NEON Sigmoid evaluation stubs with 1-step range reduction
PiperOrigin-RevId: 288321666
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 327748a..54d1172 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -562,9 +562,12 @@
src/x8-zip/x4-neon.c
src/x8-zip/xm-neon.c
src/math/sigmoid-neon-frac-p9-p10-nr1recps.c
- src/math/sigmoid-neon-lut2048-p1-nr2recps.c
- src/math/sigmoid-neon-lut64-p2-nr2recps.c
- src/math/sigmoid-neon-p5-nr2recps.c)
+ src/math/sigmoid-neon-rr1-lut2048-p1-nr2recps.c
+ src/math/sigmoid-neon-rr1-lut64-p2-nr2recps.c
+ src/math/sigmoid-neon-rr1-p5-nr2recps.c
+ src/math/sigmoid-neon-rr2-lut2048-p1-nr2recps.c
+ src/math/sigmoid-neon-rr2-lut64-p2-nr2recps.c
+ src/math/sigmoid-neon-rr2-p5-nr2recps.c)
SET(XNNPACK_NEONFMA_MICROKERNEL_SRCS
src/f32-bilinear/gen/neonfma-c4.c
@@ -665,15 +668,24 @@
src/math/expminus-neonfma-lut2048-p1.c
src/math/expminus-neonfma-lut64-p2.c
src/math/expminus-neonfma-p5.c
- src/math/sigmoid-neonfma-lut2048-p1-nr1recps1fma.c
- src/math/sigmoid-neonfma-lut2048-p1-nr2fma.c
- src/math/sigmoid-neonfma-lut2048-p1-nr2recps.c
- src/math/sigmoid-neonfma-lut64-p2-nr1recps1fma.c
- src/math/sigmoid-neonfma-lut64-p2-nr2fma.c
- src/math/sigmoid-neonfma-lut64-p2-nr2recps.c
- src/math/sigmoid-neonfma-p5-nr1recps1fma.c
- src/math/sigmoid-neonfma-p5-nr2fma.c
- src/math/sigmoid-neonfma-p5-nr2recps.c)
+ src/math/sigmoid-neonfma-rr1-lut2048-p1-nr1recps1fma.c
+ src/math/sigmoid-neonfma-rr1-lut2048-p1-nr2fma.c
+ src/math/sigmoid-neonfma-rr1-lut2048-p1-nr2recps.c
+ src/math/sigmoid-neonfma-rr1-lut64-p2-nr1recps1fma.c
+ src/math/sigmoid-neonfma-rr1-lut64-p2-nr2fma.c
+ src/math/sigmoid-neonfma-rr1-lut64-p2-nr2recps.c
+ src/math/sigmoid-neonfma-rr1-p5-nr1recps1fma.c
+ src/math/sigmoid-neonfma-rr1-p5-nr2fma.c
+ src/math/sigmoid-neonfma-rr1-p5-nr2recps.c
+ src/math/sigmoid-neonfma-rr2-lut2048-p1-nr1recps1fma.c
+ src/math/sigmoid-neonfma-rr2-lut2048-p1-nr2fma.c
+ src/math/sigmoid-neonfma-rr2-lut2048-p1-nr2recps.c
+ src/math/sigmoid-neonfma-rr2-lut64-p2-nr1recps1fma.c
+ src/math/sigmoid-neonfma-rr2-lut64-p2-nr2fma.c
+ src/math/sigmoid-neonfma-rr2-lut64-p2-nr2recps.c
+ src/math/sigmoid-neonfma-rr2-p5-nr1recps1fma.c
+ src/math/sigmoid-neonfma-rr2-p5-nr2fma.c
+ src/math/sigmoid-neonfma-rr2-p5-nr2recps.c)
SET(XNNPACK_AARCH64_NEONFMA_MICROKERNEL_SRCS
src/f32-vbinary/gen/vdiv-neon-x4.c
@@ -745,9 +757,12 @@
src/f32-spmm/gen/8x1-neonfma.c
src/f32-spmm/gen/8x2-neonfma.c
src/f32-spmm/gen/8x4-neonfma.c
- src/math/sigmoid-neonfma-lut2048-p1-div.c
- src/math/sigmoid-neonfma-lut64-p2-div.c
- src/math/sigmoid-neonfma-p5-div.c)
+ src/math/sigmoid-neonfma-rr1-lut2048-p1-div.c
+ src/math/sigmoid-neonfma-rr1-lut64-p2-div.c
+ src/math/sigmoid-neonfma-rr1-p5-div.c
+ src/math/sigmoid-neonfma-rr2-lut2048-p1-div.c
+ src/math/sigmoid-neonfma-rr2-lut64-p2-div.c
+ src/math/sigmoid-neonfma-rr2-p5-div.c)
SET(XNNPACK_AARCH64_NEONFP16ARITH_MICROKERNEL_SRCS
src/f16-gemm/gen/4x8-neonfp16arith-ld64.c