Adopt the DspTap FIR substrate via submodule

.github/workflows/book-pages.yml

@@ -43,6 +43,8 @@ jobs:
       url: ${{ steps.deployment.outputs.page_url }}
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Install Doxygen + Graphviz
         run: sudo apt-get update && sudo apt-get install -y doxygen graphviz

.github/workflows/ci-arm64.yml

@@ -26,6 +26,8 @@ jobs:
     timeout-minutes: 30
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Configure (Release)
         run: cmake -B build -DCMAKE_BUILD_TYPE=Release -DSRT_WERROR=ON

.github/workflows/ci.yml

@@ -39,6 +39,8 @@ jobs:
             werror: OFF
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Configure
         env:
@@ -68,6 +70,8 @@ jobs:
             flags: -fsanitize=thread
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Configure
         env:
@@ -109,6 +113,8 @@ jobs:
       HEXAGON_TOOLCHAIN_SHA256: "55b41922318f6331590ab7baa7f5dbdd99c109327a9c44a52c5e9878fab148c1"
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Cache toolchain
         id: cache
@@ -193,6 +199,8 @@ jobs:
     timeout-minutes: 30
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Install toolchain and QEMU
         run: >
@@ -222,6 +230,8 @@ jobs:
     timeout-minutes: 30
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Install toolchain and QEMU
         run: >
@@ -294,6 +304,8 @@ jobs:
       HEXAGON_TOOLCHAIN_SHA256: "55b41922318f6331590ab7baa7f5dbdd99c109327a9c44a52c5e9878fab148c1"
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Install toolchains and QEMU
         run: >
@@ -440,6 +452,8 @@ jobs:
     timeout-minutes: 15
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Configure
         run: >
@@ -464,6 +478,8 @@ jobs:
     timeout-minutes: 20
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Install dependencies
         run: >
@@ -493,6 +509,8 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
       # Run the TapHouse pre-commit hook — the same pinned clang-format
       # developers get from `pre-commit install` (.pre-commit-config.yaml), so
       # local and CI share one version AND one file scope (all C/C++ minus
@@ -515,6 +533,8 @@ jobs:
       MDBOOK_SHA256: "9ef07fd288ba58ff3b99d1c94e6d414d431c9a61fdb20348e5beb74b823d546b"
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Install mdBook (pinned)
         run: |

.github/workflows/compare.yml

@@ -26,6 +26,8 @@ jobs:
       HEXAGON_TOOLCHAIN_SHA256: "55b41922318f6331590ab7baa7f5dbdd99c109327a9c44a52c5e9878fab148c1"
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
+        with:
+          submodules: recursive
 
       - name: Install toolchains and QEMU
         run: >

.github/workflows/style.yml

@@ -16,6 +16,8 @@ jobs:
     timeout-minutes: 30
     steps:
       - uses: actions/checkout@v4
+        with:
+          submodules: recursive
       - name: Install tools
         run: sudo apt-get update && sudo apt-get install -y clang-tidy-18 libgtest-dev cmake python3
       - name: Configure (compile database)

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "submodules/dsptap"]
+	path = submodules/dsptap
+	url = https://github.com/tap/dsptap

CMakeLists.txt

@@ -1,12 +1,19 @@
 cmake_minimum_required(VERSION 3.24)
 project(SampleRateTap VERSION 0.1.0 LANGUAGES CXX)
 
+# Shared Tap-family substrate (tap::dsp): kaiser design math, sample-format
+# traits, FIR dot kernels, row-sum quantization, measurement instruments.
+# Pinned as a submodule per the DspTap release flow (changes land there first,
+# consumers bump the pin).
+add_subdirectory(submodules/dsptap)
+
 add_library(SampleRateTap INTERFACE)
 add_library(SampleRateTap::SampleRateTap ALIAS SampleRateTap)
 target_include_directories(SampleRateTap INTERFACE
     $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>
     $<INSTALL_INTERFACE:include>)
 target_compile_features(SampleRateTap INTERFACE cxx_std_20)
+target_link_libraries(SampleRateTap INTERFACE tap::dsp)
 
 if(PROJECT_IS_TOP_LEVEL)
     option(SRT_BUILD_TESTS "Build SampleRateTap tests" ON)

book/src/epilogue/letter.md

@@ -87,7 +87,7 @@ design target by 1/sinc(f/fs) so the composite comes out flat *and*
 zeroed:
 
 ```cpp
-{{#include ../../../include/srt/detail/kaiser.h:pw_comp_design}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/kaiser.h:pw_comp_design}}
 ```
 
