Page MenuHomePhabricator

Abe (Abe Skolnik)
User

Projects

User does not belong to any projects.

User Details

User Since
Sep 12 2016, 4:05 PM (148 w, 5 d)

Recent Activity

Dec 14 2016

Abe abandoned D25575: test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs.
Dec 14 2016, 12:00 PM
Abe abandoned D25635: detect FP contraction at test-time, disable certain subtests if detected.
Dec 14 2016, 11:59 AM

Dec 13 2016

Abe updated the diff for D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.

fixed a bug Gerolf caught

Dec 13 2016, 1:17 PM
Abe updated the diff for D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.

Trivial edit; NFC WRT prev. diff in same patch.

Dec 13 2016, 11:40 AM
Abe updated the diff for D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.

Changed the interface for the new "MachineInstr::print(...)" overload as Gerolf requested.

Dec 13 2016, 11:33 AM

Dec 12 2016

Abe updated the diff for D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.

Minor changes in source code formatting, editing according to "clang-format" output [ran with "-style=LLVM"].

Dec 12 2016, 12:35 PM
Abe updated the diff for D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.

Now using "= nullptr" instead of "= 0" in one spot.

Dec 12 2016, 11:12 AM
Abe updated the diff for D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.

Incorporated Sebastian`s further reduction of the new test case.

Dec 12 2016, 10:47 AM
Abe updated the diff for D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.

Implemented some recommendations/suggestions from Sebastian.

Dec 12 2016, 10:23 AM

Dec 9 2016

Abe updated the diff for D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.

Integrated some good suggestions from Gerolf, thereby making the fix less kludgy.

Dec 9 2016, 6:29 PM
Abe added reviewers for D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled: dexonsmith, lattner, Gerolf, echristo, bkramer.
Dec 9 2016, 5:14 PM
Abe updated D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.
Dec 9 2016, 5:09 PM
Abe updated D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.
Dec 9 2016, 5:05 PM
Abe retitled D27645: debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled from to debug output incorrect ["UNKNOWN"] when a MachineInstr is generated during the MachineCombiner pass and the relevant debug output is enabled.
Dec 9 2016, 5:05 PM

Dec 7 2016

Abe updated D27552: machine-instruction combiner uses too-simple algorithm to compute cost of one of the two alternatives, throws away combination results too often.
Dec 7 2016, 3:38 PM
Abe retitled D27552: machine-instruction combiner uses too-simple algorithm to compute cost of one of the two alternatives, throws away combination results too often from to machine-instruction combiner uses too-simple algorithm to compute cost of one of the two alternatives, throws away combination results too often.
Dec 7 2016, 3:35 PM

Oct 21 2016

Abe updated D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs.
Oct 21 2016, 8:44 AM
Abe updated the diff for D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs.

minor help-text edit

Oct 21 2016, 8:40 AM

Oct 20 2016

Abe updated the diff for D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs.

Renamed variables.

Oct 20 2016, 3:09 PM
Abe retitled D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs from test-suite: FP: by default, assume either contraction, fast-math, or both are on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs to test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs.
Oct 20 2016, 3:05 PM
Abe updated the diff for D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs.

Applied several requests/suggestions from MB & RG.

Oct 20 2016, 2:55 PM

Oct 19 2016

Abe updated the diff for D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs.

minor edit re "minisat".

Oct 19 2016, 3:47 PM
Abe retitled D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs from test-suite: FP: by default, assume either contraction, fast-math, or both are on unless the user specifies otherwise, disable tests accordingly to test-suite: FP: by default, assume either contraction, fast-math, or both are on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs.
Oct 19 2016, 11:57 AM
Abe updated the diff for D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs.

added the bulk of it

Oct 19 2016, 11:57 AM
Abe abandoned D25056: Fixed "MultiSource/Benchmarks/MiBench/telecomm-FFT" test to use an FP tolerance and check the actual output instead of hashing the output and demanding exact bitwise equality [modulo hash collisions]..
Oct 19 2016, 11:56 AM

