@@ -198,0 +81,4 @@
+*
+* On Arm, the following additional optimized functions are also provided when using `pico_double_pico`, all of which
+* saturate to the nearest representable value for too large input when converting from floating point types:
+*

@@ -198,0 +129,4 @@
+
+#include "pico.h"
+
+// PICO_CONFIG: PICO_DOUBLE_IN_RAM, Force placement of SDK provided double-precision floating point code into RAM, type=bool, default=0, group=pico_double

@@ -0,0 +1,1342 @@
+#include <stdio.h>

@@ -0,0 +309,4 @@
+    return cycle_diff(t0, t1) - INT_INPUT_COST - DOUBLE_OUTPUT_COST;
+}
+
+timer_func_def(ui2d)(volatile int32_t i) {

@@ -0,0 +1,1342 @@
+#include <stdio.h>

@@ -0,0 +757,4 @@
+#endif
+}
+
+timer_func_def(fix642double_c)(volatile int64_t i) {

@@ -0,0 +1,1286 @@
+#include <stdio.h>

@@ -198,0 +129,4 @@
+
+#include "pico.h"
+
+// PICO_CONFIG: PICO_DOUBLE_IN_RAM, Force placement of SDK provided double-precision floating point code into RAM, type=bool, default=0, group=pico_double

@@ -0,0 +309,4 @@
+    return cycle_diff(t0, t1) - INT_INPUT_COST - DOUBLE_OUTPUT_COST;
+}
+
+timer_func_def(ui2d)(volatile int32_t i) {

@@ -0,0 +757,4 @@
+#endif
+}
+
+timer_func_def(fix642double_c)(volatile int64_t i) {

@@ -0,0 +1,1286 @@
+#include <stdio.h>

@@ -198,0 +129,4 @@
+
+#include "pico.h"
+
+// PICO_CONFIG: PICO_DOUBLE_IN_RAM, Force placement of SDK provided double-precision floating point code into RAM, type=bool, default=0, group=pico_double

@@ -198,0 +81,4 @@
+*
+* On Arm, the following additional optimized functions are also provided when using `pico_double_pico`, all of which
+* saturate to the nearest representable value for too large input when converting from floating point types:
+*

@@ -18,22 +18,23 @@
 #define test_checki64(x, expected, msg) ({ if ((x) != (expected)) { printf("  %s: %lld != %lld\n", msg, (int64_t)(x), (int64_t)(expected)); stop(); } })

@@ -18,22 +18,23 @@
 #define test_checki64(x, expected, msg) ({ if ((x) != (expected)) { printf("  %s: %lld != %lld\n", msg, (int64_t)(x), (int64_t)(expected)); stop(); } })

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -135,75 +127,238 @@ extern "C" {
 *

@@ -135,75 +127,238 @@ extern "C" {
 *

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -135,75 +127,238 @@ extern "C" {
 *

@@ -135,75 +127,238 @@ extern "C" {
 *

@@ -135,75 +127,238 @@ extern "C" {
 *

@@ -442,10 +442,9 @@ int main() {
     for (double a = -100.0; a < 100.0; a += 53.103) {

@@ -18,22 +18,23 @@
 #define test_checki64(x, expected, msg) ({ if ((x) != (expected)) { printf("  %s: %lld != %lld\n", msg, (int64_t)(x), (int64_t)(expected)); stop(); } })

@@ -442,10 +442,9 @@ int main() {
     for (double a = -100.0; a < 100.0; a += 53.103) {

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -195,3 +1,1 @@
-#endif
-
-#endif
+/*

@@ -289,36 +445,53 @@ uint64_t float2ufix64(float f, int e);
 })

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -195,3 +1,1 @@
-#endif
-
-#endif
+/*

@@ -198,0 +205,4 @@
+#if !__PICO_DOUBLE_ARM_OPTIMIZED
+    // for non Arm-optimized we may as well provide the function and let the compiler handle it
+    static inline double int2double(int32_t i) { return (double)i; }
+    static inline double uint2double(uint32_t i) { return (double)i; }

@@ -195,3 +1,1 @@
-#endif
-
-#endif
+/*

@@ -0,0 +223,4 @@
+#if PICO_FLOAT_HAS_SQRTF_FAST
+    register io_ro_32 *systick_ptr = systick_value_ptr();
+    uint32_t t0 = *systick_ptr;
+    volatile float x = sqrtf_fast(a);

@@ -195,3 +1,1 @@
-#endif
-
-#endif
+/*

@@ -198,0 +123,4 @@
+*
+*   mla/fma_fast
+*
+* On RISC-V there is no custom double-precision floating point support, so `pico_double_pico` is equivalent to `pico_double_compiler`

@@ -198,0 +205,4 @@
+#if !__PICO_DOUBLE_ARM_OPTIMIZED
+    // for non Arm-optimized we may as well provide the function and let the compiler handle it
+    static inline double int2double(int32_t i) { return (double)i; }
+    static inline double uint2double(uint32_t i) { return (double)i; }

@@ -0,0 +223,4 @@
+#if PICO_FLOAT_HAS_SQRTF_FAST
+    register io_ro_32 *systick_ptr = systick_value_ptr();
+    uint32_t t0 = *systick_ptr;
+    volatile float x = sqrtf_fast(a);