* [RFC] RISC-V: Eliminate extension after for *w instructions
@ 2023-05-24 23:14 Jivan Hakobyan
2023-05-31 0:49 ` Jeff Law
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: Jivan Hakobyan @ 2023-05-24 23:14 UTC (permalink / raw)
To: gcc-patches
[-- Attachment #1.1: Type: text/plain, Size: 1279 bytes --]
`This patch tries to prevent generating unnecessary sign extension
after *w instructions like "addiw" or "divw".
The main idea of it is to add SUBREG_PROMOTED fields during expanding.
I have tested on SPEC2017 there is no regression.
Only gcc.dg/pr30957-1.c test failed.
To solve that I did some changes in loop-iv.cc, but not sure that it is
suitable.
gcc/ChangeLog:
* config/riscv/bitmanip.md (rotrdi3): New pattern.
(rotrsi3): Likewise.
(rotlsi3): Likewise.
* config/riscv/riscv-protos.h (riscv_emit_binary): New function
declaration
* config/riscv/riscv.cc (riscv_emit_binary): Removed static
* config/riscv/riscv.md (addsi3): New pattern
(subsi3): Likewise.
(negsi2): Likewise.
(mulsi3): Likewise.
(<optab>si3): New pattern for any_div.
(<optab>si3): New pattern for any_shift.
* loop-iv.cc (get_biv_step_1): Process src of extension when it
PLUS
gcc/testsuite/ChangeLog:
* testsuite/gcc.target/riscv/shift-and-2.c: New test
* testsuite/gcc.target/riscv/shift-shift-2.c: New test
* testsuite/gcc.target/riscv/sign-extend.c: New test
* testsuite/gcc.target/riscv/zbb-rol-ror-03.c: New test
--
With the best regards
Jivan Hakobyan
[-- Attachment #2: extend.diff --]
[-- Type: text/x-patch, Size: 20366 bytes --]
diff --git a/gcc/config/riscv/bitmanip.md b/gcc/config/riscv/bitmanip.md
index 96d31d92670b27d495dc5a9fbfc07e8767f40976..0430af7c95b1590308648dc4d5aaea78ada71760 100644
--- a/gcc/config/riscv/bitmanip.md
+++ b/gcc/config/riscv/bitmanip.md
@@ -304,9 +304,9 @@
[(set_attr "type" "bitmanip,load")
(set_attr "mode" "HI")])
-(define_expand "rotr<mode>3"
- [(set (match_operand:GPR 0 "register_operand")
- (rotatert:GPR (match_operand:GPR 1 "register_operand")
+(define_expand "rotrdi3"
+ [(set (match_operand:DI 0 "register_operand")
+ (rotatert:DI (match_operand:DI 1 "register_operand")
(match_operand:QI 2 "arith_operand")))]
"TARGET_ZBB || TARGET_XTHEADBB || TARGET_ZBKB"
{
@@ -322,6 +322,26 @@
"ror%i2%~\t%0,%1,%2"
[(set_attr "type" "bitmanip")])
+(define_expand "rotrsi3"
+ [(set (match_operand:SI 0 "register_operand" "=r")
+ (rotatert:SI (match_operand:SI 1 "register_operand" "r")
+ (match_operand:QI 2 "arith_operand" "rI")))]
+ "TARGET_ZBB || TARGET_ZBKB || TARGET_XTHEADBB"
+{
+ if (TARGET_XTHEADBB && !immediate_operand (operands[2], VOIDmode))
+ FAIL;
+ if (TARGET_64BIT && register_operand(operands[2], QImode))
+ {
+ rtx t = gen_reg_rtx (DImode);
+ emit_insn (gen_rotrsi3_sext (t, operands[1], operands[2]));
+ t = gen_lowpart (SImode, t);
+ SUBREG_PROMOTED_VAR_P (t) = 1;
+ SUBREG_PROMOTED_SET (t, SRP_SIGNED);
+ emit_move_insn (operands[0], t);
+ DONE;
+ }
+})
+
(define_insn "*rotrdi3"
[(set (match_operand:DI 0 "register_operand" "=r")
(rotatert:DI (match_operand:DI 1 "register_operand" "r")
@@ -330,7 +350,7 @@
"ror%i2\t%0,%1,%2"
[(set_attr "type" "bitmanip")])
-(define_insn "*rotrsi3_sext"
+(define_insn "rotrsi3_sext"
[(set (match_operand:DI 0 "register_operand" "=r")
(sign_extend:DI (rotatert:SI (match_operand:SI 1 "register_operand" "r")
(match_operand:QI 2 "arith_operand" "rI"))))]
@@ -338,7 +358,7 @@
"ror%i2%~\t%0,%1,%2"
[(set_attr "type" "bitmanip")])
-(define_insn "rotlsi3"
+(define_insn "*rotlsi3"
[(set (match_operand:SI 0 "register_operand" "=r")
(rotate:SI (match_operand:SI 1 "register_operand" "r")
(match_operand:QI 2 "register_operand" "r")))]
@@ -346,6 +366,24 @@
"rol%~\t%0,%1,%2"
[(set_attr "type" "bitmanip")])
+(define_expand "rotlsi3"
+ [(set (match_operand:SI 0 "register_operand" "=r")
+ (rotate:SI (match_operand:SI 1 "register_operand" "r")
+ (match_operand:QI 2 "register_operand" "r")))]
+ "TARGET_ZBB || TARGET_ZBKB"
+{
+ if (TARGET_64BIT)
+ {
+ rtx t = gen_reg_rtx (DImode);
+ emit_insn (gen_rotlsi3_sext (t, operands[1], operands[2]));
+ t = gen_lowpart (SImode, t);
+ SUBREG_PROMOTED_VAR_P (t) = 1;
+ SUBREG_PROMOTED_SET (t, SRP_SIGNED);
+ emit_move_insn (operands[0], t);
+ DONE;
+ }
+})
+
(define_insn "rotldi3"
[(set (match_operand:DI 0 "register_operand" "=r")
(rotate:DI (match_operand:DI 1 "register_operand" "r")
diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h
index 36419c95bbd8eebcb499ae0e02ca7aafde6c879f..de16ffd607e8e004e9b98ee9e25e4f3693818762 100644
--- a/gcc/config/riscv/riscv-protos.h
+++ b/gcc/config/riscv/riscv-protos.h
@@ -61,6 +61,7 @@ extern const char *riscv_output_return ();
extern void riscv_expand_int_scc (rtx, enum rtx_code, rtx, rtx);
extern void riscv_expand_float_scc (rtx, enum rtx_code, rtx, rtx);
extern void riscv_expand_conditional_branch (rtx, enum rtx_code, rtx, rtx);
+extern rtx riscv_emit_binary (enum rtx_code code, rtx dest, rtx x, rtx y);
#endif
extern bool riscv_expand_conditional_move (rtx, rtx, rtx, rtx);
extern rtx riscv_legitimize_call_address (rtx);
diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc
index 09fc9e5d95e611f94bc05b4851fef6f50a651c28..f14b128c9b429ee3b67c0b73e227e7a8d1f85a44 100644
--- a/gcc/config/riscv/riscv.cc
+++ b/gcc/config/riscv/riscv.cc
@@ -1404,7 +1404,7 @@ riscv_emit_set (rtx target, rtx src)
/* Emit an instruction of the form (set DEST (CODE X Y)). */
-static rtx
+rtx
riscv_emit_binary (enum rtx_code code, rtx dest, rtx x, rtx y)
{
return riscv_emit_set (dest, gen_rtx_fmt_ee (code, GET_MODE (dest), x, y));
diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md
index 124d8c95804b48a05b45b15a9cf926d27b13d029..39c0e4ec67fc7dd96b3e35f91dba26f683a6984e 100644
--- a/gcc/config/riscv/riscv.md
+++ b/gcc/config/riscv/riscv.md
@@ -512,7 +512,7 @@
[(set_attr "type" "fadd")
(set_attr "mode" "<UNITMODE>")])
-(define_insn "addsi3"
+(define_insn "*addsi3"
[(set (match_operand:SI 0 "register_operand" "=r,r")
(plus:SI (match_operand:SI 1 "register_operand" " r,r")
(match_operand:SI 2 "arith_operand" " r,I")))]
@@ -521,6 +521,24 @@
[(set_attr "type" "arith")
(set_attr "mode" "SI")])
+(define_expand "addsi3"
+ [(set (match_operand:SI 0 "register_operand" "=r,r")
+ (plus:SI (match_operand:SI 1 "register_operand" " r,r")
+ (match_operand:SI 2 "arith_operand" " r,I")))]
+ ""
+{
+ if (TARGET_64BIT)
+ {
+ rtx t = gen_reg_rtx (DImode);
+ emit_insn (gen_addsi3_extended (t, operands[1], operands[2]));