Oct 18 2016

Abe updated D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs.
Oct 18 2016, 1:51 PM
Abe retitled D25749: test-suite: by default, assume that FP contraction is on unless the user specifies otherwise, disable tests accordingly; add support for per-ISA reference outputs from to test-suite: FP: by default, assume either contraction, fast-math, or both are on unless the user specifies otherwise, disable tests accordingly.
Oct 18 2016, 1:48 PM
Abe retitled D25575: test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs from test-suite: try to force many FP-dependent tests to compile with unaggressive FP optimization or not at all, add support for per-ISA reference outputs to test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs.
Oct 18 2016, 1:09 PM

Oct 17 2016

Abe updated the diff for D25575: test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs.

rebased to:

Oct 17 2016, 2:59 PM
Abe updated the diff for D25635: detect FP contraction at test-time, disable certain subtests if detected.

implemented 2 good suggestions from MB ["u64_to_double" instead of union, exit code]

Oct 17 2016, 9:13 AM
Abe updated D25635: detect FP contraction at test-time, disable certain subtests if detected.
Oct 17 2016, 9:09 AM

Oct 14 2016

Abe updated D25635: detect FP contraction at test-time, disable certain subtests if detected.
Oct 14 2016, 4:48 PM
Abe updated the diff for D25635: detect FP contraction at test-time, disable certain subtests if detected.

added support for GCC [4.4 or higher required AFAIK]

Oct 14 2016, 4:47 PM
Abe updated D25635: detect FP contraction at test-time, disable certain subtests if detected.
Oct 14 2016, 4:43 PM
Abe retitled D25635: detect FP contraction at test-time, disable certain subtests if detected from to detect FP contraction at test-time, disable certain subtests if detected.
Oct 14 2016, 4:31 PM

Oct 13 2016

Abe updated D25575: test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs.
Oct 13 2016, 4:56 PM
Abe updated the diff for D25575: test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs.

removed attempts to force no fast-math, made it detect fast-math and disable tests accordingly instead

Oct 13 2016, 4:53 PM
Abe abandoned D25351: added "-ffp-contract=off -fno-fast-math" to lots of tests, made 1 test use its real output for reference, deleted 1 test.
Oct 13 2016, 3:57 PM
Abe added reviewers for D25575: test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs: rengolin, MatzeB.
Oct 13 2016, 1:30 PM
Abe updated subscribers of D25575: test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs.
Oct 13 2016, 1:29 PM
Abe added a reviewer for D25575: test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs: sebpop.
Oct 13 2016, 1:28 PM
Abe retitled D25575: test-suite: disable tests that fail when FP optimization is in effect, add support for per-ISA reference outputs from to test-suite: try to force many FP-dependent tests to compile with unaggressive FP optimization, add support for per-ISA reference outputs.
Oct 13 2016, 1:27 PM

Oct 6 2016

Abe retitled D25351: added "-ffp-contract=off -fno-fast-math" to lots of tests, made 1 test use its real output for reference, deleted 1 test from to added "-ffp-contract=off -fno-fast-math" to lots of tests, made 1 test use its real output for reference, deleted 1 test.
Oct 6 2016, 4:43 PM
Abe abandoned D25254: test-suite: Change extension used for reference outputs by Makefile-based harness so we can start improving how the CMake-based harness works without breaking the old system or the buildbots.
Oct 6 2016, 8:32 AM
Abe abandoned D25194: Change some of the Makefiles of the Makefile-based test harness to use "-ffp-contract=off" so the build-bots will be able to tolerate more-aggressive FP optimization by default.
Oct 6 2016, 8:31 AM

Oct 4 2016

Abe updated D25254: test-suite: Change extension used for reference outputs by Makefile-based harness so we can start improving how the CMake-based harness works without breaking the old system or the buildbots.
Oct 4 2016, 2:49 PM
Abe retitled D25254: test-suite: Change extension used for reference outputs by Makefile-based harness so we can start improving how the CMake-based harness works without breaking the old system or the buildbots from to test-suite: Change extension used for reference outputs by Makefile-based harness so we can start improving how the CMake-based harness works without breaking the old system or the buildbots.
Oct 4 2016, 2:47 PM

Oct 3 2016

Abe updated the diff for D25194: Change some of the Makefiles of the Makefile-based test harness to use "-ffp-contract=off" so the build-bots will be able to tolerate more-aggressive FP optimization by default.

