On 4/26/23 04:30, Taylor Simpson wrote:
The following instructions are tested
V6_vasrvuhubrndsat
V6_vasrvuhubsat
V6_vasrvwuhrndsat
V6_vasrvwuhsat
V6_vassign_tmp
V6_vcombine_tmp
V6_vmpyuhvs
Signed-off-by: Taylor Simpson <tsimp...@quicinc.com>
---
tests/tcg/hexagon/v69_hvx.c | 318 ++++++++++++++++++++++++++++++
tests/tcg/hexagon/Makefile.target | 3 +
2 files changed, 321 insertions(+)
create mode 100644 tests/tcg/hexagon/v69_hvx.c
diff --git a/tests/tcg/hexagon/v69_hvx.c b/tests/tcg/hexagon/v69_hvx.c
new file mode 100644
index 0000000000..051e5420df
--- /dev/null
+++ b/tests/tcg/hexagon/v69_hvx.c
@@ -0,0 +1,318 @@
+/*
+ * Copyright(c) 2023 Qualcomm Innovation Center, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <stdio.h>
+#include <stdint.h>
+#include <stdbool.h>
+#include <string.h>
+#include <limits.h>
+
+int err;
+
+#include "hvx_misc.h"
+
+#define fVROUND(VAL, SHAMT) \
+ ((VAL) + (((SHAMT) > 0) ? (1LL << ((SHAMT) - 1)) : 0))
+
+#define fVSATUB(VAL) \
+ ((((VAL) & 0xffLL) == (VAL)) ? \
+ (VAL) : \
+ ((((int32_t)(VAL)) < 0) ? 0 : 0xff))
+
+#define fVSATUH(VAL) \
+ ((((VAL) & 0xffffLL) == (VAL)) ? \
+ (VAL) : \
+ ((((int32_t)(VAL)) < 0) ? 0 : 0xffff))
+
+static void test_vasrvuhubrndsat(void)
+{
+ void *p0 = buffer0;
+ void *p1 = buffer1;
+ void *pout = output;
+
+ memset(expect, 0xaa, sizeof(expect));
+ memset(output, 0xbb, sizeof(output));
+
+ for (int i = 0; i < BUFSIZE / 2; i++) {
+ asm("v4 = vmem(%0 + #0)\n\t"
+ "v5 = vmem(%0 + #1)\n\t"
+ "v6 = vmem(%1 + #0)\n\t"
+ "v5.ub = vasr(v5:4.uh, v6.ub):rnd:sat\n\t"
+ "vmem(%2) = v5\n\t"
+ : : "r"(p0), "r"(p1), "r"(pout)
+ : "v4", "v5", "v6", "memory");
+ p0 += sizeof(MMVector) * 2;
+ p1 += sizeof(MMVector);
+ pout += sizeof(MMVector);
+
+ for (int j = 0; j < MAX_VEC_SIZE_BYTES / 2; j++) {
+ int shamt;
+ uint8_t byte0;
+ uint8_t byte1;
+
+ shamt = buffer1[i].ub[2 * j + 0] & 0x7;
+ byte0 = fVSATUB(fVROUND(buffer0[2 * i + 0].uh[j], shamt) >> shamt);
+ shamt = buffer1[i].ub[2 * j + 1] & 0x7;
+ byte1 = fVSATUB(fVROUND(buffer0[2 * i + 1].uh[j], shamt) >> shamt);
+ expect[i].uh[j] = (byte1 << 8) | (byte0 & 0xff);
+ }
+ }
+
+ check_output_h(__LINE__, BUFSIZE / 2);
+}
+
+static void test_vasrvuhubsat(void)
+{
+ void *p0 = buffer0;
+ void *p1 = buffer1;
+ void *pout = output;
+
+ memset(expect, 0xaa, sizeof(expect));
+ memset(output, 0xbb, sizeof(output));
+
+ for (int i = 0; i < BUFSIZE / 2; i++) {
+ asm("v4 = vmem(%0 + #0)\n\t"
+ "v5 = vmem(%0 + #1)\n\t"
+ "v6 = vmem(%1 + #0)\n\t"
+ "v5.ub = vasr(v5:4.uh, v6.ub):sat\n\t"
+ "vmem(%2) = v5\n\t"
+ : : "r"(p0), "r"(p1), "r"(pout)
