h48

commit 3051b2342d67dcba2fdc0010fb26f5850964c4d9
parent ade2ed050a2d472b4dd01e30666f5e7ec5030724
Author: Sebastiano Tronto <sebastiano@tronto.net>
Date:   Sat,  4 Nov 2023 10:25:13 +0100

avx2 fully works

Diffstat:
M
README.md
 | 
4
----
M
src/_trans_avx2.c
 | 
13
+++++++------
M
src/cube.c
 | 
21
+++++++++++++--------

3 files changed, 20 insertions(+), 18 deletions(-)
diff --git a/README.md b/README.md
@@ -13,10 +13,6 @@ $ make test
 
 ## TODO:
 
-### Make AVX2 work
-
-* fix inverse, flipallcorners
-
 ### Cleanup / refactor
 
 * see planner
diff --git a/src/_trans_avx2.c b/src/_trans_avx2.c
@@ -1,14 +1,15 @@
 static inline cube_t
 flipallcorners(cube_t c)
 {
-	cube_t shleft, shright, summed, newco, cleanco, ret;
+	cube_t co, shleft, shright, summed, newco, cleanco, ret;
 
-	shleft = _mm256_slli_si256(c, 1);
-	shright = _mm256_srli_si256(c, 1);
+	co = _mm256_and_si256(c, _co2_avx2);
+	shleft = _mm256_slli_epi32(co, 1);
+	shright = _mm256_srli_epi32(co, 1);
 	summed = _mm256_or_si256(shleft, shright);
-	newco = _mm256_and_si256(summed, _co_avx2);
-	cleanco = _mm256_andnot_si256(c, _co_avx2);
-	ret = _mm256_and_si256(cleanco, newco);
+	newco = _mm256_and_si256(summed, _co2_avx2);
+	cleanco = _mm256_xor_si256(c, co);
+	ret = _mm256_or_si256(cleanco, newco);
 
 	return ret;
 }
diff --git a/src/cube.c b/src/cube.c
@@ -130,10 +130,14 @@ cube_arr_t zerocube_arr = { .e = {0}, .c = {0} };
 #define _co_avx2 _mm256_set_epi8( \
     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, \
     0, 0, 0, 0, 0, 0, 0, 0, \
-    0x70, 0x70, 0x70, 0x70, 0x70, 0x70, 0x70, 0x70) 
+    0xF0, 0xF0, 0xF0, 0xF0, 0xF0, 0xF0, 0xF0, 0xF0) 
+#define _co2_avx2 _mm256_set_epi8( \
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, \
+    0, 0, 0, 0, 0, 0, 0, 0, \
+    0x60, 0x60, 0x60, 0x60, 0x60, 0x60, 0x60, 0x60) 
 #define _eo_avx2 _mm256_set_epi8( \
-    0, 0, 0, 0, 0x70, 0x70, 0x70, 0x70, \
-    0x70, 0x70, 0x70, 0x70, 0x70, 0x70, 0x70, 0x70, \
+    0, 0, 0, 0, 0x10, 0x10, 0x10, 0x10, \
+    0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, \
     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
 #define setsolved(cube) cube = _mm256_loadu_si256((__m256i_u *)&solvedcube_arr)
 #define setzero(cube)   cube = _mm256_setzero_si256()
@@ -959,6 +963,7 @@ inverse(cube_t c)
 	vi = _mm256_shuffle_epi8(vi, vi);
 	vi = _mm256_shuffle_epi8(vi, vi);
 	vi = _mm256_shuffle_epi8(vi, vi);
+	vi = _mm256_shuffle_epi8(vi, vi);
 	vi = _mm256_shuffle_epi8(vi, c);
 	vi = _mm256_shuffle_epi8(vi, vi);
 	vi = _mm256_shuffle_epi8(vi, vi);
@@ -969,9 +974,9 @@ inverse(cube_t c)
 	vi = _mm256_shuffle_epi8(vi, vi);
 	vi = _mm256_shuffle_epi8(vi, c);
 
-	vo = _mm256_and_si256(c, _mm256_or_si256(_eo_avx2, _co_avx2));
+	vo = _mm256_and_si256(c, _mm256_or_si256(_eo_avx2, _co2_avx2));
 	vo = _mm256_shuffle_epi8(vo, vi);
-	vp = _mm256_andnot_si256(_mm256_or_si256(_eo_avx2, _co_avx2), vi);
+	vp = _mm256_andnot_si256(_mm256_or_si256(_eo_avx2, _co2_avx2), vi);
 	ret = _mm256_or_si256(vp, vo);
 	
 	return flipallcorners(ret);
@@ -1016,8 +1021,8 @@ inline_compose(cube_t c1, cube_t c2)
 	eo2 = _mm256_and_si256(c2, _eo_avx2);
 	s = _mm256_shuffle_epi8(c1, c2);
 	ed = _mm256_xor_si256(s, eo2);
-	co1 = _mm256_and_si256(s, _co_avx2);
-	co2 = _mm256_and_si256(c2, _co_avx2);
+	co1 = _mm256_and_si256(s, _co2_avx2);
+	co2 = _mm256_and_si256(c2, _co2_avx2);
 	aux = _mm256_add_epi8(co1, co2);
 	cw = _mm256_set_epi8(
 	    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
@@ -1033,7 +1038,7 @@ inline_compose(cube_t c1, cube_t c2)
 	    0x60, 0x60, 0x60, 0x60, 0x60, 0x60, 0x60, 0x60
 	);
 	auz2 = _mm256_and_si256(auz1, cwccw);
-	coclean = _mm256_andnot_si256(_co_avx2, ed);
+	coclean = _mm256_andnot_si256(_co2_avx2, ed);
 	ret = _mm256_or_si256(coclean, auz2);
 #else
 	uint8_t i, piece1, piece2, p, orien, aux, auy;