Page MenuHomePhabricator

jketema (Jeroen Ketema)
User

Projects

User does not belong to any projects.

User Details

User Since
May 22 2015, 12:13 PM (371 w, 4 d)

Recent Activity

Oct 9 2017

jketema committed rL315235: Implement mem_fence on ptx.
Implement mem_fence on ptx
Oct 9 2017, 12:45 PM
jketema committed rL315228: Make ptx barrier work irrespective of the cl_mem_fence_flags.
Make ptx barrier work irrespective of the cl_mem_fence_flags
Oct 9 2017, 11:38 AM

Oct 7 2017

jketema committed rL315147: Delete empy directory.
Delete empy directory
Oct 7 2017, 2:12 AM

Oct 5 2017

jketema committed rL314998: Add vload_half helpers for ptx.
Add vload_half helpers for ptx
Oct 5 2017, 11:19 AM

Oct 4 2017

jketema committed rL314925: Add vstore_half helpers for ptx.
Add vstore_half helpers for ptx
Oct 4 2017, 12:09 PM

Oct 1 2017

jketema committed rL314634: Let get_work_dim take exactly 0 arguments.
Let get_work_dim take exactly 0 arguments
Oct 1 2017, 1:13 PM
jketema committed rL314633: Do no circularly define NULL.
Do no circularly define NULL
Oct 1 2017, 1:12 PM

Feb 12 2017

jketema committed rL294916: Move BufferPtr into the block where it it being used.
Move BufferPtr into the block where it it being used
Feb 12 2017, 1:45 PM
jketema committed rL294915: Add the correct prefixes to the cl_khr_fp64 pragma.
Add the correct prefixes to the cl_khr_fp64 pragma
Feb 12 2017, 1:43 PM

Jun 13 2016

jketema committed rL272551: [builtins] Make SOFT_FP versions of floatdidf and floatundidf compile.
[builtins] Make SOFT_FP versions of floatdidf and floatundidf compile
Jun 13 2016, 8:31 AM
jketema committed rL272550: [builtins] replace tabs by spaces and remove whitespace at end of line NFC.
[builtins] replace tabs by spaces and remove whitespace at end of line NFC
Jun 13 2016, 8:27 AM

May 10 2016

jketema committed rL269040: [OCaml] Update core test and re-enable testing.
[OCaml] Update core test and re-enable testing
May 10 2016, 4:25 AM
jketema closed D19828: [OCaml] Update core test and re-enable testing by committing rL269040: [OCaml] Update core test and re-enable testing.
May 10 2016, 4:25 AM
jketema added inline comments to D19828: [OCaml] Update core test and re-enable testing.
May 10 2016, 4:15 AM
jketema updated the diff for D19828: [OCaml] Update core test and re-enable testing.
May 10 2016, 4:14 AM

May 2 2016

jketema retitled D19828: [OCaml] Update core test and re-enable testing from to [OCaml] Update core test and re-enable testing.
May 2 2016, 1:25 PM

Apr 23 2016

jketema abandoned D19037: [llvm-c] Add the ability to create debug locations.
Apr 23 2016, 1:58 AM

Apr 13 2016

jketema added a comment to D19037: [llvm-c] Add the ability to create debug locations.

How is the scope created using that API ?

Apr 13 2016, 3:48 PM
jketema added a comment to D19037: [llvm-c] Add the ability to create debug locations.

@echristo Indeed this doesn't quite work: the generated bitcode is parsable, but invalid, because a DIScope cannot be generated. I was planning to expose DIBuilder as a follow up patch, which should resolve this, but given your comment, you think this is not the right time to do this?

Apr 13 2016, 2:58 PM
jketema added inline comments to D19037: [llvm-c] Add the ability to create debug locations.
Apr 13 2016, 2:40 PM
jketema updated D19037: [llvm-c] Add the ability to create debug locations.
Apr 13 2016, 2:39 PM

Apr 12 2016

jketema committed rL266149: Add space between words in verify-scev-maps option help message.
Add space between words in verify-scev-maps option help message
Apr 12 2016, 4:27 PM
jketema retitled D19037: [llvm-c] Add the ability to create debug locations from to [llvm-c] Add the ability to create debug locations.
Apr 12 2016, 3:22 PM

Apr 10 2016

jketema committed rL265897: [OCaml] Expose the LLVM diagnostic handler.
[OCaml] Expose the LLVM diagnostic handler
Apr 10 2016, 7:01 AM
jketema closed D18891: [OCaml] Expose the LLVM diagnostic handler by committing rL265897: [OCaml] Expose the LLVM diagnostic handler.
Apr 10 2016, 7:01 AM
jketema added a comment to D18891: [OCaml] Expose the LLVM diagnostic handler.

