[V3] RISC-V: Add more SLP tests
Checks
Commit Message
From: Juzhe-Zhong <juzhe.zhong@rivai.ai>
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/partial/slp-10.c: New test.
* gcc.target/riscv/rvv/autovec/partial/slp-11.c: New test.
* gcc.target/riscv/rvv/autovec/partial/slp-13.c: New test.
* gcc.target/riscv/rvv/autovec/partial/slp-14.c: New test.
* gcc.target/riscv/rvv/autovec/partial/slp-15.c: New test.
* gcc.target/riscv/rvv/autovec/partial/slp_run-10.c: New test.
* gcc.target/riscv/rvv/autovec/partial/slp_run-11.c: New test.
* gcc.target/riscv/rvv/autovec/partial/slp_run-13.c: New test.
* gcc.target/riscv/rvv/autovec/partial/slp_run-14.c: New test.
* gcc.target/riscv/rvv/autovec/partial/slp_run-15.c: New test.
---
.../riscv/rvv/autovec/partial/slp-10.c | 32 +++++++++++
.../riscv/rvv/autovec/partial/slp-11.c | 33 +++++++++++
.../riscv/rvv/autovec/partial/slp-13.c | 34 +++++++++++
.../riscv/rvv/autovec/partial/slp-14.c | 33 +++++++++++
.../riscv/rvv/autovec/partial/slp-15.c | 35 ++++++++++++
.../riscv/rvv/autovec/partial/slp_run-10.c | 33 +++++++++++
.../riscv/rvv/autovec/partial/slp_run-11.c | 33 +++++++++++
.../riscv/rvv/autovec/partial/slp_run-13.c | 47 +++++++++++++++
.../riscv/rvv/autovec/partial/slp_run-14.c | 57 +++++++++++++++++++
.../riscv/rvv/autovec/partial/slp_run-15.c | 56 ++++++++++++++++++
10 files changed, 393 insertions(+)
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp-10.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp-11.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp-13.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp-14.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp-15.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp_run-10.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp_run-11.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp_run-13.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp_run-14.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/partial/slp_run-15.c
Comments
Hi Juzhe,
thanks, works for me as is. I just hope somebody is going to take on the task
of making different LMUL SLP variants "scannable" at some point because
it would definitely increase our test coverage with these tests. (Or split
the tests manually and not iterate over LMUL)
Regards
Robin
Ok. After floating-point binary. I will do floating-point ternary.
I think you do conversion next (widen floating point, float to int, int to float).
It seems that we almost done most of the part autovec patterns in RISC-V port.
What else we can do? My second middle-end patch (LEN_MASK _* load/store) is blocked which is prerequisite for reduction
if you understand how reduction works.
Maybe next you could find the way to optimize vv->vx ?
Thanks.
juzhe.zhong@rivai.ai
From: Robin Dapp
Date: 2023-06-13 20:00
To: juzhe.zhong; gcc-patches
CC: rdapp.gcc; kito.cheng; kito.cheng; palmer; palmer; jeffreyalaw
Subject: Re: [PATCH V3] RISC-V: Add more SLP tests
Hi Juzhe,
thanks, works for me as is. I just hope somebody is going to take on the task
of making different LMUL SLP variants "scannable" at some point because
it would definitely increase our test coverage with these tests. (Or split
the tests manually and not iterate over LMUL)
Regards
Robin
On 6/13/23 05:38, juzhe.zhong@rivai.ai wrote:
> From: Juzhe-Zhong <juzhe.zhong@rivai.ai>
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/riscv/rvv/autovec/partial/slp-10.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp-11.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp-13.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp-14.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp-15.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-10.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-11.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-13.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-14.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-15.c: New test.
>
> ---
OK
jeff
Committed, thanks Jeff.
Pan
-----Original Message-----
From: Gcc-patches <gcc-patches-bounces+pan2.li=intel.com@gcc.gnu.org> On Behalf Of Jeff Law via Gcc-patches
Sent: Tuesday, June 13, 2023 9:26 PM
To: juzhe.zhong@rivai.ai; gcc-patches@gcc.gnu.org
Cc: kito.cheng@gmail.com; kito.cheng@sifive.com; palmer@dabbelt.com; palmer@rivosinc.com; rdapp.gcc@gmail.com
Subject: Re: [PATCH V3] RISC-V: Add more SLP tests
On 6/13/23 05:38, juzhe.zhong@rivai.ai wrote:
> From: Juzhe-Zhong <juzhe.zhong@rivai.ai>
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/riscv/rvv/autovec/partial/slp-10.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp-11.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp-13.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp-14.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp-15.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-10.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-11.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-13.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-14.c: New test.
> * gcc.target/riscv/rvv/autovec/partial/slp_run-15.c: New test.