 Two implementation details earned their scars. The cosine-series trick is
@@ -98,7 +98,7 @@ rect is applied as a running sum *after* windowing, which makes the zeros
 exact in the discrete composite:
 
 ```cpp
-{{#include ../../../include/srt/detail/kaiser.h:pw_comp_rect}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/kaiser.h:pw_comp_rect}}
 ```
 
 One bug in between is worth its paragraph, because the test suite caught
@@ -158,14 +158,14 @@ log-spaced 60 Hz–16 kHz, amplitudes pink (equal energy per octave — the
 long-run average of real program), phases chosen for bounded crest:
 
 ```cpp
-{{#include ../../../tests/support/multitone_analysis.h:pw_comb}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/analysis/multitone_analysis.h:pw_comb}}
 ```
 
 The measurement fits every tone jointly and calls everything left over —
 aliases, servo FM, noise — the residual:
 
 ```cpp
-{{#include ../../../tests/support/multitone_analysis.h:pw_metric}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/analysis/multitone_analysis.h:pw_metric}}
 ```
 
 The instrument fought back twice, and both fights are preserved in its
@@ -249,7 +249,7 @@ over with a widened tolerance earlier in this very chapter's story. The
 tricky thing is row-sum-preserving quantization:
 
 ```cpp
-{{#include ../../../include/srt/polyphase_filter.h:pw_row_sum}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/quantize.h:pw_row_sum}}
 ```
 
 After it, the measured result is the one his claim demanded: a DC input

book/src/part1/fractional-resampler.md

@@ -72,7 +72,7 @@ The fix is to split the kernel at its natural seam: blend once per frame
 into a scratch row, then run a plain dot product per channel.
 
 ```cpp
-{{#include ../../../include/srt/polyphase_filter.h:rs_dot_row}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/fir_kernels.h:rs_dot_row}}
 ```
 
 Two things about this function beyond its arithmetic. First, the comment
@@ -326,7 +326,7 @@ history to deliver all channels of tap t contiguously — the frame-major
 layout — and a register-blocked kernel:
 
 ```cpp
-{{#include ../../../include/srt/polyphase_filter.h:rs_dot_rows_frame_major}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/fir_kernels.h:rs_dot_rows_frame_major}}
 ```
 
 The measured C6 results, condensed (the full campaign, including the

book/src/part1/kaiser.md

@@ -35,7 +35,7 @@ interpolates exactly. The `sinc` in this file is that function, with the
 one hazard a numeric programmer would expect handled explicitly:
 
 ```cpp
-{{#include ../../../include/srt/detail/kaiser.h:kai_sinc}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/kaiser.h:kai_sinc}}
 ```
 
 (The 0/0 at x = 0 is a *removable* singularity — the limit is 1 — but IEEE
@@ -114,7 +114,7 @@ I₀(x) = Σₖ [ (x/2)ᵏ / k! ]²
 which converges for every finite x:
 
 ```cpp
-{{#include ../../../include/srt/detail/kaiser.h:kai_besseli0}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/kaiser.h:kai_besseli0}}
 ```
 
 Three details carry all the engineering.
@@ -155,7 +155,7 @@ constant *relative* accuracy, which is what the window formula's ratio
 ## `kaiser_beta`: an empirical fit, taken as published
 
 ```cpp
-{{#include ../../../include/srt/detail/kaiser.h:kai_beta}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/kaiser.h:kai_beta}}
 ```
 
 This is Kaiser's published fit, digit for digit — `0.1102`, `0.5842`,
@@ -182,7 +182,7 @@ harris) says taps scale linearly with attenuation and inversely with
 transition width:
 
 ```cpp
-{{#include ../../../include/srt/detail/kaiser.h:kai_estimate}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/kaiser.h:kai_estimate}}
 ```
 