Thanks for reviewing. Note that this doesn't re-enable the binding tests, as one test is still broken after these fixes: the core.ml test tries to build old-style debug information, and the new style is not even exposed in llvm-c. Maybe that part of core.ml should be disabled (or even removed?).

Apr 10 2016, 6:30 AM
jketema updated the diff for D18891: [OCaml] Expose the LLVM diagnostic handler.

Addressed new comments.

Apr 10 2016, 6:20 AM
jketema updated the diff for D18891: [OCaml] Expose the LLVM diagnostic handler.

Updated patch based on @whitequark 's feedback. I've also added a separate test to test the diagnostic handler functionality.

Apr 10 2016, 5:35 AM

Apr 8 2016

jketema added a comment to D18891: [OCaml] Expose the LLVM diagnostic handler.

Thanks for the feedback so far!

Apr 8 2016, 2:59 AM
jketema retitled D18891: [OCaml] Expose the LLVM diagnostic handler from to [OCaml] Expose the LLVM diagnostic handler.
Apr 8 2016, 2:34 AM
jketema committed rL265773: [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context}.
[llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context}
Apr 8 2016, 2:24 AM
jketema closed D18820: [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context} by committing rL265773: [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context}.
Apr 8 2016, 2:24 AM

Apr 7 2016

jketema updated the diff for D18820: [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context}.

Addressed the little nitpicks.

Apr 7 2016, 4:33 PM
jketema updated the diff for D18820: [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context}.

Remove some unnecessary duplication introduced in the previous version of this patch.

Apr 7 2016, 3:47 AM
jketema updated the diff for D18820: [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context}.

I've split out the diagnostic testing and improved it a bit as suggested. Let me know what you think.

Apr 7 2016, 3:39 AM
jketema committed rL265648: [CMake] Check for sys/types.h in config-ix.cmake.
[CMake] Check for sys/types.h in config-ix.cmake
Apr 7 2016, 1:41 AM
jketema closed D18825: [CMake] Check for sys/types.h in config-ix.cmake by committing rL265648: [CMake] Check for sys/types.h in config-ix.cmake.
Apr 7 2016, 1:41 AM

Apr 6 2016

jketema updated the diff for D18820: [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context}.

Added test for diagnostic handler

Apr 6 2016, 4:08 PM
jketema added a comment to D18820: [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context}.

I was wondering what a suitable place for a test was. Thanks for the pointer, I'll update. Once this is in I should have a first version of a patch for the OCaml bindings exposing the diagnostic handler.

Apr 6 2016, 3:35 PM
jketema retitled D18825: [CMake] Check for sys/types.h in config-ix.cmake from to [CMake] Check for sys/types.h in config-ix.cmake.
Apr 6 2016, 7:42 AM
jketema retitled D18820: [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context} from to [llvm-c] Expose LLVMContextGetDiagnostic{Handler,Context}.
Apr 6 2016, 3:25 AM

Apr 1 2016

jketema added a comment to D18675: [OCaml] Use LLVMCreateMessage with constant strings when calling llvm_raise.

Yeah something needs to be done with LLVMContextSetDiagnosticHandler anyway, as two regression tests currently fail because no handler is installed. I might do an easy fix for that first (just to get the tests back in working order) and then revisit how to properly expose the handlers.

Apr 1 2016, 1:18 AM
jketema committed rL265116: [OCaml] Use LLVMCreateMessage with constant strings when calling llvm_raise.
[OCaml] Use LLVMCreateMessage with constant strings when calling llvm_raise
Apr 1 2016, 1:01 AM
jketema closed D18675: [OCaml] Use LLVMCreateMessage with constant strings when calling llvm_raise by committing rL265116: [OCaml] Use LLVMCreateMessage with constant strings when calling llvm_raise.
Apr 1 2016, 1:01 AM
jketema committed rL265115: [OCaml] Reinstate data_layout.
[OCaml] Reinstate data_layout
Apr 1 2016, 12:59 AM
jketema closed D18677: [OCaml] Reinstate data_layout by committing rL265115: [OCaml] Reinstate data_layout.
Apr 1 2016, 12:59 AM

Mar 31 2016

jketema retitled D18677: [OCaml] Reinstate data_layout from to [OCaml] Reinstate data_layout.
Mar 31 2016, 3:33 PM
jketema retitled D18675: [OCaml] Use LLVMCreateMessage with constant strings when calling llvm_raise from to [OCaml] Use LLVMCreateMessage with constant strings when calling llvm_raise.
Mar 31 2016, 1:54 PM
jketema committed rL265001: Silence warnings in OCaml bindings.
Silence warnings in OCaml bindings
Mar 31 2016, 1:45 AM
jketema closed D18633: Silence warnings in OCaml bindings by committing rL265001: Silence warnings in OCaml bindings.
Mar 31 2016, 1:45 AM

Mar 30 2016

jketema updated the diff for D18633: Silence warnings in OCaml bindings.

