Added a copy helper

benchmarks/CMakeLists.txt

@@ -57,6 +57,9 @@ target_link_libraries(bm_ADSR benchmark absl::span absl::algorithm)
 add_executable(bm_add BM_add.cpp ${SFIZZ_SIMD_SOURCES})
 target_link_libraries(bm_add benchmark absl::span absl::algorithm)
 
+add_executable(bm_copy BM_copy.cpp ${SFIZZ_SIMD_SOURCES})
+target_link_libraries(bm_copy benchmark absl::span absl::algorithm)
+
 add_custom_target(sfizz_benchmarks)
 add_dependencies(sfizz_benchmarks 
 	bm_opf_high_vs_low 

external/Catch2

@@ -1 +1 @@
-Subproject commit ce42deb72fab2be85a862f559984580c24cb76c4
+Subproject commit b77ab74b723cbd55ee12883562e22c7e467dbcb1

external/abseil-cpp

@@ -1 +1 @@
-Subproject commit 83c1d65c90a92aa49632b9ac5a793214bb0768bc
+Subproject commit 325fd7b042ff4ec34f7dd32e602cd81ad0e24b22

sfizz/Config.h

@@ -54,4 +54,5 @@ namespace SIMDConfig {
     constexpr bool linearRamp { false };
     constexpr bool multiplicativeRamp { true };
     constexpr bool add { false };
+    constexpr bool copy { false };
 }

sfizz/SIMDDummy.cpp

@@ -106,4 +106,10 @@ template <>
 void add<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
 {
     add<float, false>(input, output);
+}
+
+template <>
+void copy<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
+{
+    copy<float, false>(input, output);
 }

sfizz/SIMDHelpers.h

@@ -325,4 +325,25 @@ void add(absl::Span<const T> input, absl::Span<T> output) noexcept
 }
 
 template <>
-void add<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept;
+void add<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept;
+
+
+template <class T>
+void snippetCopy(const T*& input, T*& output)
+{
+    *output++ = *input++;
+}
+
+template <class T, bool SIMD = SIMDConfig::copy>
+void copy(absl::Span<const T> input, absl::Span<T> output) noexcept
+{
+    ASSERT(output.size() >= input.size());
+    auto* in = input.begin();
+    auto* out = output.begin();
+    auto* sentinel = out + min(input.size(), output.size());
+    while (out < sentinel)
+        snippetCopy(in, out);
+}
+
+template <>
+void copy<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept;

sfizz/SIMDSSE.cpp

@@ -494,4 +494,26 @@ void add<float, true>(absl::Span<const float> input, absl::Span<float> output) n
 
     while (out < sentinel)
         snippetAdd<float>(in, out);
+}
+
+template <>
+void copy<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
+{
+    ASSERT(output.size() >= input.size());
+    auto* in = input.begin();
+    auto* out = output.begin();
+    auto* sentinel = out + min(input.size(), output.size());
+    const auto* lastAligned = prevAligned(sentinel);
+
+    while (unaligned(in, out) && out < lastAligned)
+        snippetCopy<float>(in, out);
+
+    while (out < lastAligned) {
+        _mm_store_ps(out, _mm_load_ps(in));
+        out += TypeAlignment;
+        in += TypeAlignment;
+    }
+
+    while (out < sentinel)
+        snippetCopy<float>(in, out);
 }

tests/SIMDHelpersT.cpp

@@ -622,3 +622,33 @@ TEST_CASE("[Helpers] Add (SIMD vs scalar)")
     add<float, true>(input, absl::MakeSpan(outputSIMD));
     REQUIRE(approxEqual<float>(outputScalar, outputSIMD));
 }
+
+TEST_CASE("[Helpers] copy")
+{
+    std::array<float, 5> input { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f };
+    std::array<float, 5> output { 1.0f, 1.0f, 1.0f, 1.0f, 1.0f };
+    copy<float, false>(input, absl::MakeSpan(output));
+    REQUIRE(output == input);
+}
+
+TEST_CASE("[Helpers] copy (SIMD)")
+{
+    std::array<float, 5> input { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f };
+    std::array<float, 5> output { 1.0f, 1.0f, 1.0f, 1.0f, 1.0f };
+    copy<float, true>(input, absl::MakeSpan(output));
+    REQUIRE(output == input);
+}
+
+TEST_CASE("[Helpers] copy (SIMD vs scalar)")
+{
+    std::vector<float> input(bigBufferSize);
+    std::vector<float> outputScalar(bigBufferSize);
+    std::vector<float> outputSIMD(bigBufferSize);
+    absl::c_iota(input, 0.0);
+    absl::c_fill(outputScalar, 0.0);
+    absl::c_fill(outputSIMD, 0.0);
+
+    add<float, false>(input, absl::MakeSpan(outputScalar));
+    add<float, true>(input, absl::MakeSpan(outputSIMD));
+    REQUIRE(approxEqual<float>(outputScalar, outputSIMD));
+}