+ t = gen_lowpart (SImode, t);
+ SUBREG_PROMOTED_VAR_P (t) = 1;
+ SUBREG_PROMOTED_SET (t, SRP_SIGNED);
+ emit_move_insn (operands[0], t);
+ DONE;
+ }
+})
+
(define_insn "adddi3"
[(set (match_operand:DI 0 "register_operand" "=r,r")
(plus:DI (match_operand:DI 1 "register_operand" " r,r")
@@ -544,7 +562,7 @@
rtx t5 = gen_reg_rtx (DImode);
rtx t6 = gen_reg_rtx (DImode);
- emit_insn (gen_addsi3 (operands[0], operands[1], operands[2]));
+ riscv_emit_binary(PLUS, operands[0], operands[1], operands[2]);
if (GET_CODE (operands[1]) != CONST_INT)
emit_insn (gen_extend_insn (t4, operands[1], DImode, SImode, 0));
else
@@ -590,7 +608,7 @@
emit_insn (gen_extend_insn (t3, operands[1], DImode, SImode, 0));
else
t3 = operands[1];
- emit_insn (gen_addsi3 (operands[0], operands[1], operands[2]));
+ riscv_emit_binary(PLUS, operands[0], operands[1], operands[2]);
emit_insn (gen_extend_insn (t4, operands[0], DImode, SImode, 0));
riscv_expand_conditional_branch (operands[3], LTU, t4, t3);
@@ -605,7 +623,7 @@
DONE;
})
-(define_insn "*addsi3_extended"
+(define_insn "addsi3_extended"
[(set (match_operand:DI 0 "register_operand" "=r,r")
(sign_extend:DI
(plus:SI (match_operand:SI 1 "register_operand" " r,r")
@@ -652,7 +670,7 @@
[(set_attr "type" "arith")
(set_attr "mode" "DI")])
-(define_insn "subsi3"
+(define_insn "*subsi3"
[(set (match_operand:SI 0 "register_operand" "= r")
(minus:SI (match_operand:SI 1 "reg_or_0_operand" " rJ")
(match_operand:SI 2 "register_operand" " r")))]
@@ -661,6 +679,24 @@
[(set_attr "type" "arith")
(set_attr "mode" "SI")])
+(define_expand "subsi3"
+ [(set (match_operand:SI 0 "register_operand" "= r")
+ (minus:SI (match_operand:SI 1 "reg_or_0_operand" " rJ")
+ (match_operand:SI 2 "register_operand" " r")))]
+ ""
+{
+ if (TARGET_64BIT)
+ {
+ rtx t = gen_reg_rtx (DImode);
+ emit_insn (gen_subsi3_extended (t, operands[1], operands[2]));
+ t = gen_lowpart (SImode, t);
+ SUBREG_PROMOTED_VAR_P (t) = 1;
+ SUBREG_PROMOTED_SET (t, SRP_SIGNED);
+ emit_move_insn (operands[0], t);
+ DONE;
+ }
+})
+
(define_expand "subv<mode>4"
[(set (match_operand:GPR 0 "register_operand" "= r")
(minus:GPR (match_operand:GPR 1 "reg_or_0_operand" " rJ")
@@ -675,7 +711,7 @@
rtx t5 = gen_reg_rtx (DImode);
rtx t6 = gen_reg_rtx (DImode);
- emit_insn (gen_subsi3 (operands[0], operands[1], operands[2]));
+ riscv_emit_binary(MINUS, operands[0], operands[1], operands[2]);
if (GET_CODE (operands[1]) != CONST_INT)
emit_insn (gen_extend_insn (t4, operands[1], DImode, SImode, 0));
else
@@ -724,7 +760,7 @@
emit_insn (gen_extend_insn (t3, operands[1], DImode, SImode, 0));
else
t3 = operands[1];
- emit_insn (gen_subsi3 (operands[0], operands[1], operands[2]));
+ riscv_emit_binary(MINUS, operands[0], operands[1], operands[2]);
emit_insn (gen_extend_insn (t4, operands[0], DImode, SImode, 0));
riscv_expand_conditional_branch (operands[3], LTU, t3, t4);
@@ -740,7 +776,7 @@
})
-(define_insn "*subsi3_extended"
+(define_insn "subsi3_extended"
[(set (match_operand:DI 0 "register_operand" "= r")
(sign_extend:DI
(minus:SI (match_operand:SI 1 "reg_or_0_operand" " rJ")
@@ -769,7 +805,7 @@
[(set_attr "type" "arith")
(set_attr "mode" "DI")])
-(define_insn "negsi2"
+(define_insn "*negsi2"
[(set (match_operand:SI 0 "register_operand" "=r")
(neg:SI (match_operand:SI 1 "register_operand" " r")))]
""
@@ -777,7 +813,24 @@
[(set_attr "type" "arith")
(set_attr "mode" "SI")])
-(define_insn "*negsi2_extended"
+(define_expand "negsi2"
+ [(set (match_operand:SI 0 "register_operand" "=r")
+ (neg:SI (match_operand:SI 1 "register_operand" " r")))]
+ ""
+{
+ if (TARGET_64BIT)
+ {
+ rtx t = gen_reg_rtx (DImode);
+ emit_insn (gen_negsi2_extended (t, operands[1]));
+ t = gen_lowpart (SImode, t);
+ SUBREG_PROMOTED_VAR_P (t) = 1;
+ SUBREG_PROMOTED_SET (t, SRP_SIGNED);
+ emit_move_insn (operands[0], t);
+ DONE;
+ }
+})
+
+(define_insn "negsi2_extended"
[(set (match_operand:DI 0 "register_operand" "=r")
(sign_extend:DI
(neg:SI (match_operand:SI 1 "register_operand" " r"))))]
@@ -813,7 +866,7 @@
[(set_attr "type" "fmul")
(set_attr "mode" "<UNITMODE>")])
-(define_insn "mulsi3"
+(define_insn "*mulsi3"
[(set (match_operand:SI 0 "register_operand" "=r")
(mult:SI (match_operand:SI 1 "register_operand" " r")
(match_operand:SI 2 "register_operand" " r")))]
@@ -822,6 +875,24 @@
[(set_attr "type" "imul")
(set_attr "mode" "SI")])
+(define_expand "mulsi3"
+ [(set (match_operand:SI 0 "register_operand" "=r")
+ (mult:SI (match_operand:SI 1 "register_operand" " r")
+ (match_operand:SI 2 "register_operand" " r")))]
+ "TARGET_ZMMUL || TARGET_MUL"
+{
+ if (TARGET_64BIT)
+ {
+ rtx t = gen_reg_rtx (DImode);
+ emit_insn (gen_mulsi3_extended (t, operands[1], operands[2]));
+ t = gen_lowpart (SImode, t);
+ SUBREG_PROMOTED_VAR_P (t) = 1;
+ SUBREG_PROMOTED_SET (t, SRP_SIGNED);
+ emit_move_insn (operands[0], t);
+ DONE;
+ }
+})
+
(define_insn "muldi3"
[(set (match_operand:DI 0 "register_operand" "=r")
(mult:DI (match_operand:DI 1 "register_operand" " r")
@@ -867,8 +938,8 @@
emit_insn (gen_smul<mode>3_highpart (hp, operands[1], operands[2]));
emit_insn (gen_mul<mode>3 (operands[0], operands[1], operands[2]));
- emit_insn (gen_ashr<mode>3 (lp, operands[0],
- GEN_INT (BITS_PER_WORD - 1)));
+ riscv_emit_binary(ASHIFTRT, lp, operands[0],
+ GEN_INT (BITS_PER_WORD - 1));
riscv_expand_conditional_branch (operands[3], NE, hp, lp);
}
@@ -922,7 +993,7 @@
DONE;
})
-(define_insn "*mulsi3_extended"
+(define_insn "mulsi3_extended"
[(set (match_operand:DI 0 "register_operand" "=r")
(sign_extend:DI
(mult:SI (match_operand:SI 1 "register_operand" " r")
@@ -1023,7 +1094,7 @@
"(TARGET_ZMMUL || TARGET_MUL) && !TARGET_64BIT"
{
rtx temp = gen_reg_rtx (SImode);
- emit_insn (gen_mulsi3 (temp, operands[1], operands[2]));
+ riscv_emit_binary(MULT, temp, operands[1], operands[2]);
emit_insn (gen_<su>mulsi3_highpart (riscv_subword (operands[0], true),
operands[1], operands[2]));
emit_insn (gen_movsi (riscv_subword (operands[0], false), temp));
@@ -1054,7 +1125,7 @@
"(TARGET_ZMMUL || TARGET_MUL) && !TARGET_64BIT"
{
rtx temp = gen_reg_rtx (SImode);
- emit_insn (gen_mulsi3 (temp, operands[1], operands[2]));
+ riscv_emit_binary(MULT, temp, operands[1], operands[2]);
emit_insn (gen_usmulsi3_highpart (riscv_subword (operands[0], true),
operands[1], operands[2]));
emit_insn (gen_movsi (riscv_subword (operands[0], false), temp));
@@ -1083,7 +1154,7 @@
;; ....................