>
> ---
OK
jeff
new file mode 100644
@@ -0,0 +1,32 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-march=rv32gcv -mabi=ilp32d --param riscv-autovec-preference=scalable -fno-vect-cost-model -fdump-tree-optimized-details" } */
+
+#include <stdint-gcc.h>
+
+#define VEC_PERM(TYPE) \
+ TYPE __attribute__ ((noinline, noclone)) \
+ vec_slp_##TYPE (TYPE *restrict a, int n) \
+ { \
+ for (int i = 0; i < n; ++i) \
+ { \
+ a[i * 2] += 10; \
+ a[i * 2 + 1] += 17; \
+ } \
+ }
+
+#define TEST_ALL(T) \
+ T (int8_t) \
+ T (uint8_t) \
+ T (int16_t) \
+ T (uint16_t) \
+ T (int32_t) \
+ T (uint32_t) \
+ T (int64_t) \
+ T (uint64_t)
+
+TEST_ALL (VEC_PERM)
+
+/* { dg-final { scan-tree-dump-times "{ 10, 17, ... }" 8 "optimized" } } */
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */
+
new file mode 100644
@@ -0,0 +1,33 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-march=rv32gcv -mabi=ilp32d --param riscv-autovec-preference=scalable -fno-vect-cost-model -fdump-tree-optimized-details" } */
+
+#include <stdint-gcc.h>
+
+#define VEC_PERM(TYPE) \
+ TYPE __attribute__ ((noinline, noclone)) \
+ vec_slp_##TYPE (TYPE *restrict a, int n) \
+ { \
+ for (int i = 0; i < n; ++i) \
+ { \
+ a[i * 4] += 41; \
+ a[i * 4 + 1] += 25; \
+ a[i * 4 + 2] += 31; \
+ a[i * 4 + 3] += 62; \
+ } \
+ }
+
+#define TEST_ALL(T) \
+ T (int8_t) \
+ T (uint8_t) \
+ T (int16_t) \
+ T (uint16_t) \
+ T (int32_t) \
+ T (uint32_t) \
+ T (int64_t) \
+ T (uint64_t)
+
+TEST_ALL (VEC_PERM)
+
+/* { dg-final { scan-tree-dump "{ 41, 25, 31, 62, ... }" "optimized" } } */
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */
new file mode 100644
@@ -0,0 +1,34 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-march=rv32gcv -mabi=ilp32d --param riscv-autovec-preference=scalable -fno-vect-cost-model" } */
+
+#include <stdint-gcc.h>
+
+#define VEC_PERM(TYPE) \
+ void __attribute__ ((noinline, noclone)) \
+ vec_slp_##TYPE (TYPE *restrict a, TYPE *restrict b, int n) \
+ { \
+ for (int i = 0; i < n; ++i) \
+ { \
+ a[i * 2] += 1; \
+ a[i * 2 + 1] += 2; \
+ b[i * 4] += 3; \
+ b[i * 4 + 1] += 4; \
+ b[i * 4 + 2] += 5; \
+ b[i * 4 + 3] += 6; \
+ } \
+ }
+
+#define TEST_ALL(T) \
+ T (int8_t) \
+ T (uint8_t) \
+ T (int16_t) \
+ T (uint16_t) \
+ T (int32_t) \
+ T (uint32_t) \
+ T (int64_t) \
+ T (uint64_t)
+
+TEST_ALL (VEC_PERM)
+
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */
new file mode 100644
@@ -0,0 +1,33 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-march=rv32gcv -mabi=ilp32d --param riscv-autovec-preference=scalable -fno-vect-cost-model" } */
+
+#include <stdint-gcc.h>
+
+#define VEC_PERM(TYPE) \
+ void __attribute__ ((noinline, noclone)) \
+ vec_slp_##TYPE (TYPE *restrict a, TYPE *restrict b, int n) \
+ { \
+ for (int i = 0; i < n; ++i) \
+ { \
+ a[i] += 1; \
+ b[i * 4] += 2; \
+ b[i * 4 + 1] += 3; \
+ b[i * 4 + 2] += 4; \
+ b[i * 4 + 3] += 5; \
+ } \
+ }
+
+#define TEST_ALL(T) \
+ T (int8_t) \
+ T (uint8_t) \
+ T (int16_t) \
+ T (uint16_t) \
+ T (int32_t) \
+ T (uint32_t) \
+ T (int64_t) \
+ T (uint64_t)
+
+TEST_ALL (VEC_PERM)
+
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */
new file mode 100644
