Rename RA_DO_CHECK to RA_CHECK

This follows RA_OPT_... and so on.

Internally rename RA_CHECK and RA_FWD to RA_CK and RA_FW respectively.

Author: lloda

.github/workflows/gcc14-no-sanitize.yml

@@ -14,7 +14,7 @@ jobs:
       matrix:
         cxxflags: ['"-O3 -fno-sanitize=all -std=c++23 -DRA_OPT_SMALLVECTOR=0"',
                    '"-O3 -fno-sanitize=all -std=c++23 -DRA_OPT_SMALLVECTOR=1"',
-                   '"-O3 -fno-sanitize=all -DRA_DO_CHECK=0 -DNDEBUG -std=c++23"']
+                   '"-O3 -fno-sanitize=all -DRA_CHECK=0 -DNDEBUG -std=c++23"']
     steps:
     - uses: actions/checkout@v4
     - name: update

.github/workflows/gcc14.yml

@@ -14,7 +14,7 @@ jobs:
       matrix:
         cxxflags: ['"-O3 -std=c++23 -DRA_OPT_SMALLVECTOR=0"',
                    '"-O3 -std=c++23 -DRA_OPT_SMALLVECTOR=1"',
-                   '"-O3 -DRA_DO_CHECK=0 -DNDEBUG -std=c++23"']
+                   '"-O3 -DRA_CHECK=0 -DNDEBUG -std=c++23"']
     steps:
     - uses: actions/checkout@v4
     - name: update

TODO

@@ -147,7 +147,7 @@ Some of these aren't bugs in the sense that I expect to solve them, but more lik
   5. [ ]
   6. [ ]
   7. [ ]
-  8. [X] test/operators.cc: Some simple expressions with scalars fail in VALUE(). The patch in ra.hh
+  8. [X] test/operators.cc: Some simple expressions with scalars fail in VAL(). The patch in ra.hh
      triggers the address sanitizer in test/ra-9.cc.
   9. [ ] test/ra-6.cc: ra::Ptr doesn't hold copies. This enables restarting (see [ra39]), so
      ra::ptr(temp) must only be used as temp. Really forbidding auto ll = ra::ptr(lvalue) would also
@@ -170,7 +170,7 @@ Some of these aren't bugs in the sense that I expect to solve them, but more lik
       (and ravel constructors also).
   17. [ ] assert() is used in some cases for runtime conditions that aren't ra::'s fault. This is
       bad because if RA_ASSERT is defined to throw, the caller probably expects to be able to handle
-      the error. On the other hand if RA_DO_CHECK is 0, we shouldn't continue. But ofc we do in
+      the error. On the other hand if RA_CHECK is 0, we shouldn't continue. But ofc we do in
       index checks etc. So is this different?
   18. [ ]
   19. [ ]

bench/bench-gemm.cc

@@ -25,7 +25,7 @@ void
 gemm1(auto && a, auto && b, auto & c)
 {
     for_each(ra::wrank<1, 2, 1>(ra::wrank<0, 1, 1>([](auto && a, auto && b, auto & c) { ra::maybe_fma(a, b, c); })),
-             RA_FWD(a), RA_FWD(b), RA_FWD(c));
+             RA_FW(a), RA_FW(b), RA_FW(c));
 }
 void
 gemm2(auto && a, auto && b, auto & c)

bench/bench-stencil1.cc

@@ -25,7 +25,7 @@ constexpr ra::Small<real, 3> mask = { 1, -2, 1 };
 #define THEOP template <class A_, class Anext_, class Astencil_> __attribute__((noinline)) \
    auto operator()(A_ & A, Anext_ & Anext, Astencil_ & Astencil)
 
-// sensitive to RA_DO_CHECK.
+// sensitive to RA_CHECK.
 struct f_raw
 {
     THEOP
@@ -37,7 +37,7 @@ struct f_raw
     };
 };
 
-// about as fast as f_raw, but no stencil. Insensitive to RA_DO_CHECK.
+// about as fast as f_raw, but no stencil. Insensitive to RA_CHECK.
 struct f_slices
 {
     THEOP
@@ -47,7 +47,7 @@ struct f_slices
     };
 };
 
-// with stencil, about as fast as f_raw. Sensitive to RA_DO_CHECK.
+// with stencil, about as fast as f_raw. Sensitive to RA_CHECK.
 struct f_stencil_explicit
 {
     THEOP

bench/bench-stencil2.cc

@@ -29,7 +29,7 @@ constexpr ra::Small<real, 3, 3> mask = { 0, 1, 0,
 #define THEOP template <class A_, class Anext_, class Astencil_> __attribute__((noinline)) \
    auto operator()(A_ & A, Anext_ & Anext, Astencil_ & Astencil)
 
