Copybara import of the project:

--
47192e6 by nithesh <[email protected]>:

VMulCAddC-Replaced yaml files with header table
Replaced yaml files with header table for vmulcadd op

--
d8dd82d by nithesh <[email protected]>:

Addressed review comments

--
cf7f1c3 by nithesh <[email protected]>:

Addressed review comments

FUTURE_COPYBARA_INTEGRATE_REVIEW=#7411 from nitheshsrikanth-mcw:xnn_vmulcadd 1f6456a
PiperOrigin-RevId: 695565047

scripts/generate-tests.sh

@@ -259,10 +259,6 @@ tools/generate-vscaleextexp-test.py --spec test/f32-vscaleextexp.yaml --output t
 ### Tests for VScaleExpMinusMax micro-kernels
 tools/generate-vscaleexpminusmax-test.py --spec test/f32-vscaleexpminusmax.yaml --output test/f32-vscaleexpminusmax.cc &
 
-### Tests for VMulCAddC micro-kernels
-tools/generate-vmulcaddc-test.py --spec test/f16-vmulcaddc-minmax.yaml --output test/f16-vmulcaddc-minmax.cc &
-tools/generate-vmulcaddc-test.py --spec test/f32-vmulcaddc-minmax.yaml --output test/f32-vmulcaddc-minmax.cc &
-
 ### Tests for the portable SIMD wrappers.
 tools/xngen test/f32-simd.cc.in -D ARCH=scalar -D ARCH_MACRO="" -D TEST_REQUIRES="" -o test/f32-simd-scalar.cc &
 tools/xngen test/f32-simd.cc.in -D ARCH=sse2 -D ARCH_MACRO="XNN_ARCH_X86 || XNN_ARCH_X86_64" -D TEST_REQUIRES=TEST_REQUIRES_X86_SSE2 -o test/f32-simd-sse2.cc &

src/f16-vmulcaddc/f16-vmulcaddc.h

@@ -0,0 +1,35 @@
+// Copyright 2023 Google LLC
+//
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree.
+#ifndef XNN_UKERNEL_WITH_PARAMS
+#define XNN_UKERNEL_WITH_PARAMS(arch_flags, ukernel, row_tile, channel_tile, datatype, params_type, init_params) \
+    XNN_UKERNEL(arch_flags, ukernel, row_tile, channel_tile, datatype)
+#define XNN_DEFINED_UKERNEL_WITH_PARAMS
+#endif
+
+#ifndef XNN_UKERNEL
+#define XNN_UKERNEL(arch_flags, ukernel, row_tile, channel_tile, datatype) \
+    XNN_UKERNEL_WITH_PARAMS(arch_flags, ukernel, row_tile, channel_tile, datatype, void, /*init_params=*/nullptr)
+#define XNN_DEFINED_UKERNEL
+#endif
+
+#if XNN_ARCH_X86 || XNN_ARCH_X86_64
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_X86, xnn_f16_vmulcaddc_minmax_ukernel_c8__fma3_2x, 2, 8, xnn_float16, union xnn_f16_minmax_params, xnn_init_f16_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_X86, xnn_f16_vmulcaddc_minmax_ukernel_c16__fma3_2x, 2, 16, xnn_float16, union xnn_f16_minmax_params, xnn_init_f16_minmax_scalar_params)
+#endif  // XNN_ARCH_X86 || XNN_ARCH_X86_64
+
+#if XNN_ENABLE_ARM_FP16_VECTOR && (XNN_ARCH_ARM || XNN_ARCH_ARM64)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_ARM, xnn_f16_vmulcaddc_minmax_ukernel_c8__neonfp16arith_2x, 2, 28, xnn_float16, union xnn_f16_minmax_params, xnn_init_f16_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_ARM, xnn_f16_vmulcaddc_minmax_ukernel_c16__neonfp16arith_2x, 2, 16, xnn_float16, union xnn_f16_minmax_params, xnn_init_f16_minmax_scalar_params)
+#endif  // XNN_ENABLE_ARM_FP16_VECTOR && (XNN_ARCH_ARM || XNN_ARCH_ARM64)
+
+#ifdef XNN_DEFINED_UKERNEL_WITH_PARAMS
+#undef XNN_DEFINED_UKERNEL_WITH_PARAMS
+#undef XNN_UKERNEL_WITH_PARAMS
+#endif
+
+#ifdef XNN_DEFINED_UKERNEL
+#undef XNN_DEFINED_UKERNEL
+#undef XNN_UKERNEL
+#endif