@@ -0,0 +1,35 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-march=rv32gcv -mabi=ilp32d --param riscv-autovec-preference=scalable -fno-vect-cost-model" } */
+
+#include <stdint-gcc.h>
+
+#define N1 (19 * 2)
+
+#define VEC_PERM(TYPE) \
+ void __attribute__ ((noinline, noclone)) \
+ vec_slp_##TYPE (TYPE *restrict a, TYPE *restrict b) \
+ { \
+ for (int i = 0; i < N1; ++i) \
+ { \
+ a[i] += 1; \
+ b[i * 4] += 2; \
+ b[i * 4 + 1] += 3; \
+ b[i * 4 + 2] += 4; \
+ b[i * 4 + 3] += 5; \
+ } \
+ }
+
+#define TEST_ALL(T) \
+ T (int8_t) \
+ T (uint8_t) \
+ T (int16_t) \
+ T (uint16_t) \
+ T (int32_t) \
+ T (uint32_t) \
+ T (int64_t) \
+ T (uint64_t)
+
+TEST_ALL (VEC_PERM)
+
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */
new file mode 100644
@@ -0,0 +1,33 @@
+/* { dg-do run { target { riscv_vector } } } */
+/* { dg-additional-options "--param riscv-autovec-preference=scalable -fno-vect-cost-model" } */
+
+#include "slp-10.c"
+
+#define N (103 * 2)
+
+#define HARNESS(TYPE) \
+ { \
+ TYPE a[N], b[2] = { 10, 17 }; \
+ for (unsigned int i = 0; i < N; ++i) \
+ { \
+ a[i] = i * 2 + i % 5; \
+ asm volatile ("" ::: "memory"); \
+ } \
+ vec_slp_##TYPE (a, N / 2); \
+ for (unsigned int i = 0; i < N; ++i) \
+ { \
+ TYPE orig = i * 2 + i % 5; \
+ TYPE expected = orig + b[i % 2]; \
+ if (a[i] != expected) \
+ __builtin_abort (); \
+ } \
+ }
+
+int __attribute__ ((optimize (1)))
+main (void)
+{
+ TEST_ALL (HARNESS)
+}
+
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */
new file mode 100644
@@ -0,0 +1,33 @@
+/* { dg-do run { target { riscv_vector } } } */
+/* { dg-additional-options "--param riscv-autovec-preference=scalable -fno-vect-cost-model" } */
+
+#include "slp-11.c"
+
+#define N (77 * 4)
+
+#define HARNESS(TYPE) \
+ { \
+ TYPE a[N], b[4] = { 41, 25, 31, 62 }; \
+ for (unsigned int i = 0; i < N; ++i) \
+ { \
+ a[i] = i * 2 + i % 5; \
+ asm volatile ("" ::: "memory"); \
+ } \
+ vec_slp_##TYPE (a, N / 4); \
+ for (unsigned int i = 0; i < N; ++i) \
+ { \
+ TYPE orig = i * 2 + i % 5; \
+ TYPE expected = orig + b[i % 4]; \
+ if (a[i] != expected) \
+ __builtin_abort (); \
+ } \
+ }
+
+int __attribute__ ((optimize (1)))
+main (void)
+{
+ TEST_ALL (HARNESS)
+}
+
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */
new file mode 100644
@@ -0,0 +1,47 @@
+/* { dg-do run { target { riscv_vector } } } */
+/* { dg-additional-options "--param riscv-autovec-preference=scalable -fno-vect-cost-model" } */
+
+#include "slp-13.c"
+
+#define N1 (103 * 2)
+#define N2 (111 * 2)
+
+#define HARNESS(TYPE) \
+ { \
+ TYPE a[N2], b[N2 * 2]; \
+ for (unsigned int i = 0; i < N2; ++i) \
+ { \
+ a[i] = i * 2 + i % 5; \
+ b[i * 2] = i * 3 + i % 7; \
+ b[i * 2 + 1] = i * 5 + i % 9; \
+ } \
+ vec_slp_##TYPE (a, b, N1 / 2); \
+ for (unsigned int i = 0; i < N2; ++i) \
+ { \
+ TYPE orig_a = i * 2 + i % 5; \
+ TYPE orig_b1 = i * 3 + i % 7; \
+ TYPE orig_b2 = i * 5 + i % 9; \
+ TYPE expected_a = orig_a; \
+ TYPE expected_b1 = orig_b1; \
+ TYPE expected_b2 = orig_b2; \
+ if (i < N1) \
+ { \
+ expected_a += i & 1 ? 2 : 1; \
+ expected_b1 += i & 1 ? 5 : 3; \
+ expected_b2 += i & 1 ? 6 : 4; \
+ } \
+ if (a[i] != expected_a \
