Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / robertknight/rten issues and pull requests
#590 - Move GEMM prepacking and output tile management into submodules
Pull Request -
State: closed - Opened by robertknight 3 days ago
#589 - Support vector inputs in MatMulInteger
Pull Request -
State: closed - Opened by robertknight 4 days ago
#588 - Unify reference implementations of `MatMul` and `MatMulInteger`
Pull Request -
State: closed - Opened by robertknight 4 days ago
#587 - Eliminate some duplicated code in `Cast` op
Pull Request -
State: closed - Opened by robertknight 5 days ago
#587 - Eliminate some duplicated code in `Cast` op
Pull Request -
State: closed - Opened by robertknight 5 days ago
#587 - Eliminate some duplicated code in `Cast` op
Pull Request -
State: closed - Opened by robertknight 5 days ago
#586 - Bump up x64 int8 tile size
Pull Request -
State: closed - Opened by robertknight 5 days ago
#586 - Bump up x64 int8 tile size
Pull Request -
State: closed - Opened by robertknight 5 days ago
#586 - Bump up x64 int8 tile size
Pull Request -
State: closed - Opened by robertknight 5 days ago
#585 - Bump tile size for wasm32 and Arm (non-dot product) int8 kernels
Pull Request -
State: closed - Opened by robertknight 5 days ago
#585 - Bump tile size for wasm32 and Arm (non-dot product) int8 kernels
Pull Request -
State: closed - Opened by robertknight 5 days ago
#585 - Bump tile size for wasm32 and Arm (non-dot product) int8 kernels
Pull Request -
State: closed - Opened by robertknight 5 days ago
#584 - Add documentation on quantization support in rten
Pull Request -
State: closed - Opened by robertknight 6 days ago
#584 - Add documentation on quantization support in rten
Pull Request -
State: closed - Opened by robertknight 6 days ago
#584 - Add documentation on quantization support in rten
Pull Request -
State: closed - Opened by robertknight 6 days ago
#583 - Avoid underflow on x64 when converting weights from i8 -> u8 in ConvInteger
Pull Request -
State: closed - Opened by robertknight 7 days ago
#582 - Generalize MatMulInteger to support other int8 type combinations in future
Pull Request -
State: closed - Opened by robertknight 7 days ago
#582 - Generalize MatMulInteger to support other int8 type combinations in future
Pull Request -
State: closed - Opened by robertknight 7 days ago
#582 - Generalize MatMulInteger to support other int8 type combinations in future
Pull Request -
State: closed - Opened by robertknight 7 days ago
#581 - Use macros to simplify invoking operators with different input types
Pull Request -
State: closed - Opened by robertknight 7 days ago
#581 - Use macros to simplify invoking operators with different input types
Pull Request -
State: closed - Opened by robertknight 7 days ago
#581 - Use macros to simplify invoking operators with different input types
Pull Request -
State: closed - Opened by robertknight 7 days ago
#580 - Support NR being a multiple of vector width in int8 GEMM
Pull Request -
State: closed - Opened by robertknight 8 days ago
#580 - Support NR being a multiple of vector width in int8 GEMM
Pull Request -
State: closed - Opened by robertknight 8 days ago
#580 - Support NR being a multiple of vector width in int8 GEMM
Pull Request -
State: closed - Opened by robertknight 8 days ago
#579 - Support prepacking weights for MatMulInteger
Pull Request -
State: open - Opened by robertknight 8 days ago
#579 - Support prepacking weights for MatMulInteger
Pull Request -
State: closed - Opened by robertknight 8 days ago
#578 - MatMulNBits support for 4-bit quantization
Issue -
State: open - Opened by robertknight 9 days ago
Labels: quantization
#578 - MatMulNBits support for 4-bit quantization
Issue -
State: open - Opened by robertknight 9 days ago
Labels: quantization
#578 - MatMulNBits support for 4-bit quantization
Issue -
State: open - Opened by robertknight 9 days ago
Labels: quantization
#577 - Support `axis` equal to tensor rank in `Flatten` operator
Pull Request -
State: closed - Opened by robertknight 9 days ago
#577 - Support `axis` equal to tensor rank in `Flatten` operator
Pull Request -
State: closed - Opened by robertknight 9 days ago
#577 - Support `axis` equal to tensor rank in `Flatten` operator
Pull Request -
State: closed - Opened by robertknight 9 days ago