Unbreak the related target test by removing add_to_pass_manager test

Mar 30 2016, 5:39 PM
jketema retitled D18633: Silence warnings in OCaml bindings from to Silence warnings in OCaml bindings.
Mar 30 2016, 4:50 PM

Feb 7 2016

jketema committed rL260026: Fix typo in default getNoPreservedMask implementation.
Fix typo in default getNoPreservedMask implementation
Feb 7 2016, 3:36 AM

Jan 4 2016

jketema committed rL256749: [MC] Fix file name in file header.
[MC] Fix file name in file header
Jan 4 2016, 4:26 AM

Oct 28 2015

jketema added inline comments to D14151: [X86][AVX] Fix lowering of X86ISD::VZEXT_MOVL for 128-bit -> 256-bit extension.
Oct 28 2015, 11:33 AM

Oct 7 2015

jketema added a comment to D13508: [ARM] Only lower to interleaved load/store if the target has NEON.

Committed. Thanks for the quick feedback!

Oct 7 2015, 7:55 AM
jketema committed rL249550: [ARM][AArch64] Only lower to interleaved load/store if the target has NEON.
[ARM][AArch64] Only lower to interleaved load/store if the target has NEON
Oct 7 2015, 7:55 AM
jketema closed D13508: [ARM] Only lower to interleaved load/store if the target has NEON by committing rL249550: [ARM][AArch64] Only lower to interleaved load/store if the target has NEON.
Oct 7 2015, 7:55 AM
jketema updated the diff for D13508: [ARM] Only lower to interleaved load/store if the target has NEON.

Also fix AArch64 and add the run line to existing test file instead (this also adds coverage for store tests).

Oct 7 2015, 7:34 AM
jketema added inline comments to D13508: [ARM] Only lower to interleaved load/store if the target has NEON.
Oct 7 2015, 6:48 AM
jketema added a comment to D13508: [ARM] Only lower to interleaved load/store if the target has NEON.

@rengolin If it is possible to conditionalize the CHECK-NOT on the run line. Is that possible? I haven't seen that before.

Oct 7 2015, 6:33 AM
jketema retitled D13508: [ARM] Only lower to interleaved load/store if the target has NEON from to [ARM] Only lower to interleaved load/store if the target has NEON.
Oct 7 2015, 6:13 AM

Oct 3 2015

jketema committed rL249253: Fix typo in README.
Fix typo in README
Oct 3 2015, 5:48 PM

Sep 30 2015

jketema added a comment to D12337: [Codegen] Ensure stack is properly aligned for call argument initialization.

Thanks for all your feedback Reid!

Sep 30 2015, 9:22 AM
jketema added a comment to D12985: [ARM] Take into account address spaces in interleaved access vectorization.

Committed. Thanks for all the feedback!

Sep 30 2015, 4:03 AM
jketema committed rL248887: [ARM][NEON] Use address space in vld([1234]|[234]lane) and vst([1234]|….
[ARM][NEON] Use address space in vld([1234]|[234]lane) and vst([1234]|…
Sep 30 2015, 3:58 AM
jketema committed rL248888: [ARM][NEON] Use address space in vld([1234]|[234]lane) and vst([1234]|….
[ARM][NEON] Use address space in vld([1234]|[234]lane) and vst([1234]|…
Sep 30 2015, 3:58 AM
jketema closed D13127: [ARM] Upgrade codegen for vld[234] and vst[234] to to communicate a 0 address space by committing rL248888: [ARM][NEON] Use address space in vld([1234]|[234]lane) and vst([1234]|….
Sep 30 2015, 3:58 AM
jketema closed D12985: [ARM] Take into account address spaces in interleaved access vectorization by committing rL248887: [ARM][NEON] Use address space in vld([1234]|[234]lane) and vst([1234]|….
Sep 30 2015, 3:58 AM

Sep 29 2015

jketema committed rL248786: Arguments spilled on the stack before a function call may have.
Arguments spilled on the stack before a function call may have
Sep 29 2015, 3:28 AM
jketema closed D12337: [Codegen] Ensure stack is properly aligned for call argument initialization by committing rL248786: Arguments spilled on the stack before a function call may have.
Sep 29 2015, 3:23 AM

Sep 28 2015

jketema updated the diff for D13127: [ARM] Upgrade codegen for vld[234] and vst[234] to to communicate a 0 address space.

