From 57a7520545134ab95f7bb0397dcbe991c906a3e9 Mon Sep 17 00:00:00 2001
From: Sebastiano Tronto <sebastiano@tronto.net>
Date: Wed, 6 Aug 2025 16:34:21 +0200
Subject: Added HTR solver

---
 src/arch/avx2.h     | 24 +++++++++++++++++++++++-
 src/arch/common.h   | 39 +++++++++++++++++++++++++++++++++++++++
 src/arch/neon.h     | 23 ++++++++++++++++++++++-
 src/arch/portable.h | 16 ++++++++++++++++
 4 files changed, 100 insertions(+), 2 deletions(-)

(limited to 'src/arch')

diff --git a/src/arch/avx2.h b/src/arch/avx2.h
index 74aad53..1c7248b 100644
--- a/src/arch/avx2.h
+++ b/src/arch/avx2.h
@@ -437,5 +437,27 @@ is_eo_even(cube_t cube)
 	e = _mm256_slli_epi16(e, 7-EOSHIFT);
 	mask = _mm256_movemask_epi8(e);
 
-	return popcount_u32(mask) % 2;
+	return popcount_u32(mask) % 2 == 0;
+}
+
+STATIC_INLINE uint64_t
+coord_epudsep(cube_t cube)
+{
+	uint8_t aux[32];
+
+	_mm256_storeu_si256((__m256i_u *)aux, cube);
+	return coord_epudsep_array(aux + 16);
+}
+
+STATIC_INLINE cube_t
+invcoord_epudsep(uint64_t i)
+{
+	cube_t cube, elow;
+	uint8_t e[32];
+
+	invcoord_epudsep_array(i, e+16);
+	elow = _mm256_load_si256((__m256i *)e);
+	cube = _mm256_set_epi64x(SOLVED_H, 0, 0, SOLVED_L);
+
+	return _mm256_or_si256(elow, cube);
 }
diff --git a/src/arch/common.h b/src/arch/common.h
index d76b65f..f06f8d7 100644
--- a/src/arch/common.h
+++ b/src/arch/common.h
@@ -29,6 +29,8 @@ STATIC_INLINE uint64_t coord_cocsep(cube_t);
 STATIC_INLINE uint64_t coord_eo(cube_t);
 STATIC_INLINE uint64_t coord_esep(cube_t);
 STATIC_INLINE cube_t invcoord_esep(uint64_t);
+STATIC_INLINE uint64_t coord_epudsep(cube_t);
+STATIC_INLINE cube_t invcoord_epudsep(uint64_t);
 
 STATIC_INLINE bool is_eo_even(cube_t);
 
@@ -38,6 +40,8 @@ STATIC_INLINE void set_eo(cube_t [static 1], uint64_t);
 
 STATIC_INLINE void invcoord_esep_array(uint64_t, uint64_t, uint8_t[static 12]);
 STATIC_INLINE cube_t invcoord_eoesep(uint64_t);
+STATIC_INLINE uint64_t coord_epudsep_array(const uint8_t [8]);
+STATIC_INLINE void invcoord_epudsep_array(uint64_t, uint8_t [8]);
 
 STATIC_INLINE uint64_t coord_cp(cube_t);
 STATIC_INLINE cube_t invcoord_cp(uint64_t);
@@ -85,3 +89,38 @@ invcoord_eoesep(uint64_t i)
 
 	return c;
 }
+
+STATIC_INLINE uint64_t
+coord_epudsep_array(const uint8_t e[8])
+{
+	uint8_t i, k, is;
+	uint64_t ret;
+
+	ret = 0;
+	k = 4;
+	for (i = 0; i < 8; i++) {
+		is = (e[i] & UINT8_C(4)) >> UINT8_C(2);
+		ret += is * binomial[7-i][k];
+		k -= is;
+	}
+
+	return ret;
+}
+
+STATIC_INLINE void
+invcoord_epudsep_array(uint64_t c, uint8_t ret[8])
+{
+	uint8_t i, k, is, x, y;
+
+	k = 4;
+	x = 0;
+	y = 4;
+	for (i = 0; i < 8; i++) {
+		is = c >= binomial[7-i][k];
+		ret[i] = is*y + (1-is)*x;
+		y += is;
+		x += 1-is;
+		c -= is * binomial[7-i][k];
+		k -= is;
+	}
+}
diff --git a/src/arch/neon.h b/src/arch/neon.h
index 2c35c58..9871125 100644
--- a/src/arch/neon.h
+++ b/src/arch/neon.h
@@ -453,7 +453,7 @@ invcoord_cp(uint64_t i)
 {
 	return (cube_t) {
 		.corner = indextoperm_8x8(i),
-		.edge = vcombine_u8(vld1_u8(SOLVED_L), vld1_u8(SOLVED_H)) 
+		.edge = vcombine_u8(vld1_u8(SOLVED_L), vld1_u8(SOLVED_H))
 	};
 }
 
@@ -515,3 +515,24 @@ is_eo_even(cube_t cube)
 
 	return count % 2 == 0;
 }
+
+STATIC_INLINE uint64_t
+coord_epudsep(cube_t cube)
+{
+	uint8_t e[8];
+
+	vst1_u8(e, vget_low_u8(cube.edge));
+	return coord_epudsep_array(e);
+}
+
+STATIC_INLINE cube_t
+invcoord_epudsep(uint64_t i)
+{
+	uint8_t e[8];
+
+	invcoord_epudsep_array(i, e);
+	return (cube_t) {
+		.corner = vld1_u8(SOLVED_L),
+		.edge = vcombine_u8(vld1_u8(e), vld1_u8(SOLVED_H))
+	};
+}
diff --git a/src/arch/portable.h b/src/arch/portable.h
index d673f15..c7f2de0 100644
--- a/src/arch/portable.h
+++ b/src/arch/portable.h
@@ -367,3 +367,19 @@ is_eo_even(cube_t cube)
 
 	return count % 2 == 0;
 }
+
+STATIC_INLINE uint64_t
+coord_epudsep(cube_t cube)
+{
+	return coord_epudsep_array(cube.edge);
+}
+
+STATIC_INLINE cube_t
+invcoord_epudsep(uint64_t c)
+{
+	cube_t ret;
+
+	ret = SOLVED_CUBE;
+	invcoord_epudsep_array(c, ret.edge);
+	return ret;
+}
-- 
cgit v1.3