;;
-(define_insn "<optab>si3"
+(define_insn "*<optab>si3"
[(set (match_operand:SI 0 "register_operand" "=r")
(any_div:SI (match_operand:SI 1 "register_operand" " r")
(match_operand:SI 2 "register_operand" " r")))]
@@ -1092,6 +1163,24 @@
[(set_attr "type" "idiv")
(set_attr "mode" "SI")])
+(define_expand "<optab>si3"
+ [(set (match_operand:SI 0 "register_operand" "=r")
+ (any_div:SI (match_operand:SI 1 "register_operand" " r")
+ (match_operand:SI 2 "register_operand" " r")))]
+ "TARGET_DIV"
+{
+ if (TARGET_64BIT)
+ {
+ rtx t = gen_reg_rtx (DImode);
+ emit_insn (gen_<optab>si3_extended (t, operands[1], operands[2]));
+ t = gen_lowpart (SImode, t);
+ SUBREG_PROMOTED_VAR_P (t) = 1;
+ SUBREG_PROMOTED_SET (t, SRP_SIGNED);
+ emit_move_insn (operands[0], t);
+ DONE;
+ }
+})
+
(define_insn "<optab>di3"
[(set (match_operand:DI 0 "register_operand" "=r")
(any_div:DI (match_operand:DI 1 "register_operand" " r")
@@ -1117,7 +1206,7 @@
DONE;
})
-(define_insn "*<optab>si3_extended"
+(define_insn "<optab>si3_extended"
[(set (match_operand:DI 0 "register_operand" "=r")
(sign_extend:DI
(any_div:SI (match_operand:SI 1 "register_operand" " r")
@@ -2032,7 +2121,7 @@
;; expand_shift_1 can do this automatically when SHIFT_COUNT_TRUNCATED is
;; defined, but use of that is discouraged.
-(define_insn "<optab>si3"
+(define_insn "*<optab>si3"
[(set (match_operand:SI 0 "register_operand" "= r")
(any_shift:SI
(match_operand:SI 1 "register_operand" " r")
@@ -2048,6 +2137,24 @@
[(set_attr "type" "shift")
(set_attr "mode" "SI")])
+(define_expand "<optab>si3"
+ [(set (match_operand:SI 0 "register_operand" "= r")
+ (any_shift:SI (match_operand:SI 1 "register_operand" " r")
+ (match_operand:QI 2 "arith_operand" " rI")))]
+ ""
+{
+ if (TARGET_64BIT)
+ {
+ rtx t = gen_reg_rtx (DImode);
+ emit_insn (gen_<optab>si3_extend (t, operands[1], operands[2]));
+ t = gen_lowpart (SImode, t);
+ SUBREG_PROMOTED_VAR_P (t) = 1;
+ SUBREG_PROMOTED_SET (t, SRP_SIGNED);
+ emit_move_insn (operands[0], t);
+ DONE;
+ }
+})
+
(define_insn "<optab>di3"
[(set (match_operand:DI 0 "register_operand" "= r")
(any_shift:DI
@@ -2082,7 +2189,7 @@
[(set_attr "type" "shift")
(set_attr "mode" "<GPR:MODE>")])
-(define_insn "*<optab>si3_extend"
+(define_insn "<optab>si3_extend"
[(set (match_operand:DI 0 "register_operand" "= r")
(sign_extend:DI
(any_shift:SI (match_operand:SI 1 "register_operand" " r")
diff --git a/gcc/loop-iv.cc b/gcc/loop-iv.cc
index 6c40db947f7f549303f8bb4d4f38aa98b6561bcc..bec1ea7e4ccf7291bb3dba91161f948e66c7bea9 100644
--- a/gcc/loop-iv.cc
+++ b/gcc/loop-iv.cc
@@ -637,7 +637,7 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg,
{
rtx set, rhs, op0 = NULL_RTX, op1 = NULL_RTX;
rtx next, nextr;
- enum rtx_code code;
+ enum rtx_code code, prev_code;
rtx_insn *insn = DF_REF_INSN (def);
df_ref next_def;
enum iv_grd_result res;
@@ -697,6 +697,23 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg,
return false;
op0 = XEXP (rhs, 0);
+
+ if (GET_CODE (op0) == PLUS)
+ {
+ rhs = op0;
+ op0 = XEXP (rhs, 0);
+ op1 = XEXP (rhs, 1);
+
+ if (CONSTANT_P (op0))
+ std::swap (op0, op1);
+
+ if (!simple_reg_p (op0) || !CONSTANT_P (op1))
+ return false;
+
+ prev_code = code;
+ code = PLUS;
+ }
+
if (!simple_reg_p (op0))
return false;
@@ -769,6 +786,11 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg,
else
*outer_step = simplify_gen_binary (code, outer_mode,
*outer_step, op1);
+
+ if (prev_code == SIGN_EXTEND)
+ *extend = IV_SIGN_EXTEND;
+ else if (prev_code == ZERO_EXTEND)
+ *extend = IV_ZERO_EXTEND;
break;
case SIGN_EXTEND:
diff --git a/gcc/testsuite/gcc.target/riscv/shift-and-2.c b/gcc/testsuite/gcc.target/riscv/shift-and-2.c
index bc01e8ef99268ff154f0759620d501cadf338561..ee9925b749859950bd90de07aa63680bd3d8a5f2 100644
--- a/gcc/testsuite/gcc.target/riscv/shift-and-2.c
+++ b/gcc/testsuite/gcc.target/riscv/shift-and-2.c
@@ -38,5 +38,24 @@ sub6 (long i, long j)
{
return i << (j & 0x3f);
}
+
+/* Test for <optab>si3_extend. */
+int
+sub7 (int i, int j) {
+ return (i << 10) & j;
+}
+
+/* Test for <optab>si3_extend. */
+unsigned
+sub8 (unsigned i, unsigned j) {
+ return (i << 10) & j;
+}
+
+/* Test for <optab>si3_extend. */
+unsigned
+sub9 (unsigned i, unsigned j) {
+ return (i >> 10) & j;
+}
+
/* { dg-final { scan-assembler-not "andi" } } */
/* { dg-final { scan-assembler-not "sext.w" } } */
diff --git a/gcc/testsuite/gcc.target/riscv/shift-shift-2.c b/gcc/testsuite/gcc.target/riscv/shift-shift-2.c
index 5f93be15ac5dc5234b75fb4004abd1fb4c37124d..bc8c4ef382897d849697e525986c4cdfa82f3699 100644
--- a/gcc/testsuite/gcc.target/riscv/shift-shift-2.c
+++ b/gcc/testsuite/gcc.target/riscv/shift-shift-2.c
@@ -38,5 +38,6 @@ sub5 (unsigned int i)
}
/* { dg-final { scan-assembler-times "slli" 5 } } */
/* { dg-final { scan-assembler-times "srli" 5 } } */
-/* { dg-final { scan-assembler-times "slliw" 1 } } */
-/* { dg-final { scan-assembler-times "srliw" 1 } } */
+/* { dg-final { scan-assembler-times ",40" 2 } } */ /* For sub5 test */
+/* { dg-final { scan-assembler-not "slliw" } } */
+/* { dg-final { scan-assembler-not "srliw" } } */
diff --git a/gcc/testsuite/gcc.target/riscv/sign-extend.c b/gcc/testsuite/gcc.target/riscv/sign-extend.c
new file mode 100644
index 0000000000000000000000000000000000000000..6f8401948330d7772dce2fe5ae1399696896a534
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/sign-extend.c
@@ -0,0 +1,81 @@
+/* { dg-do compile { target { riscv64*-*-* } } } */
+/* { dg-options "-march=rv64gc -mabi=lp64" } */
+/* { dg-skip-if "" { *-*-* } { "-O0" } } */
+
+unsigned
+foo1 (unsigned x, unsigned y, unsigned z)
+{
+ return x & (y - z);
+}
+
+int
+foo2 (int x, int y, int z)
+{
+ return x & (y - z);
+}
+
+unsigned
+foo3 (unsigned x, unsigned y, unsigned z)
+{
+ return x & (y * z);
+}
+
+int
+foo4 (int x, int y, int z)
+{
+ return x & (y * z);
+}
+
+unsigned
+foo5 (unsigned x, unsigned y)
+{
+ return x & (y / x);
+}
+
+int
+foo6 (int x, int y)
+{
+ return x & (y / x);
+}
+
+unsigned
+foo7 (unsigned x, unsigned y)
+{
+ return x & (y % x);
+}
+
+int
+foo8 (int x, int y)
+{
+ return x & (y % x);
+}
+
+int
+foo9 (int x)
+{
+ return x & (-x);
+}
+
+unsigned
+foo10 (unsigned x, unsigned y)
+{
+ return x & (y + x);
+}
+
+
+unsigned
+foo11 (unsigned x)
+{
+ return x & (15 + x);
+}
+
+/* { dg-final { scan-assembler-times "subw" 2 } } */
+/* { dg-final { scan-assembler-times "addw" 1 } } */
+/* { dg-final { scan-assembler-times "addiw" 1 } } */
+/* { dg-final { scan-assembler-times "mulw" 2 } } */
+/* { dg-final { scan-assembler-times "divw" 1 } } */
+/* { dg-final { scan-assembler-times "divuw" 1 } } */
+/* { dg-final { scan-assembler-times "remw" 1 } } */
+/* { dg-final { scan-assembler-times "remuw" 1 } } */
+/* { dg-final { scan-assembler-times "negw" 1 } } */
+/* { dg-final { scan-assembler-not "sext.w" } } */
diff --git a/gcc/testsuite/gcc.target/riscv/zbb-rol-ror-03.c b/gcc/testsuite/gcc.target/riscv/zbb-rol-ror-03.c
index b44d7fe8920b3a6bdb81329626f20a508d0ca277..e7e5cbb9a1ab7b0af5b85c71fd4e2bd56aca3c00 100644
--- a/gcc/testsuite/gcc.target/riscv/zbb-rol-ror-03.c
+++ b/gcc/testsuite/gcc.target/riscv/zbb-rol-ror-03.c
@@ -16,4 +16,5 @@ unsigned int ror(unsigned int rs1, unsigned int rs2)