#576 - Support all input tensor types in `Split` operator
Pull Request -
State: closed - Opened by robertknight 9 days ago
#576 - Support all input tensor types in `Split` operator
Pull Request -
State: closed - Opened by robertknight 9 days ago
#576 - Support all input tensor types in `Split` operator
Pull Request -
State: closed - Opened by robertknight 9 days ago
#576 - Support all input tensor types in `Split` operator
Pull Request -
State: closed - Opened by robertknight 9 days ago
#575 - Support `auto_pad=VALID`
Pull Request -
State: closed - Opened by robertknight 9 days ago
#575 - Support `auto_pad=VALID`
Pull Request -
State: closed - Opened by robertknight 9 days ago
#575 - Support `auto_pad=VALID`
Pull Request -
State: closed - Opened by robertknight 9 days ago
#575 - Support `auto_pad=VALID`
Pull Request -
State: closed - Opened by robertknight 9 days ago
#574 - Improve handling of weights with a non-reduced range in `ConvInteger`
Issue -
State: closed - Opened by robertknight 10 days ago
- 1 comment
Labels: quantization
#574 - Improve handling of weights with a non-reduced range in `ConvInteger`
Issue -
State: closed - Opened by robertknight 10 days ago
- 1 comment
Labels: quantization
#574 - Improve handling of weights with a non-reduced range in `ConvInteger`
Issue -
State: closed - Opened by robertknight 10 days ago
- 1 comment
Labels: quantization
#573 - Fix wasmtime install in CI
Pull Request -
State: closed - Opened by robertknight 10 days ago
#573 - Fix wasmtime install in CI
Pull Request -
State: closed - Opened by robertknight 10 days ago
#573 - Fix wasmtime install in CI
Pull Request -
State: closed - Opened by robertknight 10 days ago
#572 - Broadcast zero point vector when converting batched matmul to non-batched
Pull Request -
State: closed - Opened by robertknight 10 days ago
#571 - Add missing `#[inline(always)]` to `Quantize` SIMD op
Pull Request -
State: closed - Opened by robertknight 10 days ago
#570 - Implement im2col packing for int8 GEMM
Pull Request -
State: closed - Opened by robertknight 11 days ago
#569 - Expand `ReducedRangeRng` and make it reusable outside GEMM tests
Pull Request -
State: closed - Opened by robertknight 11 days ago
#568 - Add `CowTensor`, `CowNdTensor` type aliases
Pull Request -
State: closed - Opened by robertknight 11 days ago
#567 - Simplify and optimize i8/i16 interleaving SIMD ops for x64
Pull Request -
State: closed - Opened by robertknight 12 days ago
#566 - Implement `ConvInteger` operator
Pull Request -
State: closed - Opened by robertknight 13 days ago
#565 - Make panic message more helpful if `Tensor::{size, stride}` arg is invalid
Pull Request -
State: closed - Opened by robertknight 13 days ago
#564 - Fix I8 -> U8 conversion in int8 gemv transposed case
Pull Request -
State: closed - Opened by robertknight 13 days ago
#563 - Make `Simd::to_array` impls a simple transmute
Pull Request -
State: closed - Opened by robertknight 14 days ago
#562 - Remove unused `SimdInt::load_interleaved_i8`
Pull Request -
State: closed - Opened by robertknight 14 days ago
#561 - Pre-fetch next column block in int8 GEMV
Pull Request -
State: closed - Opened by robertknight 14 days ago
#560 - Fix x86 macOS build and run `cargo check` for Intel macOS in CI
Pull Request -
State: closed - Opened by robertknight 14 days ago
#559 - Optimize non-transposed int8 GEMV kernel
Pull Request -
State: closed - Opened by robertknight 15 days ago
#558 - Support generating inputs for bool tensors in ort-infer.py
Pull Request -
State: closed - Opened by robertknight 18 days ago
#557 - Revise range reduction in int8 GEMM tests to match real model inputs
Pull Request -
State: closed - Opened by robertknight 18 days ago
#556 - Implement baseline int8 kernel for WASM
Pull Request -
State: closed - Opened by robertknight 18 days ago
#555 - Skip Python lint on macOS CI
Pull Request -
State: closed - Opened by robertknight 18 days ago
#554 - Run tests under WASM in CI
Pull Request -
State: closed - Opened by robertknight 18 days ago
#553 - Adjust matmul tile size on Arm
Issue -
State: open - Opened by robertknight 18 days ago
Labels: performance
#552 - Set default number of threads to performance core count on macOS
Pull Request -
State: closed - Opened by robertknight 19 days ago
#551 - Optimize `SimdInt::saturating_cast_u8` for Arm
