ShuriZma
/
xemu
mirror of https://github.com/xemu-project/xemu.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

tests/tcg/aarch64: Add test cases for SME FMOPA (widening) 

Signed-off-by: Daniyal Khan <danikhan632@gmail.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20240717060149.204788-4-richard.henderson@linaro.org
Message-Id: 172090222034.13953.16888708708822922098-1@git.sr.ht
[rth: Split test from a larger patch, tidy assembly]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This commit is contained in:
Daniyal Khan 2024-07-17 16:01:49 +10:00 committed by Peter Maydell
parent 207d30b5fd
commit f103ecccb6
4 changed files with 185 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
tests/tcg/aarch64/Makefile.target
View File

@ -70,8 +70,9 @@ endif
# SME Tests
ifneq ($(CROSS_AS_HAS_ARMV9_SME),)
AARCH64_TESTS += sme-outprod1 sme-smopa-1 sme-smopa-2
sme-outprod1 sme-smopa-1 sme-smopa-2: CFLAGS += $(CROSS_AS_HAS_ARMV9_SME)
SME_TESTS = sme-outprod1 sme-smopa-1 sme-smopa-2 sme-fmopa-1 sme-fmopa-2 sme-fmopa-3
AARCH64_TESTS += $(SME_TESTS)
$(SME_TESTS): CFLAGS += $(CROSS_AS_HAS_ARMV9_SME)
endif
# System Registers Tests

63
tests/tcg/aarch64/sme-fmopa-1.c Normal file
View File

@ -0,0 +1,63 @@
/*
* SME outer product, 1 x 1.
* SPDX-License-Identifier: GPL-2.0-or-later
*/
#include <stdio.h>
static void foo(float *dst)
{
asm(".arch_extension sme\n\t"
"smstart\n\t"
"ptrue p0.s, vl4\n\t"
"fmov z0.s, #1.0\n\t"
/*
* An outer product of a vector of 1.0 by itself should be a matrix of 1.0.
* Note that we are using tile 1 here (za1.s) rather than tile 0.
*/
"zero {za}\n\t"
"fmopa za1.s, p0/m, p0/m, z0.s, z0.s\n\t"
/*
* Read the first 4x4 sub-matrix of elements from tile 1:
* Note that za1h should be interchangeable here.
*/
"mov w12, #0\n\t"
"mova z0.s, p0/m, za1v.s[w12, #0]\n\t"
"mova z1.s, p0/m, za1v.s[w12, #1]\n\t"
"mova z2.s, p0/m, za1v.s[w12, #2]\n\t"
"mova z3.s, p0/m, za1v.s[w12, #3]\n\t"
/*
* And store them to the input pointer (dst in the C code):
*/
"st1w {z0.s}, p0, [%0]\n\t"
"add x0, x0, #16\n\t"
"st1w {z1.s}, p0, [x0]\n\t"
"add x0, x0, #16\n\t"
"st1w {z2.s}, p0, [x0]\n\t"
"add x0, x0, #16\n\t"
"st1w {z3.s}, p0, [x0]\n\t"
"smstop"
: : "r"(dst)
: "x12", "d0", "d1", "d2", "d3", "memory");
}
int main()
{
float dst[16] = { };
foo(dst);
for (int i = 0; i < 16; i++) {
if (dst[i] != 1.0f) {
goto failure;
}
}
/* success */
return 0;
failure:
for (int i = 0; i < 16; i++) {
printf("%f%c", dst[i], i % 4 == 3 ? '\n' : ' ');
}
return 1;
}

