tony-tye (please use t-tye instead)
User

Projects

User does not belong to any projects.

User Details

User Since
Apr 15 2016, 11:26 AM (91 w, 5 d)

Recent Activity

Mar 22 2017

tony-tye edited reviewers for D21723: [RFC] Enhance synchscope representation, added: t-tye; removed: tony-tye.
Mar 22 2017, 6:10 PM
tony-tye edited reviewers for D24623: AMDGPU: Implement memory model, added: t-tye; removed: tony-tye.
Mar 22 2017, 6:10 PM
tony-tye edited reviewers for D28531: AMDGPU : Fix ISA Version Definitions., added: t-tye; removed: tony-tye.
Mar 22 2017, 6:08 PM
tony-tye edited reviewers for D29911: AMDGPU : Recalculate SGPRs when trap handler is supported., added: t-tye; removed: tony-tye.
Mar 22 2017, 6:07 PM
tony-tye accepted D29948: [AMDGPU] Restructure code object metadata creation.

LGTM

Mar 22 2017, 3:35 PM

Mar 15 2017

tony-tye added a comment to D30970: [AMDGPU] Do not emit isa info as code object metadata.

LGTM.

Do we have a general guideline what should be in metadata and what should not? Do we assume that end user should be able to manually craft the metadata when they write the ISA assembly by hand?

I'm not sure why we need to encode some of these static hardware properties into it. Why doesn't the runtime/debugger know these already?

Mar 15 2017, 9:51 AM

Mar 13 2017

tony-tye added a comment to D21723: [RFC] Enhance synchscope representation.

Just a wild thought:-) Since address spaces and memory scopes are both target defined concepts, it seems best for the target to define the enumeration of the ones it supports. Would it make sense for the data layout to not only give the properties of the address spaces, but also give their textual names, together with the textual names of the memory scopes supported by the target? The memory scopes would include the memory scope to use for singlethread and crossthread.

Mar 13 2017, 9:31 PM
tony-tye accepted D21723: [RFC] Enhance synchscope representation.

LGTM

Mar 13 2017, 12:45 PM

Feb 22 2017

tony-tye accepted D29671: [DebugInfo] Add address space when creating DIDerivedTypes.

LGTM

Feb 22 2017, 12:46 PM
tony-tye accepted D30232: Update TrapCode based on Trap Handler ABI..

LGTM

Feb 22 2017, 12:05 PM

Feb 21 2017

tony-tye requested changes to D30232: Update TrapCode based on Trap Handler ABI..
Feb 21 2017, 6:37 PM
tony-tye added inline comments to D30227: AMDGPU: Change m0 initialization handling to help LDS.
Feb 21 2017, 5:34 PM
tony-tye accepted D29670: [DebugInfo] Emit address space with DW_AT_address_class attribute for pointer and reference types.

LGTM

Feb 21 2017, 2:27 PM
tony-tye added inline comments to D29671: [DebugInfo] Add address space when creating DIDerivedTypes.
Feb 21 2017, 2:27 PM
tony-tye accepted D29913: AMDGPU : Update AMDGPU Trap Handler ABI..

LGTM

Feb 21 2017, 10:34 AM

Feb 17 2017

tony-tye added inline comments to D29913: AMDGPU : Update AMDGPU Trap Handler ABI..
Feb 17 2017, 4:33 PM
tony-tye added inline comments to D29913: AMDGPU : Update AMDGPU Trap Handler ABI..
Feb 17 2017, 3:52 PM
tony-tye requested changes to D29913: AMDGPU : Update AMDGPU Trap Handler ABI..
Feb 17 2017, 2:31 PM

Feb 16 2017

tony-tye accepted D29673: [DebugInfo] Append extended dereferencing mechanism to variables' DIExpression for targets that support more than one address space.

LGTM

Feb 16 2017, 7:45 PM
tony-tye added inline comments to D29671: [DebugInfo] Add address space when creating DIDerivedTypes.
Feb 16 2017, 7:40 PM

Feb 14 2017

tony-tye requested changes to D29913: AMDGPU : Update AMDGPU Trap Handler ABI..
Feb 14 2017, 3:33 PM
tony-tye added inline comments to D29911: AMDGPU : Recalculate SGPRs when trap handler is supported..
Feb 14 2017, 2:45 PM

Feb 9 2017

tony-tye accepted D29741: [AMDGPU] Calculate number of min/max SGPRs/VGPRs for WavesPerEU instead of using switch statement.

