h48

commit 861a8a5fc4b62e5f70081f8e72b0854889a6cec6
parent 8c281789799a3954e94c7eeb10fb09e526005e1f
Author: Sebastiano Tronto <sebastiano@tronto.net>
Date:   Tue, 26 Dec 2023 16:46:01 +0100

Merge branch 'master' of tronto.net:h48

Diffstat:
M
configure.sh
 | 
9
++++++---
M
cube.c
 | 
19
++++++++++++++++---

2 files changed, 22 insertions(+), 6 deletions(-)
diff --git a/configure.sh b/configure.sh
@@ -1,9 +1,12 @@
 #!/bin/sh
 
-detectarch() { cc -march=native -dM -E - </dev/null | grep "$1"; }
-detectsan() { cc -fsanitize=$1 -dM -E -x c - </dev/null | grep "SANITIZE"; }
+cc=${CC:-cc}
+
+detectarch() { ${cc} -march=native -dM -E - </dev/null | grep "$1"; }
+detectsan() { ${cc} -fsanitize=$1 -dM -E -x c - </dev/null | grep "SANITIZE"; }
 
 [ -n "$(detectarch __AVX2__)" ] && detected="AVX2"
+[ -n "$(detectarch __ARM_NEON)" ] && detected="NEON"
 
 TYPE=${TYPE-"$detected"}
 
@@ -26,5 +29,5 @@ echo "";
 echo "CFLAGS = $CFLAGS";
 echo "DBGFLAGS = $DBGFLAGS";
 echo "";
-echo "CC = ${CC:-cc}"
+echo "CC = ${cc}"
 } > config.mk
diff --git a/cube.c b/cube.c
@@ -485,11 +485,9 @@ Section: AVX2 fast methods
 
 This section contains performance-critical methods that rely on AVX2
 intructions such as routines for moving or transforming the cube.
-
-Note: the #ifdef below is closed in the next section.
 ******************************************************************************/
 
-#ifdef CUBE_AVX2
+#if defined(CUBE_AVX2)
 
 #include <immintrin.h>
 
@@ -651,6 +649,21 @@ coord_fast_eo(cube_fast_t c)
 	return mask >> 17;
 }
 
+/******************************************************************************
+Section: ARM NEON fast methods
+
+This section contains performance-critical methods that rely on ARM NEON
+intructions such as routines for moving or transforming the cube.
+******************************************************************************/
+
+#elif defined(CUBE_NEON)
+
+typedef struct {
+	uint8x16_t corner;
+	uint8x16_t edge;
+} cube_fast_t;
+
+/* TODO! */
 
 /******************************************************************************
 Section: portable fast methods