Fix CI comments that name the wrong source of -march=native

.github/workflows/continuous.yml

@@ -74,10 +74,10 @@ jobs:
         uses: SimenB/github-actions-cpu-cores@v2
         id: cpu-cores
 
-      # FindSIMD, MeshFEMCore, and Tight-Inclusion all compile with
-      # -march=native, which bakes the building runner's ISA into every cached
-      # object file. ccache hashes the literal flag string, not what it expands
-      # to, so restored objects are only safe on a runner with the same ISA.
+      # FindSIMD and Tight-Inclusion both compile with -march=native, which
+      # bakes the building runner's ISA into every cached object file. ccache
+      # hashes the literal flag string, not what it expands to, so restored
+      # objects are only safe on a runner with the same ISA.
       # The CPU model name has not been enough to guarantee that: EPYC 9V74
       # runners restored each other's objects and crashed with SIGILL. On Linux
       # we also hash the flags the compiler expands -march=native to, which is

.github/workflows/coverage.yml

@@ -37,10 +37,10 @@ jobs:
         uses: SimenB/github-actions-cpu-cores@v2
         id: cpu-cores
 
-      # FindSIMD, MeshFEMCore, and Tight-Inclusion all compile with
-      # -march=native, which bakes the building runner's ISA into every cached
-      # object file. ccache hashes the literal flag string, not what it expands
-      # to, so restored objects are only safe on a runner with the same ISA.
+      # FindSIMD and Tight-Inclusion both compile with -march=native, which
+      # bakes the building runner's ISA into every cached object file. ccache
+      # hashes the literal flag string, not what it expands to, so restored
+      # objects are only safe on a runner with the same ISA.
       # The CPU model name has not been enough to guarantee that: EPYC 9V74
       # runners restored each other's objects and crashed with SIGILL. We also
       # hash the flags the compiler expands -march=native to, which is exactly

.github/workflows/cuda.yml

@@ -51,10 +51,10 @@ jobs:
         uses: SimenB/github-actions-cpu-cores@v2
         id: cpu-cores
 
-      # FindSIMD, MeshFEMCore, and Tight-Inclusion all compile with
-      # -march=native, which bakes the building runner's ISA into every cached
-      # object file. ccache hashes the literal flag string, not what it expands
-      # to, so restored objects are only safe on a runner with the same ISA.
+      # FindSIMD and Tight-Inclusion both compile with -march=native, which
+      # bakes the building runner's ISA into every cached object file. ccache
+      # hashes the literal flag string, not what it expands to, so restored
+      # objects are only safe on a runner with the same ISA.
       # The CPU model name has not been enough to guarantee that: EPYC 9V74
       # runners restored each other's objects and crashed with SIGILL. We also
       # hash the flags the compiler expands -march=native to, which is exactly

.github/workflows/docs.yml

@@ -90,12 +90,13 @@ jobs:
           done
           cd ..
 
-      # We build ipctk without ccache on purpose. Its dependencies compile with
-      # -march=native, and ccache hashes that flag string, not the CPU it expands
-      # to. A cached object could therefore be restored onto a runner that lacks
-      # some of its instructions, which crashed `import ipctk` with SIGILL when
-      # Sphinx started. The tradeoff is a cold build of about 5 minutes instead of
-      # about 1, which is fine because docs only build on pushes to main.
+      # We build ipctk without ccache on purpose. It compiles with -march=native
+      # (from FindSIMD, plus Tight-Inclusion's own flags), and ccache hashes that
+      # flag string, not the CPU it expands to. A cached object could therefore
+      # be restored onto a runner that lacks some of its instructions, which
+      # crashed `import ipctk` with SIGILL when Sphinx started. The tradeoff is a
+      # cold build of about 5 minutes instead of about 1, which is fine because
+      # docs only build on pushes to main.
       - name: Build ipctk
         run: pip install --verbose .
 

.github/workflows/python.yml

@@ -64,10 +64,10 @@ jobs:
           choco install ccache
           "CACHE_PATH=${env:LOCALAPPDATA}\ccache" | Out-File -FilePath $env:GITHUB_ENV -Append
 
-      # FindSIMD, MeshFEMCore, and Tight-Inclusion all compile with
-      # -march=native, which bakes the building runner's ISA into every cached
-      # object file. ccache hashes the literal flag string, not what it expands
-      # to, so restored objects are only safe on a runner with the same ISA.
+      # FindSIMD and Tight-Inclusion both compile with -march=native, which
+      # bakes the building runner's ISA into every cached object file. ccache
+      # hashes the literal flag string, not what it expands to, so restored
+      # objects are only safe on a runner with the same ISA.
       # The CPU model name has not been enough to guarantee that: EPYC 9V74
       # runners restored each other's objects and crashed with SIGILL. On Linux
       # we also hash the flags the compiler expands -march=native to, which is