Issues

• f64: QuantizePow34Bands recomputes pow34ZeroBound on every call

  tphakala/simd#342 · Filter by author tphakala opened on Nov 7, 2026

• f32, f64: RealFFTPower docs say gc does not fuse on amd64, false at GOAMD64=v3

  tphakala/simd#340 · Filter by author tphakala opened on Nov 7, 2026

• f64: QuantizePow34Bands has no test that faults on reads or writes past sum(widths)

  tphakala/simd#339 · Filter by author tphakala opened on Nov 7, 2026

• f64: no test pins that bit-identical ops actually dispatch to their SIMD kernels

  tphakala/simd#338 · Filter by author tphakala opened on Nov 7, 2026

• f64: amd64 impl snapshot in go_impl_amd64_test.go omits affineImpl and convolveDecimateImpl

  tphakala/simd#334 · Filter by author tphakala opened on Nov 7, 2026

• f32: faster NEON Pow, zero lanes in the Pow fast path, vectorized partial RealFFT unravel

  tphakala/simd#331 · Filter by author tphakala opened on Nov 6, 2026

• f32: recover the small-shape MinIdxOfSumRows slowdown from the shared fused kernels (#328)

  tphakala/simd#330 · Filter by author tphakala opened on Nov 5, 2026
  1

• amd64: legacy SSE MOVQ/MOVD while YMM state is dirty in i8, i16 and i32 AVX2 kernels

  tphakala/simd#327 · Filter by author tphakala opened on Nov 5, 2026

• amd64: function-variable dispatch makes caller stack buffers escape to the heap

  tphakala/simd#326 · Filter by author tphakala opened on Nov 5, 2026
  1

• f64: measure the ~2% STFT/IRFFT slowdown from the shared FFT core (#322)

  tphakala/simd#325 · Filter by author tphakala opened on Oct 19, 2026

• test: f64 tests do not compile for 32-bit targets, and CI has no 32-bit leg

  tphakala/simd#324 · Filter by author tphakala opened on Oct 19, 2026

• ci: the AVX-512 SDE job can hang for hours when a test binary does not exit under sde64

  tphakala/simd#320 · Filter by author tphakala opened on Oct 16, 2026
  1

• Go 1.27 arm64 assembler: NEON mnemonics now cover much of our hand-encoded WORD set (informational)

  tphakala/simd#287 · Filter by author tphakala opened on Sep 8, 2026
  1

• ConvolveValidMaxAbs: add AVX-512 fused kernel when hardware is available

  tphakala/simd#141 · Filter by author tphakala opened on Jun 14, 2026
  1

• Tracking: extend the int8/byte SIMD surface (i8, and a possible u8)

  enhancement
  tphakala/simd#132 · Filter by author tphakala opened on Jun 13, 2026
  4