56
tests/tcg/aarch64/sme-fmopa-2.c Normal file
View File

@ -0,0 +1,56 @@
/*
* SME outer product, FZ vs FZ16
* SPDX-License-Identifier: GPL-2.0-or-later
*/
#include <stdint.h>
#include <stdio.h>
static void test_fmopa(uint32_t *result)
{
asm(".arch_extension sme\n\t"
"smstart\n\t" /* Z*, P* and ZArray cleared */
"ptrue p2.b, vl16\n\t" /* Limit vector length to 16 */
"ptrue p5.b, vl16\n\t"
"movi d0, #0x00ff\n\t" /* fp16 denormal */
"movi d16, #0x00ff\n\t"
"mov w15, #0x0001000000\n\t" /* FZ=1, FZ16=0 */
"msr fpcr, x15\n\t"
"fmopa za3.s, p2/m, p5/m, z16.h, z0.h\n\t"
"mov w15, #0\n\t"
"st1w {za3h.s[w15, 0]}, p2, [%0]\n\t"
"add %0, %0, #16\n\t"
"st1w {za3h.s[w15, 1]}, p2, [%0]\n\t"
"mov w15, #2\n\t"
"add %0, %0, #16\n\t"
"st1w {za3h.s[w15, 0]}, p2, [%0]\n\t"
"add %0, %0, #16\n\t"
"st1w {za3h.s[w15, 1]}, p2, [%0]\n\t"
"smstop"
: "+r"(result) :
: "x15", "x16", "p2", "p5", "d0", "d16", "memory");
}
int main(void)
{
uint32_t result[4 * 4] = { };
test_fmopa(result);
if (result[0] != 0x2f7e0100) {
printf("Test failed: Incorrect output in first 4 bytes\n"
"Expected: %08x\n"
"Got: %08x\n",
0x2f7e0100, result[0]);
return 1;
}
for (int i = 1; i < 16; ++i) {
if (result[i] != 0) {
printf("Test failed: Non-zero word at position %d\n", i);
return 1;
}
}
return 0;
}

63
tests/tcg/aarch64/sme-fmopa-3.c Normal file
View File

@ -0,0 +1,63 @@
/*
* SME outer product, [ 1 2 3 4 ] squared
* SPDX-License-Identifier: GPL-2.0-or-later
*/
#include <stdio.h>
#include <stdint.h>
#include <string.h>
#include <math.h>
static const float i_1234[4] = {
1.0f, 2.0f, 3.0f, 4.0f
};
static const float expected[4] = {
4.515625f, 5.750000f, 6.984375f, 8.218750f
};
static void test_fmopa(float *result)
{
asm(".arch_extension sme\n\t"
"smstart\n\t" /* ZArray cleared */
"ptrue p2.b, vl16\n\t" /* Limit vector length to 16 */
"ld1w {z0.s}, p2/z, [%1]\n\t"
"mov w15, #0\n\t"
"mov za3h.s[w15, 0], p2/m, z0.s\n\t"
"mov za3h.s[w15, 1], p2/m, z0.s\n\t"
"mov w15, #2\n\t"
"mov za3h.s[w15, 0], p2/m, z0.s\n\t"
"mov za3h.s[w15, 1], p2/m, z0.s\n\t"
"msr fpcr, xzr\n\t"
"fmopa za3.s, p2/m, p2/m, z0.h, z0.h\n\t"
"mov w15, #0\n\t"
"st1w {za3h.s[w15, 0]}, p2, [%0]\n"
"add %0, %0, #16\n\t"
"st1w {za3h.s[w15, 1]}, p2, [%0]\n\t"
"mov w15, #2\n\t"
"add %0, %0, #16\n\t"
"st1w {za3h.s[w15, 0]}, p2, [%0]\n\t"
"add %0, %0, #16\n\t"
"st1w {za3h.s[w15, 1]}, p2, [%0]\n\t"
"smstop"
: "+r"(result) : "r"(i_1234)
: "x15", "x16", "p2", "d0", "memory");
}
int main(void)
{
float result[4 * 4] = { };
int ret = 0;
test_fmopa(result);
for (int i = 0; i < 4; i++) {
float actual = result[i];
if (fabsf(actual - expected[i]) > 0.001f) {
printf("Test failed at element %d: Expected %f, got %f\n",
i, expected[i], actual);
ret = 1;
}
}
return ret;
}