aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorRobin Dapp <rdapp@ventanamicro.com>2024-05-13 22:09:35 +0200
committerRobin Dapp <rdapp@ventanamicro.com>2024-05-31 21:54:48 +0200
commitaf4bf422a699de0e7af5a26e02997d313e7301a6 (patch)
tree506b9345bd5aabbc9a934b3c7eba03ac3a9dd2ec
parent9781885a624f3e29634d95c14cd10940cefb1a5a (diff)
downloadgcc-af4bf422a699de0e7af5a26e02997d313e7301a6.zip
gcc-af4bf422a699de0e7af5a26e02997d313e7301a6.tar.gz
gcc-af4bf422a699de0e7af5a26e02997d313e7301a6.tar.bz2
RISC-V: Add vwsll combine helpers.
This patch enables the usage of vwsll in autovec context by adding the necessary combine patterns and tests. gcc/ChangeLog: * config/riscv/autovec-opt.md (*vwsll_zext1_<mode>): New pattern. (*vwsll_zext2_<mode>): Ditto. (*vwsll_zext1_scalar_<mode>): Ditto. (*vwsll_zext1_trunc_<mode>): Ditto. (*vwsll_zext2_trunc_<mode>): Ditto. (*vwsll_zext1_trunc_scalar_<mode>): Ditto. * config/riscv/vector-crypto.md: Make pattern similar to other narrowing/widening patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/autovec/binop/vwsll-1.c: New test. * gcc.target/riscv/rvv/autovec/binop/vwsll-run.c: New test. * gcc.target/riscv/rvv/autovec/binop/vwsll-template.h: New test.
-rw-r--r--gcc/config/riscv/autovec-opt.md126
-rwxr-xr-xgcc/config/riscv/vector-crypto.md2
-rw-r--r--gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-1.c10
-rw-r--r--gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-run.c67
-rw-r--r--gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-template.h49
5 files changed, 251 insertions, 3 deletions
diff --git a/gcc/config/riscv/autovec-opt.md b/gcc/config/riscv/autovec-opt.md
index 04f85d8..bc6af04 100644
--- a/gcc/config/riscv/autovec-opt.md
+++ b/gcc/config/riscv/autovec-opt.md
@@ -1467,5 +1467,127 @@
operands, operands[4]);
DONE;
}
- [(set_attr "type" "vector")]
-)
+ [(set_attr "type" "vector")])
+
+;; vzext.vf2 + vsll = vwsll.
+(define_insn_and_split "*vwsll_zext1_<mode>"
+ [(set (match_operand:VWEXTI 0 "register_operand" "=vr ")
+ (ashift:VWEXTI
+ (zero_extend:VWEXTI
+ (match_operand:<V_DOUBLE_TRUNC> 1 "register_operand" " vr "))
+ (match_operand:<V_DOUBLE_TRUNC> 2 "vector_shift_operand" "vrvk")))]
+ "TARGET_ZVBB && can_create_pseudo_p ()"
+ "#"
+ "&& 1"
+ [(const_int 0)]
+ {
+ insn_code icode = code_for_pred_vwsll (<MODE>mode);
+ riscv_vector::emit_vlmax_insn (icode, riscv_vector::BINARY_OP, operands);
+ DONE;
+ }
+ [(set_attr "type" "vwsll")])
+
+(define_insn_and_split "*vwsll_zext2_<mode>"
+ [(set (match_operand:VWEXTI 0 "register_operand" "=vr ")
+ (ashift:VWEXTI
+ (zero_extend:VWEXTI
+ (match_operand:<V_DOUBLE_TRUNC> 1 "register_operand" " vr "))
+ (zero_extend:VWEXTI
+ (match_operand:<V_DOUBLE_TRUNC> 2 "vector_shift_operand" "vrvk"))))]
+ "TARGET_ZVBB && can_create_pseudo_p ()"
+ "#"
+ "&& 1"
+ [(const_int 0)]
+ {
+ insn_code icode = code_for_pred_vwsll (<MODE>mode);
+ riscv_vector::emit_vlmax_insn (icode, riscv_vector::BINARY_OP, operands);
+ DONE;
+ }
+ [(set_attr "type" "vwsll")])
+
+
+(define_insn_and_split "*vwsll_zext1_scalar_<mode>"
+ [(set (match_operand:VWEXTI 0 "register_operand" "=vr")
+ (ashift:VWEXTI
+ (zero_extend:VWEXTI
+ (match_operand:<V_DOUBLE_TRUNC> 1 "register_operand" " vr"))
+ (match_operand:<VEL> 2 "vector_scalar_shift_operand" " rK")))]
+ "TARGET_ZVBB && can_create_pseudo_p ()"
+ "#"
+ "&& 1"
+ [(const_int 0)]
+ {
+ if (GET_CODE (operands[2]) == SUBREG)
+ operands[2] = SUBREG_REG (operands[2]);
+ insn_code icode = code_for_pred_vwsll_scalar (<MODE>mode);
+ riscv_vector::emit_vlmax_insn (icode, riscv_vector::BINARY_OP, operands);
+ DONE;
+ }
+ [(set_attr "type" "vwsll")])
+
+;; For
+;; uint16_t dst;
+;; uint8_t a, b;
+;; dst = vwsll (a, b)
+;; we seem to create
+;; aa = (int) a;
+;; bb = (int) b;
+;; dst = (short) vwsll (aa, bb);
+;; The following patterns help to combine this idiom into one vwsll.