/* { dg-final { scan-assembler-times "rolw" 1 } } */
/* { dg-final { scan-assembler-times "rorw" 1 } } */
-/* { dg-final { scan-assembler-not "and" } } */
\ No newline at end of file
+/* { dg-final { scan-assembler-not "and" } } */
+/* { dg-final { scan-assembler-not "sext.w" } } */
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [RFC] RISC-V: Eliminate extension after for *w instructions 2023-05-24 23:14 [RFC] RISC-V: Eliminate extension after for *w instructions Jivan Hakobyan @ 2023-05-31 0:49 ` Jeff Law 2023-06-04 3:59 ` Jeff Law 2023-06-07 20:00 ` Jeff Law 2 siblings, 0 replies; 4+ messages in thread From: Jeff Law @ 2023-05-31 0:49 UTC (permalink / raw) To: Jivan Hakobyan, gcc-patches On 5/24/23 17:14, Jivan Hakobyan via Gcc-patches wrote: > > gcc/ChangeLog: > * config/riscv/bitmanip.md (rotrdi3): New pattern. > (rotrsi3): Likewise. > (rotlsi3): Likewise. > * config/riscv/riscv-protos.h (riscv_emit_binary): New function > declaration > * config/riscv/riscv.cc (riscv_emit_binary): Removed static > * config/riscv/riscv.md (addsi3): New pattern > (subsi3): Likewise. > (negsi2): Likewise. > (mulsi3): Likewise. > (<optab>si3): New pattern for any_div. > (<optab>si3): New pattern for any_shift. > * loop-iv.cc (get_biv_step_1): Process src of extension when it > PLUS > > gcc/testsuite/ChangeLog: > * testsuite/gcc.target/riscv/shift-and-2.c: New test > * testsuite/gcc.target/riscv/shift-shift-2.c: New test > * testsuite/gcc.target/riscv/sign-extend.c: New test > * testsuite/gcc.target/riscv/zbb-rol-ror-03.c: New test > > > -- With the best regards Jivan Hakobyan > > > diff --git a/gcc/loop-iv.cc b/gcc/loop-iv.cc > index 6c40db947f7f549303f8bb4d4f38aa98b6561bcc..bec1ea7e4ccf7291bb3dba91161f948e66c7bea9 100644 > --- a/gcc/loop-iv.cc > +++ b/gcc/loop-iv.cc > @@ -637,7 +637,7 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg, > { > rtx set, rhs, op0 = NULL_RTX, op1 = NULL_RTX; > rtx next, nextr; > - enum rtx_code code; > + enum rtx_code code, prev_code; > rtx_insn *insn = DF_REF_INSN (def); > df_ref next_def; > enum iv_grd_result res; > @@ -697,6 +697,23 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg, > return false; > > op0 = XEXP (rhs, 0); > + > + if (GET_CODE (op0) == PLUS) > + { > + rhs = op0; > + op0 = XEXP (rhs, 0); > + op1 = XEXP (rhs, 1); > + > + if (CONSTANT_P (op0)) > + std::swap (op0, op1); > + > + if (!simple_reg_p (op0) || !CONSTANT_P (op1)) > + return false; > + > + prev_code = code; > + code = PLUS; > + } > + > if (!simple_reg_p (op0)) > return false; > > @@ -769,6 +786,11 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg, > else > *outer_step = simplify_gen_binary (code, outer_mode, > *outer_step, op1); > + > + if (prev_code == SIGN_EXTEND) > + *extend = IV_SIGN_EXTEND; > + else if (prev_code == ZERO_EXTEND) > + *extend = IV_ZERO_EXTEND; > break; So I'm still working through the basics of loop-iv.cc, but one thing I did notice is that prev_code may be used without being properly initialized. You should probably initialize it to "UNKNOWN". Jeff ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [RFC] RISC-V: Eliminate extension after for *w instructions 2023-05-24 23:14 [RFC] RISC-V: Eliminate extension after for *w instructions Jivan Hakobyan 2023-05-31 0:49 ` Jeff Law @ 2023-06-04 3:59 ` Jeff Law 2023-06-07 20:00 ` Jeff Law 2 siblings, 0 replies; 4+ messages in thread From: Jeff Law @ 2023-06-04 3:59 UTC (permalink / raw) To: Jivan Hakobyan, gcc-patches On 5/24/23 17:14, Jivan Hakobyan via Gcc-patches wrote: > `This patch tries to prevent generating unnecessary sign extension > after *w instructions like "addiw" or "divw". > > The main idea of it is to add SUBREG_PROMOTED fields during expanding. > > I have tested on SPEC2017 there is no regression. > Only gcc.dg/pr30957-1.c test failed. > To solve that I did some changes in loop-iv.cc, but not sure that it is > suitable. So this generally looks good and I did some playing around with it over the weekend. It's generally a win, though it can result in performance regressions under the "right" circumstances. The case I was looking at was omnetpp in spec. I suspect you didn't see it because we have a parameter which increases the threshold for when a string comparison against a constant string should be inlined. I suspect you aren't using that param. I inherited that param usage and haven't tried to eliminate it yet. If you compile a test like this: #include <string.h> int foo (char *x) { return strcmp (x, "lowerLayout"); } With -O2 --param builtin-string-cmp-inline-length=100 You'll see the regression. This is reasonable representation of the code in omnetpp. It's actually a fairly interesting little problem. We end up with overlapping lifetimes for the 32 and 64bit results. The register allocator doesn't really try hard to detect the case where it can ignore a conflict because the two objects hold the same value. I remember kicking this around with Vlad at least 10 years ago and we concluded (based on experience and data at the time) that this case wasn't that important to handle. Anyway... Another approach to this problem would be to twiddle the strcmp expander to work in word_mode, then convert the word mode result to the final mode at the end of the sequence. I need to ponder the semantics of this a bit more, but if the semantics are right, it seems like it might be a viable solution. I briefly looked at the big improvement in leela (33b instructions, roughly 1.7% of the dynamic count removed). The hope was that if I looked at the cases where we improved that they would all be shifts, rotates and the like and we could consider a more limited version of your patch. But it was quite clear that the improvement in leela as due to better handling of 32bit additions. So that idea's a non-starter. I didn't look at the big gains in xz (they're smaller in absolute count terms, but larger in percentage of instructions removed). So I'm going to play a bit more with the expander for comparisons against constant strings. If done correctly it might actually be an improvement on other targets as well. Jeff > > > gcc/ChangeLog: > * config/riscv/bitmanip.md (rotrdi3): New pattern. > (rotrsi3): Likewise. > (rotlsi3): Likewise. > * config/riscv/riscv-protos.h (riscv_emit_binary): New function > declaration > * config/riscv/riscv.cc (riscv_emit_binary): Removed static > * config/riscv/riscv.md (addsi3): New pattern > (subsi3): Likewise. > (negsi2): Likewise. > (mulsi3): Likewise. > (<optab>si3): New pattern for any_div. > (<optab>si3): New pattern for any_shift. > * loop-iv.cc (get_biv_step_1): Process src of extension when it > PLUS > > gcc/testsuite/ChangeLog: > * testsuite/gcc.target/riscv/shift-and-2.c: New test > * testsuite/gcc.target/riscv/shift-shift-2.c: New test > * testsuite/gcc.target/riscv/sign-extend.c: New test > * testsuite/gcc.target/riscv/zbb-rol-ror-03.c: New test > > ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [RFC] RISC-V: Eliminate extension after for *w instructions 2023-05-24 23:14 [RFC] RISC-V: Eliminate extension after for *w instructions Jivan Hakobyan 2023-05-31 0:49 ` Jeff Law 2023-06-04 3:59 ` Jeff Law @ 