backport from mainline 2014-06-06 Uros Bizjak ubiz...@gmail.com PR target/61423 * config/i386/i386.md (*floatunssimode2_i387_with_xmm): New define_insn_and_split pattern, merged from *floatunssimode2_1 and corresponding splitters. Zero extend general register or memory input operand to XMM temporary. Enable for TARGET_SSE2 and TARGET_INTER_UNIT_MOVES_TO_VEC only. (floatunssimode2): Update expander predicate. testsuite/ChangeLog: Backport from mainline 2014-06-06 Uros Bizjak ubiz...@gmail.com PR target/61423 * gcc.target/i386/pr61423.c: New test. Added: branches/gcc-4_9-branch/gcc/testsuite/gcc.target/i386/pr61423.c Modified: branches/gcc-4_9-branch/gcc/ChangeLog branches/gcc-4_9-branch/gcc/config/i386/i386.md branches/gcc-4_9-branch/gcc/testsuite/ChangeLog
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 Bug 61423 depends on bug 61446, which changed state. Bug 61446 Summary: [4.10 Regression]: ICE in copyprop_hardreg_forward_1, at regcprop.c with -O2 -m32 -march=corei7 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61446 What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 --- Comment #11 from Jeffrey A. Law law at redhat dot com --- I expect the REE bug to be fixed tomorrow AM (my time). It's a trivial issue.
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 Jakub Jelinek jakub at gcc dot gnu.org changed: What|Removed |Added CC||jakub at gcc dot gnu.org --- Comment #9 from Jakub Jelinek jakub at gcc dot gnu.org --- Shouldn't this patch be reverted until REE is fixed?
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 Jeffrey A. Law law at redhat dot com changed: What|Removed |Added CC||law at redhat dot com --- Comment #10 from Jeffrey A. Law law at redhat dot com --- Either way will cause a failure. This is #1 on my hit list after the regex problem that's affecting mysql.
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 --- Comment #8 from Uroš Bizjak ubizjak at gmail dot com --- (In reply to Uroš Bizjak from comment #5) Created attachment 32903 [details] Testcase that fails in REE pass This testcase fails with patched gcc in REE pass: This is now PR61446.
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423
Richard Biener rguenth at gcc dot gnu.org changed:
What|Removed |Added
Keywords||wrong-code
Target||i?86-*-*
Status|UNCONFIRMED |NEW
Last reconfirmed||2014-06-06
Component|c |target
Ever confirmed|0 |1
--- Comment #2 from Richard Biener rguenth at gcc dot gnu.org ---
Confirmed. Works with -mfpmath=sse. Initial RTL looks ok to me:
(insn 21 20 22 (set (reg:SI 99 [ stmp_sum_5.8 ])
(vec_select:SI (reg:V4SI 98 [ vect_sum_5.9 ])
(parallel [
(const_int 0 [0])
]))) -1
(nil))
(insn 22 21 23 (parallel [
(set (reg:DF 97 [ t ])
(unsigned_float:DF (reg:SI 99 [ stmp_sum_5.8 ])))
(clobber (mem/c:DI (plus:SI (reg/f:SI 78 virtual-stack-vars)
(const_int -8 [0xfff8])) [0 S8 A64]))
(clobber (scratch:SI))
]) t.c:18 -1
(nil))
(insn 23 22 24 (set (reg:DF 94 [ retval ])
(reg:DF 97 [ t ])) t.c:19 -1
(nil))
So it must be a bogus *floatunssidf2_1 pattern
(insn 22 21 28 4 (parallel [
(set (reg:DF 8 st [orig:97 t ] [97])
(unsigned_float:DF (reg:SI 21 xmm0 [orig:99 stmp_sum_5.8 ]
[99])))
(clobber (mem/c:DI (plus:SI (reg/f:SI 7 sp)
(const_int 8 [0x8])) [0 S8 A64]))
(clobber (scratch:SI))
]) t.c:18 211 {*floatunssidf2_1}
(nil))
split to
(insn 39 38 40 4 (set (mem/c:DI (plus:SI (reg/f:SI 7 sp)
(const_int 8 [0x8])) [0 S8 A64])
(reg:DI 21 xmm0 [orig:99 stmp_sum_5.8 ] [99])) t.c:18 89
{*movdi_internal}
(nil))
(insn 40 39 28 4 (set (reg:DF 8 st [orig:97 t ] [97])
(float:DF (mem/c:DI (plus:SI (reg/f:SI 7 sp)
(const_int 8 [0x8])) [0 S8 A64]))) t.c:18 206
{*floatdidf2_i387}
(nil))
note the use of a DImode memory but the missing zero-extend of xmm0:DI.