+
+(define_insn_and_split "*vwsll_zext1_trunc_<mode>"
+ [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vr ")
+ (truncate:<V_DOUBLE_TRUNC>
+ (ashift:VQEXTI
+ (zero_extend:VQEXTI
+ (match_operand:<V_QUAD_TRUNC> 1 "register_operand" " vr "))
+ (match_operand:VQEXTI 2 "vector_shift_operand" "vrvk"))))]
+ "TARGET_ZVBB && can_create_pseudo_p ()"
+ "#"
+ "&& 1"
+ [(const_int 0)]
+ {
+ insn_code icode = code_for_pred_vwsll (<V_DOUBLE_TRUNC>mode);
+ riscv_vector::emit_vlmax_insn (icode, riscv_vector::BINARY_OP, operands);
+ DONE;
+ }
+ [(set_attr "type" "vwsll")])
+
+(define_insn_and_split "*vwsll_zext2_trunc_<mode>"
+ [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vr ")
+ (truncate:<V_DOUBLE_TRUNC>
+ (ashift:VQEXTI
+ (zero_extend:VQEXTI
+ (match_operand:<V_QUAD_TRUNC> 1 "register_operand" " vr "))
+ (zero_extend:VQEXTI
+ (match_operand:<V_QUAD_TRUNC> 2 "vector_shift_operand" "vrvk")))))]
+ "TARGET_ZVBB && can_create_pseudo_p ()"
+ "#"
+ "&& 1"
+ [(const_int 0)]
+ {
+ insn_code icode = code_for_pred_vwsll (<V_DOUBLE_TRUNC>mode);
+ riscv_vector::emit_vlmax_insn (icode, riscv_vector::BINARY_OP, operands);
+ DONE;
+ }
+ [(set_attr "type" "vwsll")])
+
+(define_insn_and_split "*vwsll_zext1_trunc_scalar_<mode>"
+ [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vr ")
+ (truncate:<V_DOUBLE_TRUNC>
+ (ashift:VQEXTI
+ (zero_extend:VQEXTI
+ (match_operand:<V_QUAD_TRUNC> 1 "register_operand" " vr "))
+ (match_operand:<VEL> 2 "vector_scalar_shift_operand" " rK"))))]
+ "TARGET_ZVBB && can_create_pseudo_p ()"
+ "#"
+ "&& 1"
+ [(const_int 0)]
+ {
+ if (GET_CODE (operands[2]) == SUBREG)
+ operands[2] = SUBREG_REG (operands[2]);
+ insn_code icode = code_for_pred_vwsll_scalar (<V_DOUBLE_TRUNC>mode);
+ riscv_vector::emit_vlmax_insn (icode, riscv_vector::BINARY_OP, operands);
+ DONE;
+ }
+ [(set_attr "type" "vwsll")])
diff --git a/gcc/config/riscv/vector-crypto.md b/gcc/config/riscv/vector-crypto.md
index e474ddf..24822e2 100755
--- a/gcc/config/riscv/vector-crypto.md
+++ b/gcc/config/riscv/vector-crypto.md
@@ -298,7 +298,7 @@
(match_operand:<V_DOUBLE_TRUNC> 4 "register_operand" "vr"))
(match_operand:VWEXTI 2 "vector_merge_operand" "0vu")))]
"TARGET_ZVBB"
- "vwsll.vv\t%0,%3,%4%p1"
+ "vwsll.v%o4\t%0,%3,%4%p1"
[(set_attr "type" "vwsll")
(set_attr "mode" "<V_DOUBLE_TRUNC>")])
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-1.c
new file mode 100644
index 0000000..a2e5b4f
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-1.c
@@ -0,0 +1,10 @@
+/* { dg-do compile } */
+/* { dg-add-options "riscv_v" } */
+/* { dg-add-options "riscv_zvbb" } */
+/* { dg-additional-options "-std=c99 -fno-vect-cost-model" } */
+
+#include "vwsll-template.h"
+
+/* { dg-final { scan-assembler-times {\tvwsll\.vv} 3 } } */
+/* { dg-final { scan-assembler-times {\tvwsll\.vx} 3 } } */
+/* { dg-final { scan-assembler-times {\tvwsll\.vi} 3 } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-run.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-run.c
new file mode 100644
index 0000000..ddb8461
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-run.c
@@ -0,0 +1,67 @@
+/* { dg-do run } */
+/* { dg-require-effective-target "riscv_zvbb_ok" } */
+/* { dg-add-options "riscv_v" } */
+/* { dg-add-options "riscv_zvbb" } */