-// sensitive to RA_DO_CHECK.
+// sensitive to RA_CHECK.
 struct f_raw
 {
     THEOP
@@ -45,7 +45,7 @@ struct f_raw
     };
 };
 
-// about as fast as f_raw, but no stencil. Insensitive to RA_DO_CHECK.
+// about as fast as f_raw, but no stencil. Insensitive to RA_CHECK.
 struct f_slices
 {
     THEOP
@@ -57,7 +57,7 @@ struct f_slices
     };
 };
 
-// with stencil, about as fast as f_raw. Sensitive to RA_DO_CHECK.
+// with stencil, about as fast as f_raw. Sensitive to RA_CHECK.
 struct f_stencil_explicit
 {
     THEOP

bench/bench-stencil3.cc

@@ -33,7 +33,7 @@ constexpr ra::Small<real, 3, 3, 3> mask = { 0, 0, 0,  0, 1, 0,  0, 0, 0,
 #define THEOP template <class A_, class Anext_, class Astencil_> __attribute__((noinline)) \
    auto operator()(A_ & A, Anext_ & Anext, Astencil_ & Astencil)
 
-// sensitive to RA_DO_CHECK.
+// sensitive to RA_CHECK.
 struct f_raw
 {
     THEOP
@@ -51,7 +51,7 @@ struct f_raw
     };
 };
 
-// about as fast as f_raw, but no stencil. Insensitive to RA_DO_CHECK.
+// about as fast as f_raw, but no stencil. Insensitive to RA_CHECK.
 struct f_slices
 {
     THEOP
@@ -63,7 +63,7 @@ struct f_slices
     };
 };
 
-// with stencil, about as fast as f_raw. Sensitive to RA_DO_CHECK.
+// with stencil, about as fast as f_raw. Sensitive to RA_CHECK.
 struct f_stencil_explicit
 {
     THEOP

box/genfrom.cc

@@ -8,12 +8,11 @@
 // later version.
 
 // Atm view(i ...) only beats i of rank ≤1 and only when all i are beatable. This is a sandbox for a version that beats any combination of scalar or view<Seq> of any rank, and only returns an expr for the unbeatable subscripts.
-// There is an additional issue that eg A2(i1) (indicating ranks) returns a nested expression, that is, a rank-1 expr where each element is a rank-1 view. We'd prefer if the result was rank 2 and not nested, iow, the value type of the result should always be the same as that of A.
+// There is an additional issue that eg A2(unbeatable-i1) (indicating ranks) returns a nested expression, that is, a rank-1 expr where each element is a rank-1 view. It should instead be rank 2 and not nested, iow, the value type of the result should always be the same as that of A.
+// 1) Need to split the static and the dynamic parts so some/most of the routine can be used for Small or Big
+// 2) Need to find a way to use len like in Ptr-based iota.
 
 #include "ra/test.hh"
-#include <iomanip>
-#include <chrono>
-#include <span>
 
 using std::cout, std::endl, std::flush;
 
@@ -25,6 +24,25 @@ constexpr auto ii(dim_t (&&s)[rank])
     return ViewBig<Seq<dim_t>, rank>(s, Seq<dim_t>{0});
 }
 
+template <std::integral auto ... i>
+constexpr auto ii(ra::ilist_t<i ...>)
+{
+    return ViewSmall<Seq<dim_t>, ra::ic_t<ra::default_dims(std::array<ra::dim_t, sizeof...(i)>{i...})>>(Seq<dim_t>{0});
+}
+
+template <class A> concept is_iota_static = requires (A a)
+{
+    []<class I, class Dimv>(ViewSmall<Seq<I>, Dimv> const &){}(a);
+    requires UNB!=ra::size(a); // exclude UNB from beating to let B=A(... i ...) use B's len. FIXME
+};
+
+template <class A> concept is_iota_dynamic = requires (A a)
+{
+    []<class I, rank_t RANK>(ViewBig<Seq<I>, RANK> const &){}(a);
+};
+
+template <class A> concept is_iota_any = is_iota_dynamic<A> || is_iota_static<A>;
+
 template <class A, class ... I>
 constexpr auto
 genfrom(A && a, I && ...  i)
@@ -39,6 +57,6 @@ int main()
 {
     ra::TestRecorder tr(std::cout);
     cout << ra::ii({3, 4}) << endl;
+    cout << ra::ii(ra::ilist<3, 4>) << endl;
     return tr.summary();
 }
-x

docs/index.html

@@ -314,7 +314,7 @@ The following @code{#define}s affect the behavior of @code{ra::}.
 