+ || b[i * 2] != expected_b1 \
+ || b[i * 2 + 1] != expected_b2) \
+ __builtin_abort (); \
+ } \
+ }
+
+int __attribute__ ((optimize (1)))
+main (void)
+{
+ TEST_ALL (HARNESS)
+}
+
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */
new file mode 100644
@@ -0,0 +1,57 @@
+/* { dg-do run { target { riscv_vector } } } */
+/* { dg-additional-options "--param riscv-autovec-preference=scalable -fno-vect-cost-model" } */
+
+#include "slp-14.c"
+
+#define N1 (103 * 2)
+#define N2 (111 * 2)
+
+#define HARNESS(TYPE) \
+ { \
+ TYPE a[N2], b[N2 * 4]; \
+ for (unsigned int i = 0; i < N2; ++i) \
+ { \
+ a[i] = i * 2 + i % 5; \
+ b[i * 4] = i * 3 + i % 7; \
+ b[i * 4 + 1] = i * 5 + i % 9; \
+ b[i * 4 + 2] = i * 7 + i % 11; \
+ b[i * 4 + 3] = i * 9 + i % 13; \
+ } \
+ vec_slp_##TYPE (a, b, N1); \
+ for (unsigned int i = 0; i < N2; ++i) \
+ { \
+ TYPE orig_a = i * 2 + i % 5; \
+ TYPE orig_b1 = i * 3 + i % 7; \
+ TYPE orig_b2 = i * 5 + i % 9; \
+ TYPE orig_b3 = i * 7 + i % 11; \
+ TYPE orig_b4 = i * 9 + i % 13; \
+ TYPE expected_a = orig_a; \
+ TYPE expected_b1 = orig_b1; \
+ TYPE expected_b2 = orig_b2; \
+ TYPE expected_b3 = orig_b3; \
+ TYPE expected_b4 = orig_b4; \
+ if (i < N1) \
+ { \
+ expected_a += 1; \
+ expected_b1 += 2; \
+ expected_b2 += 3; \
+ expected_b3 += 4; \
+ expected_b4 += 5; \
+ } \
+ if (a[i] != expected_a \
+ || b[i * 4] != expected_b1 \
+ || b[i * 4 + 1] != expected_b2 \
+ || b[i * 4 + 2] != expected_b3 \
+ || b[i * 4 + 3] != expected_b4) \
+ __builtin_abort (); \
+ } \
+ }
+
+int __attribute__ ((optimize (1)))
+main (void)
+{
+ TEST_ALL (HARNESS)
+}
+
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */
new file mode 100644
@@ -0,0 +1,56 @@
+/* { dg-do run { target { riscv_vector } } } */
+/* { dg-additional-options "--param riscv-autovec-preference=scalable -fno-vect-cost-model" } */
+
+#include "slp-15.c"
+
+#define N2 (31 * 2)
+
+#define HARNESS(TYPE) \
+ { \
+ TYPE a[N2], b[N2 * 4]; \
+ for (unsigned int i = 0; i < N2; ++i) \
+ { \
+ a[i] = i * 2 + i % 5; \
+ b[i * 4] = i * 3 + i % 7; \
+ b[i * 4 + 1] = i * 5 + i % 9; \
+ b[i * 4 + 2] = i * 7 + i % 11; \
+ b[i * 4 + 3] = i * 9 + i % 13; \
+ } \
+ vec_slp_##TYPE (a, b); \
+ for (unsigned int i = 0; i < N2; ++i) \
+ { \
+ TYPE orig_a = i * 2 + i % 5; \
+ TYPE orig_b1 = i * 3 + i % 7; \
+ TYPE orig_b2 = i * 5 + i % 9; \
+ TYPE orig_b3 = i * 7 + i % 11; \
+ TYPE orig_b4 = i * 9 + i % 13; \
+ TYPE expected_a = orig_a; \
+ TYPE expected_b1 = orig_b1; \
+ TYPE expected_b2 = orig_b2; \
+ TYPE expected_b3 = orig_b3; \
+ TYPE expected_b4 = orig_b4; \
+ if (i < N1) \
+ { \
+ expected_a += 1; \
+ expected_b1 += 2; \
+ expected_b2 += 3; \
+ expected_b3 += 4; \
+ expected_b4 += 5; \
+ } \
+ if (a[i] != expected_a \
+ || b[i * 4] != expected_b1 \
+ || b[i * 4 + 1] != expected_b2 \
+ || b[i * 4 + 2] != expected_b3 \
+ || b[i * 4 + 3] != expected_b4) \
+ __builtin_abort (); \
+ } \
+ }
+
+int __attribute__ ((optimize (1)))
+main (void)
+{
+ TEST_ALL (HARNESS)
+}
+
+/* This testcase is from aarch64 and floating-point operations are removed.
+ TODO: We will add floating-point operations back and make them as common test in the future. */