+/* { dg-additional-options "-std=c99 -fno-vect-cost-model" } */
+
+#include "vwsll-template.h"
+
+#include <assert.h>
+
+#define SZ 512
+
+#define RUN(TYPE1, TYPE2, VAL) \
+ TYPE1 dst##TYPE1[SZ]; \
+ TYPE2 a##TYPE2[SZ]; \
+ TYPE2 b##TYPE2[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ dst##TYPE1[i] = 0; \
+ a##TYPE2[i] = VAL; \
+ b##TYPE2[i] = i % 4; \
+ } \
+ vwsll_vv##TYPE1 (dst##TYPE1, a##TYPE2, b##TYPE2, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (dst##TYPE1[i] == (VAL << (i % 4)));
+
+#define RUN2(TYPE1, TYPE2, VAL) \
+ TYPE1 dst2##TYPE1[SZ]; \
+ TYPE2 a2##TYPE2[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ dst2##TYPE1[i] = 0; \
+ a2##TYPE2[i] = VAL; \
+ } \
+ TYPE2 b2##TYPE2 = 7; \
+ vwsll_vx##TYPE1 (dst2##TYPE1, a2##TYPE2, b2##TYPE2, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (dst2##TYPE1[i] == (VAL << b2##TYPE2));
+
+#define RUN3(TYPE1, TYPE2, VAL) \
+ TYPE1 dst3##TYPE1[SZ]; \
+ TYPE2 a3##TYPE2[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ dst3##TYPE1[i] = 0; \
+ a3##TYPE2[i] = VAL; \
+ } \
+ vwsll_vi##TYPE1 (dst3##TYPE1, a3##TYPE2, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (dst3##TYPE1[i] == (VAL << 6));
+
+#define RUN_ALL() \
+ RUN (uint16_t, uint8_t, 2) \
+ RUN (uint32_t, uint16_t, 2) \
+ RUN (uint64_t, uint32_t, 4) \
+ RUN2 (uint16_t, uint8_t, 8) \
+ RUN2 (uint32_t, uint16_t, 8) \
+ RUN2 (uint64_t, uint32_t, 10) \
+ RUN3 (uint16_t, uint8_t, 255) \
+ RUN3 (uint32_t, uint16_t, 34853) \
+ RUN3 (uint64_t, uint32_t, 1794394)
+
+int
+main ()
+{
+ RUN_ALL ()
+}
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-template.h b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-template.h
new file mode 100644
index 0000000..376cbae
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/vwsll-template.h
@@ -0,0 +1,49 @@
+/* { dg-do compile } */
+/* { dg-add-options "riscv_v" } */
+/* { dg-add-options "riscv_zvbb" } */
+/* { dg-additional-options "-std=c99 -fno-vect-cost-model" } */
+
+#include <stdint-gcc.h>
+
+#define TEST1_TYPE(TYPE1, TYPE2) \
+ __attribute__ ((noipa)) void vwsll_vv##TYPE1 (TYPE1 *restrict dst, \
+ TYPE2 *restrict a, \
+ TYPE2 *restrict b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = (TYPE1) a[i] << b[i]; \
+ }
+
+#define TEST2_TYPE(TYPE1, TYPE2) \
+ __attribute__ ((noipa)) void vwsll_vx##TYPE1 (TYPE1 *restrict dst, \
+ TYPE2 *restrict a, TYPE2 b, \
+ int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = (TYPE1) a[i] << b; \
+ }
+
+#define TEST3_TYPE(TYPE1, TYPE2) \
+ __attribute__ ((noipa)) void vwsll_vi##TYPE1 (TYPE1 *restrict dst, \
+ TYPE2 *restrict a, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = (TYPE1) a[i] << 6; \
+ }
+
+#define TEST_ALL() \
+ TEST1_TYPE (uint16_t, uint8_t) \
+ TEST1_TYPE (uint32_t, uint16_t) \
+ TEST1_TYPE (uint64_t, uint32_t) \
+ TEST2_TYPE (uint16_t, uint8_t) \
+ TEST2_TYPE (uint32_t, uint16_t) \
+ TEST2_TYPE (uint64_t, uint32_t) \
+ TEST3_TYPE (uint16_t, uint8_t) \
+ TEST3_TYPE (uint32_t, uint16_t) \
+ TEST3_TYPE (uint64_t, uint32_t)
+
+TEST_ALL()
+
+/* { dg-final { scan-assembler-times {\tvwsll\.vv} 3 } } */
+/* { dg-final { scan-assembler-times {\tvwsll\.vx} 3 } } */
+/* { dg-final { scan-assembler-times {\tvwsll\.vi} 3 } } */