2023-06-07 20:00 ` Jeff Law 2 siblings, 0 replies; 4+ messages in thread From: Jeff Law @ 2023-06-07 20:00 UTC (permalink / raw) To: Jivan Hakobyan, gcc-patches [-- Attachment #1: Type: text/plain, Size: 7068 bytes --] On 5/24/23 17:14, Jivan Hakobyan via Gcc-patches wrote: > Subject: > [RFC] RISC-V: Eliminate extension after for *w instructions > From: > Jivan Hakobyan via Gcc-patches <gcc-patches@gcc.gnu.org> > Date: > 5/24/23, 17:14 > > To: > gcc-patches@gcc.gnu.org > > > `This patch tries to prevent generating unnecessary sign extension > after *w instructions like "addiw" or "divw". > > The main idea of it is to add SUBREG_PROMOTED fields during expanding. > > I have tested on SPEC2017 there is no regression. > Only gcc.dg/pr30957-1.c test failed. > To solve that I did some changes in loop-iv.cc, but not sure that it is > suitable. > > > gcc/ChangeLog: > * config/riscv/bitmanip.md (rotrdi3): New pattern. > (rotrsi3): Likewise. > (rotlsi3): Likewise. > * config/riscv/riscv-protos.h (riscv_emit_binary): New function > declaration > * config/riscv/riscv.cc (riscv_emit_binary): Removed static > * config/riscv/riscv.md (addsi3): New pattern > (subsi3): Likewise. > (negsi2): Likewise. > (mulsi3): Likewise. > (<optab>si3): New pattern for any_div. > (<optab>si3): New pattern for any_shift. > * loop-iv.cc (get_biv_step_1): Process src of extension when it > PLUS > > gcc/testsuite/ChangeLog: > * testsuite/gcc.target/riscv/shift-and-2.c: New test > * testsuite/gcc.target/riscv/shift-shift-2.c: New test > * testsuite/gcc.target/riscv/sign-extend.c: New test > * testsuite/gcc.target/riscv/zbb-rol-ror-03.c: New test > > > -- With the best regards Jivan Hakobyan > > > extend.diff > > diff --git a/gcc/config/riscv/bitmanip.md b/gcc/config/riscv/bitmanip.md > index 96d31d92670b27d495dc5a9fbfc07e8767f40976..0430af7c95b1590308648dc4d5aaea78ada71760 100644 > --- a/gcc/config/riscv/bitmanip.md > +++ b/gcc/config/riscv/bitmanip.md > @@ -304,9 +304,9 @@ > [(set_attr "type" "bitmanip,load") > (set_attr "mode" "HI")]) > > -(define_expand "rotr<mode>3" > - [(set (match_operand:GPR 0 "register_operand") > - (rotatert:GPR (match_operand:GPR 1 "register_operand") > +(define_expand "rotrdi3" > + [(set (match_operand:DI 0 "register_operand") > + (rotatert:DI (match_operand:DI 1 "register_operand") > (match_operand:QI 2 "arith_operand")))] > "TARGET_ZBB || TARGET_XTHEADBB || TARGET_ZBKB" The condition for this expander needs to be adjusted. Previously it used the GPR iterator. The GPR iterator is defined like this: (define_mode_iterator GPR [SI (DI "TARGET_64BIT")]) Note how the DI case is conditional on TARGET_64BIT. This impacts the HAVE_* macros that are generated from the MD file in insn-flags.h: #define HAVE_rotrsi3 (TARGET_ZBB || TARGET_XTHEADBB || TARGET_ZBKB) #define HAVE_rotrdi3 ((TARGET_ZBB || TARGET_XTHEADBB || TARGET_ZBKB) && (TARGET_64BIT)) Note how the rotrdi3 has the && (TARGET_64BIT) on the end. With your change we would expose rotrdi3 independent of TARGET_64BIT which is not what we want. Sorry I didn't catch that earlier. I'll fix this minor problem. > @@ -544,7 +562,7 @@ > rtx t5 = gen_reg_rtx (DImode); > rtx t6 = gen_reg_rtx (DImode); > > - emit_insn (gen_addsi3 (operands[0], operands[1], operands[2])); > + riscv_emit_binary(PLUS, operands[0], operands[1], operands[2]); Just a note. In GCC we always emit a space between the function name and the open parenthesis for its argument list. I fixed a few of these. > @@ -867,8 +938,8 @@ > > emit_insn (gen_smul<mode>3_highpart (hp, operands[1], operands[2])); > emit_insn (gen_mul<mode>3 (operands[0], operands[1], operands[2])); > - emit_insn (gen_ashr<mode>3 (lp, operands[0], > - GEN_INT (BITS_PER_WORD - 1))); > + riscv_emit_binary(ASHIFTRT, lp, operands[0], > + GEN_INT (BITS_PER_WORD - 1)); Another formatting nit. When we wrap lines for an argument list, we line up the arguments. So something like this frobit (a, b, c d, e, f); Obviously that's not a great example as it doesn't need wrapping, but it should clearly show how we indent things in this case. I've fixed up this nit. > diff --git a/gcc/loop-iv.cc b/gcc/loop-iv.cc > index 6c40db947f7f549303f8bb4d4f38aa98b6561bcc..bec1ea7e4ccf7291bb3dba91161f948e66c7bea9 100644 > --- a/gcc/loop-iv.cc > +++ b/gcc/loop-iv.cc > @@ -637,7 +637,7 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg, > { > rtx set, rhs, op0 = NULL_RTX, op1 = NULL_RTX; > rtx next, nextr; > - enum rtx_code code; > + enum rtx_code code, prev_code; So as I mentioned earlier, PREV_CODE might be used without being initialized. I've initialized it to "UNKNOWN" which is a special RTX code which can be used for this purpose. If we are changing a target independent file the standard is that we bootstrap and regression test on at least one primary platform such as x86_64 linux. This would have been caught by that bootstrap process as it's a pretty simple uninitialized object use to analyze. > rtx_insn *insn = DF_REF_INSN (def); > df_ref next_def; > enum iv_grd_result res; > @@ -697,6 +697,23 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg, > return false; > > op0 = XEXP (rhs, 0); > + > + if (GET_CODE (op0) == PLUS) I've added a few comments before this code indicating why it was added. And WRT formatting, we strongly prefer to use a tab rather than 8 spaces. I think it's pretty annoying myself, but I live with it as it's the standard for the GCC project. I've fixed these nits as well. WRT the impacts. We haven't run this on the emulator, but we can easily look at the instruction counts for spec2017. It's a clear improvement with the biggest improvements in the xz benchmark (1-2%), with others in the .5% range. There is one regression of note with how we expand comparisons against string constants. I'm going to file a bug for that so it doesn't get lost. In my opinion the benefits of this patch outweigh the impact of that minor regression in omnetpp. I'll file a bug for the omnetpp regression. What will be really interesting will be to test the TARGET_REP_EXTENDED patch from Philipp again. I wouldn't be surprised at all if much of the benefit of TARGET_REP_EXTENDED is negated by exposing the target's actual semantics like you've done. This regressed the zba-shNadd-05 test, which isn't a big surprise given this test was designed to test a pattern which is sensitive to the structure of 32bit arithmetic. I've adjusted the appropriate define_insn_and_split so that it matches the updated RTL. There's a couple of other splitters in bitmanip.md which may need similar treatment. But I don't see cases for them in the testsuite, so it's hard to know if those splitters are useful and to test if we've got any adjustments to them correct. Bootstrapped and regression tested on x86 linux. Regression tested on riscv64 as well. Attached is the version I pushed to the trunk. Thanks again, Jeff [-- Attachment #2: P --] [-- Type: text/plain, Size: 23821 bytes --] commit 99bfdb072e67fa3fe294d86b4b2a9f686f8d9705 Author: Jeff Law <jlaw@ventanamicro.com> Date: Wed Jun 7 13:40:16 2023 -0600 RISC-V: Eliminate extension after for *w instructions This patch tries to prevent generating unnecessary