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 Uroš Bizjak ubizjak at gmail dot com changed: What|Removed |Added Status|NEW |ASSIGNED Assignee|unassigned at gcc dot gnu.org |ubizjak at gmail dot com --- Comment #3 from Uroš Bizjak ubizjak at gmail dot com --- Looking into it.
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 --- Comment #4 from Uroš Bizjak ubizjak at gmail dot com --- Created attachment 32901 -- https://gcc.gnu.org/bugzilla/attachment.cgi?id=32901action=edit Proposed patch This patch fies invalid code gen, but unfortunately uncovers a problem in REE pass and fails bootstrap when configured --with-arch=core-avx-i --with-cpu=core-avx-i.
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423
--- Comment #5 from Uroš Bizjak ubizjak at gmail dot com ---
Created attachment 32903
-- https://gcc.gnu.org/bugzilla/attachment.cgi?id=32903action=edit
Testcase that fails in REE pass
This testcase fails with patched gcc in REE pass:
cc1 -O2 -m32 -march=corei7 libgcc2.i
/home/uros/gcc-svn/trunk/libgcc/libgcc2.c: In function ‘__fixunssfdi’:
/home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1492:1: error: insn does not satisfy
its constraints:
}
^
(insn 54 11 47 2 (set (reg:DI 0 ax)
(reg:DI 21 xmm0)) /home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1444 89
{*movdi_internal}
(expr_list:REG_UNUSED (reg:DI 0 ax)
(nil)))
/home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1492:1: internal compiler error: in
copyprop_hardreg_forward_1, at regcprop.c:776
...
_split2 pass produces correct sequence:
(insn 46 11 47 2 (set (reg:DI 21 xmm0 [118])
(zero_extend:DI (reg/v:SI 0 ax [orig:85 hi ] [85])))
/home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1452 133 {*zero_extendsidi2}
(nil))
(insn 47 46 48 2 (set (mem/c:DI (reg/f:SI 7 sp) [0 S8 A64])
(reg:DI 21 xmm0 [118])) /home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1452
89 {*movdi_internal}
(nil))
(insn 48 47 13 2 (set (reg:DF 9 st(1) [orig:101 D.6895 ] [101])
(float:DF (mem/c:DI (reg/f:SI 7 sp) [0 S8 A64])))
/home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1452 206 {*floatdidf2_i387}
(nil))
and _ree pass converts this to:
(insn 11 45 54 2 (set (reg:DI 21 xmm0)
(zero_extend:DI (reg:SI 2 cx [99])))
/home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1444 133 {*zero_extendsidi2}
(nil))
(insn 54 11 47 2 (set (reg:DI 0 ax)
(reg:DI 21 xmm0)) /home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1444 -1
(nil))
(insn 47 54 48 2 (set (mem/c:DI (reg/f:SI 7 sp) [0 S8 A64])
(reg:DI 21 xmm0 [118])) /home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1452
89 {*movdi_internal}
(nil))
(insn 48 47 13 2 (set (reg:DF 9 st(1) [orig:101 D.6895 ] [101])
(float:DF (mem/c:DI (reg/f:SI 7 sp) [0 S8 A64])))
/home/uros/gcc-svn/trunk/libgcc/libgcc2.c:1452 206 {*floatdidf2_i387}
(nil))
Please note ivalid (insn 54).
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 Uroš Bizjak ubizjak at gmail dot com changed: What|Removed |Added CC||law at gcc dot gnu.org --- Comment #6 from Uroš Bizjak ubizjak at gmail dot com --- Jeff, can you please look at what goes wrong in REE pass?
[Bug target/61423] Incorrect conversion from unsigned int to floating point
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61423 --- Comment #7 from uros at gcc dot gnu.org --- Author: uros Date: Fri Jun 6 17:45:10 2014 New Revision: 211321 URL: http://gcc.gnu.org/viewcvs?rev=211321root=gccview=rev Log: PR target/61423 * config/i386/i386.md (*floatunssimode2_i387_with_xmm): New define_insn_and_split pattern, merged from *floatunssimode2_1 and corresponding splitters. Zero extend general register or memory input operand to XMM temporary. Enable for TARGET_SSE2 and TARGET_INTER_UNIT_MOVES_TO_VEC only. (floatunssimode2): Update expander predicate. testsuite/ChangeLog: PR target/61423 * gcc.target/i386/pr61423.c: New test. Added: trunk/gcc/testsuite/gcc.target/i386/pr61423.c Modified: trunk/gcc/ChangeLog trunk/gcc/config/i386/i386.md trunk/gcc/testsuite/ChangeLog