+ : "v4", "v5", "v6", "memory");
+ p0 += sizeof(MMVector) * 2;
+ p1 += sizeof(MMVector);
+ pout += sizeof(MMVector);
+
+ for (int j = 0; j < MAX_VEC_SIZE_BYTES / 2; j++) {
+ int shamt;
+ uint8_t byte0;
+ uint8_t byte1;
+
+ shamt = buffer1[i].ub[2 * j + 0] & 0x7;
+ byte0 = fVSATUB(buffer0[2 * i + 0].uh[j] >> shamt);
+ shamt = buffer1[i].ub[2 * j + 1] & 0x7;
+ byte1 = fVSATUB(buffer0[2 * i + 1].uh[j] >> shamt);
+ expect[i].uh[j] = (byte1 << 8) | (byte0 & 0xff);
+ }
+ }
+
+ check_output_h(__LINE__, BUFSIZE / 2);
+}
+
+static void test_vasrvwuhrndsat(void)
+{
+ void *p0 = buffer0;
+ void *p1 = buffer1;
+ void *pout = output;
+
+ memset(expect, 0xaa, sizeof(expect));
+ memset(output, 0xbb, sizeof(output));
+
+ for (int i = 0; i < BUFSIZE / 2; i++) {
+ asm("v4 = vmem(%0 + #0)\n\t"
+ "v5 = vmem(%0 + #1)\n\t"
+ "v6 = vmem(%1 + #0)\n\t"
+ "v5.uh = vasr(v5:4.w, v6.uh):rnd:sat\n\t"
+ "vmem(%2) = v5\n\t"
+ : : "r"(p0), "r"(p1), "r"(pout)
+ : "v4", "v5", "v6", "memory");
+ p0 += sizeof(MMVector) * 2;
+ p1 += sizeof(MMVector);
+ pout += sizeof(MMVector);
+
+ for (int j = 0; j < MAX_VEC_SIZE_BYTES / 2; j++) {
+ int shamt;
+ uint16_t half0;
+ uint16_t half1;
+
+ shamt = buffer1[i].uh[2 * j + 0] & 0xf;
+ half0 = fVSATUH(fVROUND(buffer0[2 * i + 0].w[j], shamt) >> shamt);
+ shamt = buffer1[i].uh[2 * j + 1] & 0xf;
+ half1 = fVSATUH(fVROUND(buffer0[2 * i + 1].w[j], shamt) >> shamt);
+ expect[i].w[j] = (half1 << 16) | (half0 & 0xffff);
I think we want MAX_VEC_SIZE_BYTES / 4 as the upper bound for this loop,
we currently
overflow since we're accessing words.
+ }
+ }
+
+ check_output_w(__LINE__, BUFSIZE / 2);
+}
+
+static void test_vasrvwuhsat(void)
+{
+ void *p0 = buffer0;
+ void *p1 = buffer1;
+ void *pout = output;
+
+ memset(expect, 0xaa, sizeof(expect));
+ memset(output, 0xbb, sizeof(output));
+
+ for (int i = 0; i < BUFSIZE / 2; i++) {
+ asm("v4 = vmem(%0 + #0)\n\t"
+ "v5 = vmem(%0 + #1)\n\t"
+ "v6 = vmem(%1 + #0)\n\t"
+ "v5.uh = vasr(v5:4.w, v6.uh):sat\n\t"
+ "vmem(%2) = v5\n\t"
+ : : "r"(p0), "r"(p1), "r"(pout)
+ : "v4", "v5", "v6", "memory");
+ p0 += sizeof(MMVector) * 2;
+ p1 += sizeof(MMVector);
+ pout += sizeof(MMVector);
+
+ for (int j = 0; j < MAX_VEC_SIZE_BYTES / 2; j++) {
+ int shamt;
+ uint16_t half0;
+ uint16_t half1;
+
+ shamt = buffer1[i].uh[2 * j + 0] & 0xf;
+ half0 = fVSATUH(buffer0[2 * i + 0].w[j] >> shamt);
+ shamt = buffer1[i].uh[2 * j + 1] & 0xf;
+ half1 = fVSATUH(buffer0[2 * i + 1].w[j] >> shamt);
+ expect[i].w[j] = (half1 << 16) | (half0 & 0xffff);
Same here.
Otherwise,
Reviewed-by: Anton Johansson <a...@rev.ng>