 Note what the signature normalizes to: transition width *as a fraction of
@@ -222,7 +222,7 @@ UB, however silly the input.
 ## `design_prototype`: where all of it lands
 
 ```cpp
-{{#include ../../../include/srt/detail/kaiser.h:kai_prototype}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/kaiser.h:kai_prototype}}
 ```
 
 One pass, one output array, but four decisions are packed into these lines.
@@ -292,7 +292,7 @@ was rejected, and since the reasoning is a design artifact it is kept where
 refactors will trip over it:
 
 ```cpp
-{{#include ../../../include/srt/detail/kaiser.h:kai_design_note}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/kaiser.h:kai_design_note}}
 ```
 
 Present the alternative fairly, because it *almost* works:

book/src/part1/sample-traits.md

@@ -156,15 +156,15 @@ precisely at the row where the response matters most.
 So the coefficients trade one precision bit for one headroom bit:
 
 ```cpp
-{{#include ../../../include/srt/sample_traits.h:st_q15_coeff}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/sample_traits.h:st_q15_coeff}}
 ```
 
 with the conversion doing round-half-away-from-zero and saturating at the
 integer limits (the *design* is checked separately; saturation here is a
 belt against future filter specs, not an expected event):
 
 ```cpp
-{{#include ../../../include/srt/sample_traits.h:st_roundsat}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/sample_traits.h:st_roundsat}}
 ```
 
 What did the traded bit cost? Quantizing coefficients to Q14 puts the
@@ -190,7 +190,7 @@ and that is their virtue: `Q15::make_coeff(1.0) == 16384` is the sentence
 Here is the Q15 multiply-accumulate:
 
 ```cpp
-{{#include ../../../include/srt/sample_traits.h:st_q15_mac}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/sample_traits.h:st_q15_mac}}
 ```
 
 Two things are chosen here. The product is computed in `int32_t` — a
@@ -218,7 +218,7 @@ blended-row rewrite could both be verified *bit-exact* rather than
 All of the rounding budget is spent in one place:
 
 ```cpp
-{{#include ../../../include/srt/sample_traits.h:st_q15_finalize}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/sample_traits.h:st_q15_finalize}}
 ```
 
 The accumulator holds a Q29 value (Q0.15 sample × Q1.14 coefficient); the
@@ -256,7 +256,7 @@ accumulator worth having on the targets this path exists for.
 So each product gives up 16 bits *before* joining the sum:
 
 ```cpp
-{{#include ../../../include/srt/sample_traits.h:st_q31_mac}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/sample_traits.h:st_q31_mac}}
 ```
 
 Now redo the bound: Q45 products have worst-case magnitude 2⁴⁵, and

book/src/part3/c3-c5.md

@@ -172,7 +172,7 @@ The subtle part of C4 is not the intrinsic; it is the **gate**. Here is
 the actual block from `include/srt/polyphase_filter.h`, pulled in live:
 
 ```cpp
-{{#include ../../../include/srt/polyphase_filter.h:opt_smlald_gate}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/fir_kernels.h:opt_smlald_gate}}
 ```
 
 Read the condition: DSP extension present *and MVE absent*. The naive gate

book/src/part3/c6.md

@@ -145,7 +145,7 @@ demonstrably keeps in registers. The library's tile is a template on K,
 taken live from `include/srt/polyphase_filter.h`:
 
 ```cpp
-{{#include ../../../include/srt/polyphase_filter.h:opt_dot_tile}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/fir_kernels.h:opt_dot_tile}}
 ```
 
 The doc comment carries the 2.8× number in the header itself — the trap
@@ -181,7 +181,7 @@ cascade of tiles — 8, then 4, then 2, then 1 — each a specialization of
 the same template:
 
 ```cpp
-{{#include ../../../include/srt/polyphase_filter.h:opt_dot_rows}}
+{{#include ../../../submodules/dsptap/include/tap/dsp/fir_kernels.h:opt_dot_rows}}
 ```
 
 Twelve channels is 8+4. Sixteen is 8+8. The deployment shapes exercise