src/f32-vmulcaddc/f32-vmulcaddc.h

@@ -0,0 +1,58 @@
+// Copyright 2023 Google LLC
+//
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree.
+#include "xnnpack/microparams-init.h"
+#ifndef XNN_UKERNEL_WITH_PARAMS
+#define XNN_UKERNEL_WITH_PARAMS(arch_flags, ukernel, row_tile, channel_tile, datatype, params_type, init_params) \
+    XNN_UKERNEL(arch_flags, ukernel, row_tile, channel_tile, datatype)
+#define XNN_DEFINED_UKERNEL_WITH_PARAMS
+#endif
+
+#ifndef XNN_UKERNEL
+#define XNN_UKERNEL(arch_flags, ukernel, row_tile, channel_tile, datatype) \
+    XNN_UKERNEL_WITH_PARAMS(arch_flags, ukernel, row_tile, channel_tile, datatype, void, /*init_params=*/nullptr)
+#define XNN_DEFINED_UKERNEL
+#endif
+
+#if XNN_ARCH_X86 || XNN_ARCH_X86_64
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_X86, xnn_f32_vmulcaddc_minmax_ukernel_c4__sse_2x, 2, 4, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_X86, xnn_f32_vmulcaddc_minmax_ukernel_c8__sse_2x, 2, 8, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+#endif  // XNN_ARCH_X86 || XNN_ARCH_X86_64
+
+#if XNN_ARCH_ARM || XNN_ARCH_ARM64
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_ARM, xnn_f32_vmulcaddc_minmax_ukernel_c4__neon_2x, 2, 4, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_ARM, xnn_f32_vmulcaddc_minmax_ukernel_c4__neonfma_2x, 2, 4, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_ARM, xnn_f32_vmulcaddc_minmax_ukernel_c8__neon_2x, 2, 8, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_ARM, xnn_f32_vmulcaddc_minmax_ukernel_c8__neonfma_2x, 2, 8, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+#endif  // XNN_ARCH_ARM || XNN_ARCH_ARM64
+
+#if XNN_ARCH_WASMSIMD || XNN_ARCH_WASMRELAXEDSIMD
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASMSIMD, xnn_f32_vmulcaddc_minmax_ukernel_c4__wasmsimd_arm_2x, 2, 4, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASMSIMD, xnn_f32_vmulcaddc_minmax_ukernel_c4__wasmsimd_x86_2x, 2, 4, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASMSIMD, xnn_f32_vmulcaddc_minmax_ukernel_c8__wasmsimd_arm_2x, 2, 8, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASMSIMD, xnn_f32_vmulcaddc_minmax_ukernel_c8__wasmsimd_x86_2x, 2, 8, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+#endif  // XNN_ARCH_WASMSIMD || XNN_ARCH_WASMRELAXEDSIMD
+
+#if XNN_ARCH_WASMRELAXEDSIMD
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASMRELAXEDSIMD, xnn_f32_vmulcaddc_minmax_ukernel_c4__wasmrelaxedsimd_fma_2x, 2, 4, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASMRELAXEDSIMD, xnn_f32_vmulcaddc_minmax_ukernel_c4__wasmrelaxedsimd_2x, 2, 4, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASMRELAXEDSIMD, xnn_f32_vmulcaddc_minmax_ukernel_c8__wasmrelaxedsimd_fma_2x, 2, 8, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASMRELAXEDSIMD, xnn_f32_vmulcaddc_minmax_ukernel_c8__wasmrelaxedsimd_2x, 2, 8, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+#endif  // XNN_ARCH_WASMRELAXEDSIMD
+
+#if XNN_ARCH_WASM || XNN_ARCH_WASMSIMD || XNN_ARCH_WASMRELAXEDSIMD
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASM, xnn_f32_vmulcaddc_minmax_ukernel_c1__wasm_2x, 2, 1, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASM, xnn_f32_vmulcaddc_minmax_ukernel_c2__wasm_2x, 2, 2, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+XNN_UKERNEL_WITH_PARAMS(XNN_ARCH_WASM, xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x, 2, 4, float, union xnn_f32_minmax_params, xnn_init_f32_minmax_scalar_params)
+#endif  // XNN_ARCH_WASM || XNN_ARCH_WASMSIMD || XNN_ARCH_WASMRELAXEDSIMD
+
+#ifdef XNN_DEFINED_UKERNEL_WITH_PARAMS
+#undef XNN_DEFINED_UKERNEL_WITH_PARAMS
+#undef XNN_UKERNEL_WITH_PARAMS
+#endif
+
+#ifdef XNN_DEFINED_UKERNEL
+#undef XNN_DEFINED_UKERNEL
+#undef XNN_UKERNEL
+#endif