Pull Request -
State: closed - Opened by robertknight 19 days ago
#550 - Exclude `Conv` op from quantization in ort-quantize.py
Pull Request -
State: closed - Opened by robertknight 19 days ago
#549 - Reduce use of `unsafe` in portable SIMD library
Issue -
State: open - Opened by robertknight 19 days ago
- 1 comment
Labels: safety
#549 - Reduce use of `unsafe` in portable SIMD library
Issue -
State: open - Opened by robertknight 19 days ago
- 1 comment
Labels: safety
#548 - Refactor depthwise convolution to support arch/data type-specific kernels
Pull Request -
State: closed - Opened by robertknight 19 days ago
#548 - Refactor depthwise convolution to support arch/data type-specific kernels
Pull Request -
State: closed - Opened by robertknight 19 days ago
#547 - Add build with AVX-512 enabled in CI
Pull Request -
State: closed - Opened by robertknight 20 days ago
#547 - Add build with AVX-512 enabled in CI
Pull Request -
State: closed - Opened by robertknight 20 days ago
#546 - Update notes about data type and quantization support in the docs
Pull Request -
State: closed - Opened by robertknight 20 days ago
#546 - Update notes about data type and quantization support in the docs
Pull Request -
State: closed - Opened by robertknight 20 days ago
#545 - Simplify running tests against all supported f32 kernels
Pull Request -
State: closed - Opened by robertknight 20 days ago
#544 - Add int8 fallback kernel for older Arm CPUs that don't support UDOT
Pull Request -
State: closed - Opened by robertknight 20 days ago
- 1 comment
#544 - Add int8 fallback kernel for older Arm CPUs that don't support UDOT
Pull Request -
State: closed - Opened by robertknight 20 days ago
- 1 comment
#543 - Convert Arm int8 kernels to use UDOT instead of USDOT
Pull Request -
State: closed - Opened by robertknight 21 days ago
#543 - Convert Arm int8 kernels to use UDOT instead of USDOT
Pull Request -
State: closed - Opened by robertknight 21 days ago
#542 - Add initial aarch64 int8 kernel using USDOT
Pull Request -
State: closed - Opened by robertknight 22 days ago
#542 - Add initial aarch64 int8 kernel using USDOT
Pull Request -
State: closed - Opened by robertknight 22 days ago
#541 - Implement optimized int8 gemv kernel using portable SIMD
Pull Request -
State: closed - Opened by robertknight 25 days ago
#540 - Fuse quantized matmul subgraphs for dynamic quantization
Issue -
State: open - Opened by robertknight 27 days ago
#539 - Fix wrong B zero points being passed to kernel in int8 gemv
Pull Request -
State: closed - Opened by robertknight 28 days ago
#538 - Vectorize QuantizeLinear, second stage of DynamicQuantizeLinear
Pull Request -
State: closed - Opened by robertknight 28 days ago
- 1 comment
#537 - Add AVX-512 int8 GEMM using VNNI
Pull Request -
State: closed - Opened by robertknight 29 days ago
#536 - Use fused multiply-add instructions when WASM relaxed-simd feature is enabled
Pull Request -
State: closed - Opened by robertknight about 1 month ago
#535 - Implement u8 x i8 -> i32 GEMM kernel for x86_64 using AVX2 intrinsics
Pull Request -
State: closed - Opened by robertknight about 1 month ago
#534 - Add helper to ensure packing buffer is fully written, remove unused col-major packing code
Pull Request -
State: closed - Opened by robertknight about 1 month ago
#533 - Extract slice-casting functions into own module, remove some dead code
Pull Request -
State: closed - Opened by robertknight about 1 month ago
#532 - Add small optimizations for QuantizeLinear
Pull Request -
State: closed - Opened by robertknight about 1 month ago
#531 - Compute input min/max with a single vectorized pass in DynamicQuantizeLinear
Pull Request -
State: closed - Opened by robertknight about 1 month ago
- 1 comment
#530 - Support subgraphs and use safer quantization settings in ort-quantize.py
Pull Request -
State: closed - Opened by robertknight about 1 month ago
#529 - Add a basic benchmark for int8 matmul
Pull Request -
State: closed - Opened by robertknight about 1 month ago
#528 - Implement minimal u8 x i8 -> i32 quantized GEMM support
Pull Request -
State: closed - Opened by robertknight about 1 month ago
#527 - Refactor reference GEMM implementation and make it generic
Pull Request -
State: closed - Opened by robertknight about 1 month ago