 @itemize
 @c FIXME The flag should only apply to dynamic checks.
-@item @code{RA_DO_CHECK} (default 1)
+@item @code{RA_CHECK} (default 1)
 
 If 1, check shape agreement (e.g. @code{Big<int, 1> @{2, 3@} + Big<int, 1> @{1, 2, 3@}}) and random array accesses (e.g. @code{Small<int, 2> a = 0; int i = 10; a[i] = 0;}). See @ref{Error handling}.
 
@@ -1761,26 +1761,26 @@ Runtime error handling in @code{ra::} is controlled by two macros.
 @itemize
 @item @code{RA_ASSERT(cond, ...)} ---
 check that @code{cond} evaluates to true in the @code{ra::} namespace. The other arguments are informative.
-@item @code{RA_DO_CHECK} ---
+@item @code{RA_CHECK} ---
 must have one of the values 0, 1, or 2.
 @end itemize
 
 They work as follows:
 
 @itemize
-@item If @code{RA_DO_CHECK} is 0, runtime checks are skipped.
-@item If @code{RA_DO_CHECK} is not 0, runtime checks are done.
+@item If @code{RA_CHECK} is 0, runtime checks are skipped.
+@item If @code{RA_CHECK} is not 0, runtime checks are done.
     @itemize
     @item If @code{RA_ASSERT} is defined, using @code{RA_ASSERT}.
-    @item If @code{RA_ASSERT} isn't defined, the method depends on the value of @code{RA_DO_CHECK}. The two options are 1 (plain @code{assert}) and 2 (prints the informative arguments and aborts). Other values are an error.
+    @item If @code{RA_ASSERT} isn't defined, the method depends on the value of @code{RA_CHECK}. The two options are 1 (plain @code{assert}) and 2 (prints the informative arguments and aborts). Other values are an error.
     @end itemize
 @end itemize
 
-@code{ra::} contains uses of @code{assert} for checking invariants or for sanity checks that are separate from uses of @code{RA_ASSERT}. Those can be disabled in the usual way with @option{-DNDEBUG}, but note that @option{-DNDEBUG} will also disable any @code{assert}s that are a result of @code{RA_DO_CHECK=1}.
+@code{ra::} contains uses of @code{assert} for checking invariants or for sanity checks that are separate from uses of @code{RA_ASSERT}. Those can be disabled in the usual way with @option{-DNDEBUG}, but note that @option{-DNDEBUG} will also disable any @code{assert}s that are a result of @code{RA_CHECK=1}.
 
-The performance cost of the runtime checks depends on the program. Without custom @code{RA_ASSERT}, @code{RA_DO_CHECK=1} usually has an acceptable cost, but @code{RA_DO_CHECK=2} is usually more expensive, because having the arguments around to print can prevent optimization. The default is @code{RA_DO_CHECK=1}.
+The performance cost of the runtime checks depends on the program. Without custom @code{RA_ASSERT}, @code{RA_CHECK=1} usually has an acceptable cost, but @code{RA_CHECK=2} is usually more expensive, because having the arguments around to print can prevent optimization. The default is @code{RA_CHECK=1}.
 
-The following example shows how errors might be reported depending on @code{RA_DO_CHECK}.
+The following example shows how errors might be reported depending on @code{RA_CHECK}.
 
 @cartouche
 @verbatim
@@ -1792,11 +1792,11 @@ cout << (a+b) << endl;
 
 @c FIXME we'd get 'Bad shapes (-1922222222 3)' on sum(a+b) bc of the check only at top strategy. Improve that.
 @itemize
-@item @code{RA_DO_CHECK=2}
+@item @code{RA_CHECK=2}
 @verbatim
 *** ra::./ra/expr.hh:436,13 (check()) Bad shapes (10 3)(40 3). ***
 @end verbatim
-@item @code{RA_DO_CHECK=1}
+@item @code{RA_CHECK=1}
 @verbatim
 agreement: ./ra/expr.hh:436: constexpr void ra::Match<std::tuple<_Ts ...>, std::tuple<std::integral_constant<int, I>...> >::validate() const [with P = {ra::Cell<int, const ra::
 SmallArray<ra::Dim, std::tuple<std::integral_constant<int, 2> >, std::tuple<std::integral_constant<int, 1> >, std::tuple<ra::Dim, ra::Dim> >&, std::integral_constant<int, 0> >,
@@ -1924,7 +1924,7 @@ Item 1
 Item 2
 @end enumerate
 
-@section The cost of @code{RA_DO_CHECK}
+@section The cost of @code{RA_CHECK}
 
 FIXME