Removed some CXXFLAGS lines [they were unneeded acc. to Sebastian Pop].

Oct 3 2016, 3:17 PM
Abe retitled D25194: Change some of the Makefiles of the Makefile-based test harness to use "-ffp-contract=off" so the build-bots will be able to tolerate more-aggressive FP optimization by default from to Change some of the Makefiles of the Makefile-based test harness to use "-ffp-contract=off" so the build-bots will be able to tolerate more-aggressive FP optimization by default.
Oct 3 2016, 9:25 AM

Sep 28 2016

Abe retitled D25056: Fixed "MultiSource/Benchmarks/MiBench/telecomm-FFT" test to use an FP tolerance and check the actual output instead of hashing the output and demanding exact bitwise equality [modulo hash collisions]. from to Fixed "MultiSource/Benchmarks/MiBench/telecomm-FFT" test to use an FP tolerance and check the actual output instead of hashing the output and demanding exact bitwise equality [modulo hash collisions]..
Sep 28 2016, 5:17 PM

Sep 22 2016

Abe updated the diff for D24481: make “#pragma STDC FP_CONTRACT” on by default.

Minor edits for style-guidelines conformance.

Sep 22 2016, 9:42 AM

Sep 21 2016

Abe updated the diff for D24481: make “#pragma STDC FP_CONTRACT” on by default.

Combined "fp-contract-pragma___on-by-default___-O0___aarch64-backend.c" and "fp-contract-pragma___on-by-default___-O1...3___aarch64-backend.c" into a single file ["fp-contract-pragma___on-by-default.c"].

Sep 21 2016, 2:02 PM

Sep 20 2016

Abe updated the diff for D24481: make “#pragma STDC FP_CONTRACT” on by default.

Collapsed 4 test-case files that didn`t really need to be separate into 1 file, as suggested by Yaxun Liu.

Sep 20 2016, 11:28 AM
Abe added a comment to D24481: make “#pragma STDC FP_CONTRACT” on by default.

Is it possible to merge cfe/trunk/test/CodeGen/fp-contract-pragma___on-by-default___-O[0-3]___aarch64-backend.c as one and remove cfe/trunk/test/CodeGen/fp-contract-pragma___on-by-default___-O1...3___aarch64-backend.h? They look the same except the RUN commands. Why not using one file with multiple RUn commands?

Sep 20 2016, 10:43 AM
Abe updated the diff for D24481: make “#pragma STDC FP_CONTRACT” on by default.

Removed some comments that I felt were good for clarity but at least 2 people disagreed.

Sep 20 2016, 10:27 AM

Sep 16 2016

Abe added a reviewer for D24481: make “#pragma STDC FP_CONTRACT” on by default: yaxunl.
Sep 16 2016, 12:25 PM

Sep 14 2016

Abe removed a reviewer for D24481: make “#pragma STDC FP_CONTRACT” on by default: llvm-commits.
Sep 14 2016, 10:48 AM

Sep 13 2016

Abe updated the diff for D24481: make “#pragma STDC FP_CONTRACT” on by default.

Patch tweaking: renamed path prefixes from "clang/" to "cfe/trunk/", added _lots_ more context. Now the context shows up in Phabricator rather than "Context not available."

Sep 13 2016, 11:05 AM
Abe updated the diff for D24481: make “#pragma STDC FP_CONTRACT” on by default.

Fixed "fp_contract_7" of "clang/test/CodeGen/fp-contract-pragma.cpp" so it actually _does_ have multiple _not-provably-dead_ uses of the result of the multiplication [this nuance might be important in the future if a DCE pass will occur before the IR is tested].

Sep 13 2016, 10:21 AM
Abe updated the diff for D24481: make “#pragma STDC FP_CONTRACT” on by default.

Added improved testing [both as improvements to existing tests and 1 entirely-new test] {copied from / in imitation of} https://reviews.llvm.org/D14200.

Sep 13 2016, 10:03 AM

Sep 12 2016

Abe retitled D24481: make “#pragma STDC FP_CONTRACT” on by default from to make “#pragma STDC FP_CONTRACT” on by default.
Sep 12 2016, 4:35 PM