LGTM

Feb 9 2017, 1:42 PM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Feb 9 2017, 12:45 PM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Feb 9 2017, 9:25 AM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Feb 9 2017, 8:15 AM
tony-tye added inline comments to D29741: [AMDGPU] Calculate number of min/max SGPRs/VGPRs for WavesPerEU instead of using switch statement.
Feb 9 2017, 12:38 AM
tony-tye added inline comments to D29741: [AMDGPU] Calculate number of min/max SGPRs/VGPRs for WavesPerEU instead of using switch statement.
Feb 9 2017, 12:08 AM

Feb 8 2017

tony-tye added inline comments to D29741: [AMDGPU] Calculate number of min/max SGPRs/VGPRs for WavesPerEU instead of using switch statement.
Feb 8 2017, 5:53 PM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Feb 8 2017, 5:23 PM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Feb 8 2017, 5:17 PM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Feb 8 2017, 5:16 PM

Feb 7 2017

tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Feb 7 2017, 3:13 PM
tony-tye added inline comments to D29673: [DebugInfo] Append extended dereferencing mechanism to variables' DIExpression for targets that support more than one address space.
Feb 7 2017, 2:24 PM

Feb 2 2017

tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Feb 2 2017, 10:00 PM

Jan 31 2017

tony-tye added a comment to D26010: AMDGPU : Add trap handler support..

Also need to use the trap handler ABI query to see if there is a trap handler, and if there is add the TRAP_HANDLER_SGPR_COUNT to the number of SGPRs budgeted for the wave in determining the number of waves per EU calculation. TRAP_HANDLER_SGPR_COUNT is 16 for GFX6 onwards.

Jan 31 2017, 3:41 PM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Jan 31 2017, 3:32 PM

Jan 28 2017

tony-tye added inline comments to D29214: [AMDGPU] Internalize non-kernel symbols.
Jan 28 2017, 12:42 AM

Jan 27 2017

tony-tye added inline comments to D29214: [AMDGPU] Internalize non-kernel symbols.
Jan 27 2017, 11:18 PM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Jan 27 2017, 9:25 PM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Jan 27 2017, 2:34 PM
tony-tye added inline comments to D29214: [AMDGPU] Internalize non-kernel symbols.
Jan 27 2017, 11:31 AM

Jan 26 2017

tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Jan 26 2017, 9:10 PM
tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Jan 26 2017, 5:32 PM
tony-tye added inline comments to D28937: AMDGPU: Treat 0 as private null pointer in addrspacecast lowering.
Jan 26 2017, 4:42 PM

Jan 24 2017

tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Jan 24 2017, 12:59 PM

Nov 8 2016

tony-tye added inline comments to D26196: Add support for non-zero null pointer for C and OpenCL.
Nov 8 2016, 10:57 AM

Nov 2 2016

tony-tye added a comment to D25998: AMDGPU/SI: Don't use non-0 waitcnt values when waiting on Flat instructions.

Feedback on overall pass.

Nov 2 2016, 12:01 AM

Oct 27 2016

tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Oct 27 2016, 4:16 PM
tony-tye updated subscribers of D25944: [AMDGPU][CodeGen] To improve CGEMM performance: combine LDS reads..
Oct 27 2016, 3:28 PM

Oct 26 2016

tony-tye added inline comments to D26010: AMDGPU : Add trap handler support..
Oct 26 2016, 5:39 PM
tony-tye requested changes to D26010: AMDGPU : Add trap handler support..
Oct 26 2016, 3:38 PM

Oct 25 2016

tony-tye accepted D25788: AMDGPU/SI: Don't emit multi-dword flat memory ops when they might access scratch.

LGTM

Oct 25 2016, 7:17 PM

Oct 18 2016

tony-tye accepted D25746: [AMDGPU] Make note record name a static const member of target streamer..

LGTM

Oct 18 2016, 2:32 PM

Oct 17 2016

tony-tye added inline comments to D25693: [AMDGPU] Emit constant address space data in .rodata section and use relocations instead of fixups (amdhsa only).
Oct 17 2016, 1:11 PM

Sep 21 2016

tony-tye added a comment to D21723: [RFC] Enhance synchscope representation.

I would argue that even without (2) there is a case for using standard LLVM atomic instructions as it makes it possible to generate the same IR for a given language regardless of whether the target supports scopes. It seems CLANG is moving towards generating LLVM IR atomics directly rather than calls to built-ins so this approach would support that. Currently LLVM already supports two scopes. It also makes code generation much simpler as the existing machine instructions can be used and so avoid creating a large number of pseudo instructions. The patches D24577 and D24623 that are under review are pretty simple and do that. There would be a lot more code if intrinsics had to be used.

Sep 21 2016, 9:30 AM

Jun 10 2016

tony-tye added a comment to D21215: [docs] Update AMDGPU relocation information.

The old R_AMDGPU_ABS32_LO and the new R_AMDGPU_ABS32 are in fact the same thing. The & 0xffffffff is implicitly done because the result of the R_AMDGPU_ABS32 is a word32 not a word64. So putting in the & 0xffffffff is redundant. The other ABI documents (such as for the x86) do not put in the & 0xffffffff so it seemed best to follow their conventions. R_AMDGPU_ABS32_HI and R_AMDGPU_ABS32 both return 32 bits as they are both defined as word32. They differ in that R_AMDGPU_ABS32_HI takes the address and shifts it right by 32 bits which effectively means that the top 32 bits of the 64 bit address are returned, not the bottom 32 bits.

Jun 10 2016, 10:42 PM