Sync with updates to the llvm part of this patch

Sep 28 2015, 8:05 AM
jketema updated the diff for D12985: [ARM] Take into account address spaces in interleaved access vectorization.

This extends the patch to cover all of vld[1234], vls[234]lane, vst[1234], and vst[234]lane. All tests now use the updated intrinsic names, and new tests have been added to test auto-upgrading.

Sep 28 2015, 8:03 AM

Sep 24 2015

jketema updated the diff for D12985: [ARM] Take into account address spaces in interleaved access vectorization.

This now upgrades the vld[234] and vst[234] intrinsics to take into account an address space, as discussed.

Sep 24 2015, 3:47 AM
jketema retitled D13127: [ARM] Upgrade codegen for vld[234] and vst[234] to to communicate a 0 address space from to [ARM] Upgrade codegen for vld[234] and vst[234] to to communicate a 0 address space.
Sep 24 2015, 3:43 AM
jketema added a comment to D12337: [Codegen] Ensure stack is properly aligned for call argument initialization.

Bump

Sep 24 2015, 1:19 AM

Sep 23 2015

jketema added inline comments to D12985: [ARM] Take into account address spaces in interleaved access vectorization.
Sep 23 2015, 2:21 AM

Sep 22 2015

jketema added inline comments to D12985: [ARM] Take into account address spaces in interleaved access vectorization.
Sep 22 2015, 6:16 AM
jketema added inline comments to D12985: [ARM] Take into account address spaces in interleaved access vectorization.
Sep 22 2015, 4:42 AM

Sep 21 2015

jketema committed rL248208: [ARM] Do not scale vext with a factor.
[ARM] Do not scale vext with a factor
Sep 21 2015, 1:29 PM
jketema closed D12974: [ARM] Do not scale vext with a factor by committing rL248208: [ARM] Do not scale vext with a factor.
Sep 21 2015, 1:29 PM
jketema added a comment to D12974: [ARM] Do not scale vext with a factor.

So can this be committed, or does someone else still need to look at this?

Sep 21 2015, 9:31 AM
jketema added inline comments to D12985: [ARM] Take into account address spaces in interleaved access vectorization.
Sep 21 2015, 9:30 AM

Sep 19 2015

jketema updated the diff for D12337: [Codegen] Ensure stack is properly aligned for call argument initialization.

Hi Reid,

Sep 19 2015, 6:41 AM

Sep 18 2015

jketema retitled D12985: [ARM] Take into account address spaces in interleaved access vectorization from to [ARM] Take into account address spaces in interleaved access vectorization.
Sep 18 2015, 2:31 PM
jketema retitled D12974: [ARM] Do not scale vext with a factor from to [ARM] Do not scale vext with a factor.
Sep 18 2015, 10:24 AM

Sep 15 2015

jketema added inline comments to D12337: [Codegen] Ensure stack is properly aligned for call argument initialization.
Sep 15 2015, 1:48 PM
jketema added inline comments to D12337: [Codegen] Ensure stack is properly aligned for call argument initialization.
Sep 15 2015, 1:35 PM
jketema added a comment to D12337: [Codegen] Ensure stack is properly aligned for call argument initialization.

bump

Sep 15 2015, 1:29 PM

Sep 10 2015

jketema closed D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.

Thanks for committing James.

Sep 10 2015, 1:55 PM

Sep 9 2015

jketema requested review of D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.

If this is all ok now, please commit on my behalf.

Sep 9 2015, 2:04 PM
jketema updated the diff for D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.

Hi Ahmed,

Sep 9 2015, 2:03 PM

Sep 7 2015

jketema added a comment to D12337: [Codegen] Ensure stack is properly aligned for call argument initialization.

Bump

Sep 7 2015, 2:03 PM
jketema added a comment to D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.

Bump

Sep 7 2015, 2:03 PM

Sep 3 2015

jketema added a comment to D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.

Hi Ahmed,

Sep 3 2015, 5:57 PM
jketema updated the diff for D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.

Update tests.

Sep 3 2015, 5:52 PM
jketema requested review of D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.

Is this ok now with the added tests?

Sep 3 2015, 2:48 PM

Sep 2 2015

jketema added inline comments to D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.
Sep 2 2015, 4:20 AM
jketema updated the diff for D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.

I've added two test cases based on the suggestion by Ahmed.

Sep 2 2015, 4:15 AM
jketema added inline comments to D12334: [ARM] Do not use vtrn for vectorshuffle if the order is reversed.
Sep 2 2015, 3:30 AM