sign extension after *w instructions like "addiw" or "divw". The main idea of it is to add SUBREG_PROMOTED fields during expanding. I have tested on SPEC2017 there is no regression. Only gcc.dg/pr30957-1.c test failed. To solve that I did some changes in loop-iv.cc, but not sure that it is suitable. gcc/ChangeLog: * config/riscv/bitmanip.md (rotrdi3, rotrsi3, rotlsi3): New expanders. (rotrsi3_sext): Expose generator. (rotlsi3 pattern): Hide generator. * config/riscv/riscv-protos.h (riscv_emit_binary): New function declaration. * config/riscv/riscv.cc (riscv_emit_binary): Removed static * config/riscv/riscv.md (addsi3, subsi3, negsi2): Hide generator. (mulsi3, <optab>si3): Likewise. (addsi3, subsi3, negsi2, mulsi3, <optab>si3): New expanders. (addv<mode>4, subv<mode>4, mulv<mode>4): Use riscv_emit_binary. (<u>mulsidi3): Likewise. (addsi3_extended, subsi3_extended, negsi2_extended): Expose generator. (mulsi3_extended, <optab>si3_extended): Likewise. (splitter for shadd feeding divison): Update RTL pattern to account for changes in how 32 bit ops are expanded for TARGET_64BIT. * loop-iv.cc (get_biv_step_1): Process src of extension when it PLUS. gcc/testsuite/ChangeLog: * gcc.target/riscv/shift-and-2.c: New tests. * gcc.target/riscv/shift-shift-2.c: Adjust expected output. * gcc.target/riscv/sign-extend.c: New test. * gcc.target/riscv/zbb-rol-ror-03.c: Adjust expected output. Co-authored-by: Jeff Law <jlaw@ventanamicro.com> diff --git a/gcc/config/riscv/bitmanip.md b/gcc/config/riscv/bitmanip.md index 96d31d92670..c42e7b890db 100644 --- a/gcc/config/riscv/bitmanip.md +++ b/gcc/config/riscv/bitmanip.md @@ -47,10 +47,10 @@ (define_insn "*shNadd" ; implicit sign-extensions. (define_split [(set (match_operand:DI 0 "register_operand") - (sign_extend:DI (div:SI (plus:SI (subreg:SI (ashift:DI (match_operand:DI 1 "register_operand") - (match_operand:QI 2 "imm123_operand")) 0) - (subreg:SI (match_operand:DI 3 "register_operand") 0)) - (subreg:SI (match_operand:DI 4 "register_operand") 0)))) + (sign_extend:DI (div:SI (plus:SI (ashift:SI (subreg:SI (match_operand:DI 1 "register_operand") 0) + (match_operand:QI 2 "imm123_operand")) + (subreg:SI (match_operand:DI 3 "register_operand") 0)) + (subreg:SI (match_operand:DI 4 "register_operand") 0)))) (clobber (match_operand:DI 5 "register_operand"))] "TARGET_64BIT && TARGET_ZBA" [(set (match_dup 5) (plus:DI (ashift:DI (match_dup 1) (match_dup 2)) (match_dup 3))) @@ -304,11 +304,11 @@ (define_insn "*zero_extendhi<GPR:mode>2_zbb" [(set_attr "type" "bitmanip,load") (set_attr "mode" "HI")]) -(define_expand "rotr<mode>3" - [(set (match_operand:GPR 0 "register_operand") - (rotatert:GPR (match_operand:GPR 1 "register_operand") +(define_expand "rotrdi3" + [(set (match_operand:DI 0 "register_operand") + (rotatert:DI (match_operand:DI 1 "register_operand") (match_operand:QI 2 "arith_operand")))] - "TARGET_ZBB || TARGET_XTHEADBB || TARGET_ZBKB" + "TARGET_64BIT && (TARGET_ZBB || TARGET_XTHEADBB || TARGET_ZBKB)" { if (TARGET_XTHEADBB && !immediate_operand (operands[2], VOIDmode)) FAIL; @@ -322,6 +322,26 @@ (define_insn "*rotrsi3" "ror%i2%~\t%0,%1,%2" [(set_attr "type" "bitmanip")]) +(define_expand "rotrsi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (rotatert:SI (match_operand:SI 1 "register_operand" "r") + (match_operand:QI 2 "arith_operand" "rI")))] + "TARGET_ZBB || TARGET_ZBKB || TARGET_XTHEADBB" +{ + if (TARGET_XTHEADBB && !immediate_operand (operands[2], VOIDmode)) + FAIL; + if (TARGET_64BIT && register_operand (operands[2], QImode)) + { + rtx t = gen_reg_rtx (DImode); + emit_insn (gen_rotrsi3_sext (t, operands[1], operands[2])); + t = gen_lowpart (SImode, t); + SUBREG_PROMOTED_VAR_P (t) = 1; + SUBREG_PROMOTED_SET (t, SRP_SIGNED); + emit_move_insn (operands[0], t); + DONE; + } +}) + (define_insn "*rotrdi3" [(set (match_operand:DI 0 "register_operand" "=r") (rotatert:DI (match_operand:DI 1 "register_operand" "r") @@ -330,7 +350,7 @@ (define_insn "*rotrdi3" "ror%i2\t%0,%1,%2" [(set_attr "type" "bitmanip")]) -(define_insn "*rotrsi3_sext" +(define_insn "rotrsi3_sext" [(set (match_operand:DI 0 "register_operand" "=r") (sign_extend:DI (rotatert:SI (match_operand:SI 1 "register_operand" "r") (match_operand:QI 2 "arith_operand" "rI"))))] @@ -338,7 +358,7 @@ (define_insn "*rotrsi3_sext" "ror%i2%~\t%0,%1,%2" [(set_attr "type" "bitmanip")]) -(define_insn "rotlsi3" +(define_insn "*rotlsi3" [(set (match_operand:SI 0 "register_operand" "=r") (rotate:SI (match_operand:SI 1 "register_operand" "r") (match_operand:QI 2 "register_operand" "r")))] @@ -346,6 +366,24 @@ (define_insn "rotlsi3" "rol%~\t%0,%1,%2" [(set_attr "type" "bitmanip")]) +(define_expand "rotlsi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (rotate:SI (match_operand:SI 1 "register_operand" "r") + (match_operand:QI 2 "register_operand" "r")))] + "TARGET_ZBB || TARGET_ZBKB" +{ + if (TARGET_64BIT) + { + rtx t = gen_reg_rtx (DImode); + emit_insn (gen_rotlsi3_sext (t, operands[1], operands[2])); + t = gen_lowpart (SImode, t); + SUBREG_PROMOTED_VAR_P (t) = 1; + SUBREG_PROMOTED_SET (t, SRP_SIGNED); + emit_move_insn (operands[0], t); + DONE; + } +}) + (define_insn "rotldi3" [(set (match_operand:DI 0 "register_operand" "=r") (rotate:DI (match_operand:DI 1 "register_operand" "r") diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h index 9782f1794fb..38e4125424b 100644 --- a/gcc/config/riscv/riscv-protos.h +++ b/gcc/config/riscv/riscv-protos.h @@ -61,6 +61,7 @@ extern const char *riscv_output_return (); extern void riscv_expand_int_scc (rtx, enum rtx_code, rtx, rtx); extern void riscv_expand_float_scc (rtx, enum rtx_code, rtx, rtx); extern void riscv_expand_conditional_branch (rtx, enum rtx_code, rtx, rtx); +extern rtx riscv_emit_binary (enum rtx_code code, rtx dest, rtx x, rtx y); #endif extern bool riscv_expand_conditional_move (rtx, rtx, rtx, rtx); extern rtx riscv_legitimize_call_address (rtx); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 60ebd9903e5..de30bf4e567 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1415,7 +1415,7 @@ riscv_emit_set (rtx target, rtx src) /* Emit an instruction of the form (set DEST (CODE X Y)). */ -static rtx +rtx riscv_emit_binary (enum rtx_code code, rtx dest, rtx x, rtx y) { return riscv_emit_set (dest, gen_rtx_fmt_ee (code, GET_MODE (dest), x, y)); diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index be960583101..38b8fba2a53 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -513,7 +513,7 @@ (define_insn "add<mode>3" [(set_attr "type" "fadd") (set_attr "mode" "<UNITMODE>")]) -(define_insn "addsi3" +(define_insn "*addsi3" [(set (match_operand:SI 0 "register_operand" "=r,r") (plus:SI (match_operand:SI 1 "register_operand" " r,r") (match_operand:SI 2 "arith_operand" " r,I")))] @@ -522,6 +522,24 @@ (define_insn "addsi3" [(set_attr "type" "arith") (set_attr "mode" "SI")]) +(define_expand "addsi3" + [(set (match_operand:SI 0 "register_operand" "=r,r") + (plus:SI (match_operand:SI 1 "register_operand" " r,r") + (match_operand:SI 2 "arith_operand" " r,I")))] + "" +{ + if (TARGET_64BIT) + { + rtx t = gen_reg_rtx (DImode); + emit_insn (gen_addsi3_extended (t, operands[1], operands[2])); + t = gen_lowpart (SImode, t); + SUBREG_PROMOTED_VAR_P (t) = 1; + SUBREG_PROMOTED_SET (t, SRP_SIGNED); + emit_move_insn (operands[0], t); + DONE; + } +}) + (define_insn "adddi3" [(set (match_operand:DI 0 "register_operand" "=r,r") (plus:DI (match_operand:DI 1 "register_operand" " r,r") @@ -545,7 +563,7 @@ (define_expand "addv<mode>4" rtx t5 = gen_reg_rtx (DImode); rtx t6 = gen_reg_rtx (DImode); - emit_insn (gen_addsi3 (operands[0], operands[1], operands[2])); + riscv_emit_binary (PLUS, operands[0], operands[1], operands[2]); if (GET_CODE (operands[1]) != CONST_INT) emit_insn (gen_extend_insn (t4, operands[1], DImode, SImode, 0)); else @@ -591,7 +609,7 @@ (define_expand "uaddv<mode>4" emit_insn (gen_extend_insn (t3, operands[1], DImode, SImode, 0)); else t3 = operands[1]; - emit_insn (gen_addsi3 (operands[0], operands[1], operands[2])); + riscv_emit_binary (PLUS, operands[0], operands[1], operands[2]); emit_insn (gen_extend_insn (t4, operands[0], DImode, SImode, 0)); riscv_expand_conditional_branch (operands[3], LTU, t4, t3); @@ -606,7 +624,7 @@ (define_expand "uaddv<mode>4" DONE; }) -(define_insn "*addsi3_extended" +(define_insn "addsi3_extended" [(set (match_operand:DI 0 "register_operand" "=r,r") (sign_extend:DI (plus:SI (match_operand:SI 1 "register_operand" " r,r") @@ -653,7 +671,7 @@ (define_insn "subdi3" [(set_attr "type" "arith") (set_attr "mode" "DI")]) -(define_insn "subsi3" +(define_insn "*subsi3" [(set (match_operand:SI 0 "register_operand" "= r") (minus:SI (match_operand:SI 1 "reg_or_0_operand" " rJ") (match_operand:SI 2 "register_operand" " r")))] @@ -662,6 +680,24 @@ (define_insn "subsi3" [(set_attr "type" "arith") (set_attr "mode" "SI")]) +(define_expand "subsi3" + [(set (match_operand:SI 0 "register_operand" "= r") + (minus:SI (match_operand:SI 1 "reg_or_0_operand" " rJ") + (match_operand:SI 2 "register_operand" " r")))] + "" +{ + if (TARGET_64BIT) + { + rtx t = gen_reg_rtx (DImode); + emit_insn (gen_subsi3_extended (t, operands[1], operands[2])); + t = gen_lowpart (SImode, t); + SUBREG_PROMOTED_VAR_P (t) = 1; + SUBREG_PROMOTED_SET (t, SRP_SIGNED); + emit_move_insn (operands[0], t); + DONE; + } +}) + (define_expand "subv<mode>4" [(set (match_operand:GPR 0 "register_operand" "= r") (minus:GPR (match_operand:GPR 1 "reg_or_0_operand" " rJ") @@ -676,7 +712,7 @@ (define_expand "subv<mode>4" rtx t5 = gen_reg_rtx (DImode); rtx t6 = gen_reg_rtx (DImode); - emit_insn (gen_subsi3 (operands[0], operands[1], operands[2])); + riscv_emit_binary (MINUS, operands[0], operands[1], operands[2]); if (GET_CODE (operands[1]) != CONST_INT) emit_insn (gen_extend_insn (t4, operands[1], DImode, SImode, 0)); else @@ -725,7 +761,7 @@ (define_expand "usubv<mode>4" emit_insn (gen_extend_insn (t3, operands[1], DImode, SImode, 0)); else t3 = operands[1]; - emit_insn (gen_subsi3 (operands[0], operands[1], operands[2])); + riscv_emit_binary (MINUS, operands[0], operands[1], operands[2]); emit_insn (gen_extend_insn (t4, operands[0], DImode, SImode, 0)); riscv_expand_conditional_branch (operands[3], LTU, t3, t4); @@ -741,7 +777,7 @@ (define_expand "usubv<mode>4" }) -(define_insn "*subsi3_extended" +(define_insn "subsi3_extended" [(set (match_operand:DI 0 "register_operand" "= r") (sign_extend:DI (minus:SI (match_operand:SI 1 "reg_or_0_operand" " rJ") @@ -770,7 +806,7 @@ (define_insn "negdi2" [(set_attr "type" "arith") (set_attr "mode" "DI")]) -(define_insn "negsi2" +(define_insn "*negsi2" [(set (match_operand:SI 0 "register_operand" "=r") (neg:SI (match_operand:SI 1 "register_operand" " r")))] "" @@ -778,7 +814,24 @@ (define_insn "negsi2" [(set_attr "type" "arith") (set_attr "mode" "SI")]) -(define_insn "*negsi2_extended" +(define_expand "negsi2" + [(set (match_operand:SI 0 "register_operand" "=r") + (neg:SI (match_operand:SI 1 "register_operand" " r")))] + "" +{ + if (TARGET_64BIT) + { + rtx t = gen_reg_rtx (DImode); + emit_insn (gen_negsi2_extended (t, operands[1])); + t = gen_lowpart (SImode, t); + SUBREG_PROMOTED_VAR_P (t) = 1; + SUBREG_PROMOTED_SET (t, SRP_SIGNED); + emit_move_insn (operands[0], t); + DONE; + } +}) + +(define_insn "negsi2_extended" [(set (match_operand:DI 0 "register_operand" "=r") (sign_extend:DI (neg:SI (match_operand:SI 1 "register_operand" " r"))))] @@ -814,7 +867,7 @@ (define_insn "mul<mode>3" [(set_attr "type" "fmul") (set_attr "mode" "<UNITMODE>")]) -(define_insn "mulsi3" +(define_insn "*mulsi3" [(set (match_operand:SI 0 "register_operand" "=r") (mult:SI (match_operand:SI 1 "register_operand" " r") (match_operand:SI 2 "register_operand" " r")))] @@ -823,6 +876,24 @@ (define_insn "mulsi3" [(set_attr "type" "imul") (set_attr "mode" "SI")]) +(define_expand "mulsi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (mult:SI (match_operand:SI 1 "register_operand" " r") + (match_operand:SI 2 "register_operand" " r")))] + "TARGET_ZMMUL || TARGET_MUL" +{ + if (TARGET_64BIT) + { + rtx t = gen_reg_rtx (DImode); + emit_insn (gen_mulsi3_extended (t, operands[1], operands[2])); + t = gen_lowpart (SImode, t); + SUBREG_PROMOTED_VAR_P (t) = 1; + SUBREG_PROMOTED_SET (t, SRP_SIGNED); + emit_move_insn (operands[0], t); + DONE; + } +}) + (define_insn "muldi3" [(set (match_operand:DI 0 "register_operand" "=r") (mult:DI (match_operand:DI 1 "register_operand" " r") @@ -868,8 +939,8 @@ (define_expand "mulv<mode>4" emit_insn (gen_smul<mode>3_highpart (hp, operands[1], operands[2])); emit_insn (gen_mul<mode>3 (operands[0], operands[1], operands[2])); - emit_insn (gen_ashr<mode>3 (lp, operands[0], - GEN_INT (BITS_PER_WORD - 1))); + riscv_emit_binary (ASHIFTRT, lp, operands[0], + GEN_INT (BITS_PER_WORD - 1)); riscv_expand_conditional_branch (operands[3], NE, hp, lp); } @@ -923,7 +994,7 @@ (define_expand "umulv<mode>4" DONE; }) -(define_insn "*mulsi3_extended" +(define_insn "mulsi3_extended" [(set (match_operand:DI 0 "register_operand" "=r") (sign_extend:DI (mult:SI (match_operand:SI 1 "register_operand" " r") @@ -1024,7 +1095,7 @@ (define_expand "<u>mulsidi3" "(TARGET_ZMMUL || TARGET_MUL) && !TARGET_64BIT" { rtx temp = gen_reg_rtx (SImode); - emit_insn (gen_mulsi3 (temp, operands[1], operands[2])); + riscv_emit_binary (MULT, temp, operands[1], operands[2]); emit_insn (gen_<su>mulsi3_highpart (riscv_subword (operands[0], true), operands[1], operands[2])); emit_insn (gen_movsi (riscv_subword (operands[0], false), temp)); @@ -1055,7 +1126,7 @@ (define_expand "usmulsidi3" "(TARGET_ZMMUL || TARGET_MUL) && !TARGET_64BIT" { rtx temp = gen_reg_rtx (SImode); - emit_insn (gen_mulsi3 (temp, operands[1], operands[2])); + riscv_emit_binary (MULT, temp, operands[1], operands[2]); emit_insn (gen_usmulsi3_highpart (riscv_subword (operands[0], true), operands[1], operands[2])); emit_insn (gen_movsi (riscv_subword (operands[0], false), temp)); @@ -1084,7 +1155,7 @@ (define_insn "usmulsi3_highpart" ;; .................... ;; -(define_insn "<optab>si3" +(define_insn "*<optab>si3" [(set (match_operand:SI 0 "register_operand" "=r") (any_div:SI (match_operand:SI 1 "register_operand" " r") (match_operand:SI 2 "register_operand" " r")))] @@ -1093,6 +1164,24 @@ (define_insn "<optab>si3" [(set_attr "type" "idiv") (set_attr "mode" "SI")]) +(define_expand "<optab>si3" + [(set (match_operand:SI 0 "register_operand" "=r") + (any_div:SI (match_operand:SI 1 "register_operand" " r") + (match_operand:SI 2 "register_operand" " r")))] + "TARGET_DIV" +{ + if (TARGET_64BIT) + { + rtx t = gen_reg_rtx (DImode); + emit_insn (gen_<optab>si3_extended (t, operands[1], operands[2])); + t = gen_lowpart (SImode, t); + SUBREG_PROMOTED_VAR_P (t) = 1; + SUBREG_PROMOTED_SET (t, SRP_SIGNED); + emit_move_insn (operands[0], t); + DONE; + } +}) + (define_insn "<optab>di3" [(set (match_operand:DI 0 "register_operand" "=r") (any_div:DI (match_operand:DI 1 "register_operand" " r") @@ -1118,7 +1207,7 @@ (define_expand "<u>divmod<mode>4" DONE; }) -(define_insn "*<optab>si3_extended" +(define_insn "<optab>si3_extended" [(set (match_operand:DI 0 "register_operand" "=r") (sign_extend:DI (any_div:SI (match_operand:SI 1 "register_operand" " r") @@ -2072,7 +2161,7 @@ (define_insn "riscv_pause" ;; expand_shift_1 can do this automatically when SHIFT_COUNT_TRUNCATED is ;; defined, but use of that is discouraged. -(define_insn "<optab>si3" +(define_insn "*<optab>si3" [(set (match_operand:SI 0 "register_operand" "= r") (any_shift:SI (match_operand:SI 1 "register_operand" " r") @@ -2088,6 +2177,24 @@ (define_insn "<optab>si3" [(set_attr "type" "shift") (set_attr "mode" "SI")]) +(define_expand "<optab>si3" + [(set (match_operand:SI 0 "register_operand" "= r") + (any_shift:SI (match_operand:SI 1 "register_operand" " r") + (match_operand:QI 2 "arith_operand" " rI")))] + "" +{ + if (TARGET_64BIT) + { + rtx t = gen_reg_rtx (DImode); + emit_insn (gen_<optab>si3_extend (t, operands[1], operands[2])); + t = gen_lowpart (SImode, t); + SUBREG_PROMOTED_VAR_P (t) = 1; + SUBREG_PROMOTED_SET (t, SRP_SIGNED); + emit_move_insn (operands[0], t); + DONE; + } +}) + (define_insn "<optab>di3" [(set (match_operand:DI 0 "register_operand" "= r") (any_shift:DI @@ -2122,7 +2229,7 @@ (define_insn_and_split "*<optab><GPR:mode>3_mask_1" [(set_attr "type" "shift") (set_attr "mode" "<GPR:MODE>")]) -(define_insn "*<optab>si3_extend" +(define_insn "<optab>si3_extend" [(set (match_operand:DI 0 "register_operand" "= r") (sign_extend:DI (any_shift:SI (match_operand:SI 1 "register_operand" " r") diff --git a/gcc/loop-iv.cc b/gcc/loop-iv.cc index 6c40db947f7..858c5ee84f2 100644 --- a/gcc/loop-iv.cc +++ b/gcc/loop-iv.cc @@ -637,7 +637,7 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg, { rtx set, rhs, op0 = NULL_RTX, op1 = NULL_RTX; rtx next, nextr; - enum rtx_code code; + enum rtx_code code, prev_code = UNKNOWN; rtx_insn *insn = DF_REF_INSN (def); df_ref next_def; enum iv_grd_result res; @@ -697,6 +697,27 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg, return false; op0 = XEXP (rhs, 0); + + /* rv64 wraps SImode arithmetic inside an extension to DImode. + This matches the actual hardware semantics. So peek inside + the extension and see if we have simple arithmetic that we + can analyze. */ + if (GET_CODE (op0) == PLUS) + { + rhs = op0; + op0 = XEXP (rhs, 0); + op1 = XEXP (rhs, 1); + + if (CONSTANT_P (op0)) + std::swap (op0, op1); + + if (!simple_reg_p (op0) || !CONSTANT_P (op1)) + return false; + + prev_code = code; + code = PLUS; + } + if (!simple_reg_p (op0)) return false; @@ -769,6 +790,11 @@ get_biv_step_1 (df_ref def, scalar_int_mode outer_mode, rtx reg, else *outer_step = simplify_gen_binary (code, outer_mode, *outer_step, op1); + + if (prev_code == SIGN_EXTEND) + *extend = IV_SIGN_EXTEND; + else if (prev_code == ZERO_EXTEND) + *extend = IV_ZERO_EXTEND; break; case SIGN_EXTEND: diff --git a/gcc/testsuite/gcc.target/riscv/shift-and-2.c b/gcc/testsuite/gcc.target/riscv/shift-and-2.c index bc01e8ef992..ee9925b7498 100644 --- a/gcc/testsuite/gcc.target/riscv/shift-and-2.c +++ b/gcc/testsuite/gcc.target/riscv/shift-and-2.c @@ -38,5 +38,24 @@ sub6 (long i, long j) { return i << (j & 0x3f); } + +/* Test for <optab>si3_extend. */ +int +sub7 (int i, int j) { + return (i << 10) & j; +} + +/* Test for <optab>si3_extend. */ +unsigned +sub8 (unsigned i, unsigned j) { + return (i << 10) & j; +} + +/* Test for <optab>si3_extend. */ +unsigned +sub9 (unsigned i, unsigned j) { + return (i >> 10) & j; +} + /* { dg-final { scan-assembler-not "andi" } } */ /* { dg-final { scan-assembler-not "sext.w" } } */ diff --git a/gcc/testsuite/gcc.target/riscv/shift-shift-2.c b/gcc/testsuite/gcc.target/riscv/shift-shift-2.c index 5f93be15ac5..bc8c4ef3828 100644 --- a/gcc/testsuite/gcc.target/riscv/shift-shift-2.c +++ b/gcc/testsuite/gcc.target/riscv/shift-shift-2.c @@ -38,5 +38,6 @@ sub5 (unsigned int i) } /* { dg-final { scan-assembler-times "slli" 5 } } */ /* { dg-final { scan-assembler-times "srli" 5 } } */ -/* { dg-final { scan-assembler-times "slliw" 1 } } */ -/* { dg-final { scan-assembler-times "srliw" 1 } } */ +/* { dg-final { scan-assembler-times ",40" 2 } } */ /* For sub5 test */ +/* { dg-final { scan-assembler-not "slliw" } } */ +/* { dg-final { scan-assembler-not "srliw" } } */ diff --git a/gcc/testsuite/gcc.target/riscv/sign-extend.c b/gcc/testsuite/gcc.target/riscv/sign-extend.c new file mode 100644 index 00000000000..6f840194833 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/sign-extend.c @@ -0,0 +1,81 @@ +/* { dg-do compile { target { riscv64*-*-* } } } */ +/* { dg-options "-march=rv64gc -mabi=lp64" } */ +/* { dg-skip-if "" { *-*-* } { "-O0" } } */ + +unsigned +foo1 (unsigned x, unsigned y, unsigned z) +{ + return x & (y - z); +} + +int +foo2 (int x, int y, int z) +{ + return x & (y - z); +} + +unsigned +foo3 (unsigned x, unsigned y, unsigned z) +{ + return x & (y * z); +} + +int +foo4 (int x, int y, int z) +{ + return x & (y * z); +} + +unsigned +foo5 (unsigned x, unsigned y) +{ + return x & (y / x); +} + +int +foo6 (int x, int y) +{ + return x & (y / x); +} + +unsigned +foo7 (unsigned x, unsigned y) +{ + return x & (y % x); +} + +int +foo8 (int x, int y) +{ + return x & (y % x); +} + +int +foo9 (int x) +{ + return x & (-x); +} + +unsigned +foo10 (unsigned x, unsigned y) +{ + return x & (y + x); +} + + +unsigned +foo11 (unsigned x) +{ + return x & (15 + x); +} + +/* { dg-final { scan-assembler-times "subw" 2 } } */ +/* { dg-final { scan-assembler-times "addw" 1 } } */ +/* { dg-final { scan-assembler-times "addiw" 1 } } */ +/* { dg-final { scan-assembler-times "mulw" 2 } } */ +/* { dg-final { scan-assembler-times "divw" 1 } } */ +/* { dg-final { scan-assembler-times "divuw" 1 } } */ +/* { dg-final { scan-assembler-times "remw" 1 } } */ +/* { dg-final { scan-assembler-times "remuw" 1 } } */ +/* { dg-final { scan-assembler-times "negw" 1 } } */ +/* { dg-final { scan-assembler-not "sext.w" } } */ diff --git a/gcc/testsuite/gcc.target/riscv/zbb-rol-ror-03.c b/gcc/testsuite/gcc.target/riscv/zbb-rol-ror-03.c index b44d7fe8920..e7e5cbb9a1a 100644 --- a/gcc/testsuite/gcc.target/riscv/zbb-rol-ror-03.c +++ b/gcc/testsuite/gcc.target/riscv/zbb-rol-ror-03.c @@ -16,4 +16,5 @@ unsigned int ror(unsigned int rs1, unsigned int rs2) /* { dg-final { scan-assembler-times "rolw" 1 } } */ /* { dg-final { scan-assembler-times "rorw" 1 } } */ -/* { dg-final { scan-assembler-not "and" } } */ \ No newline at end of file +/* { dg-final { scan-assembler-not "and" } } */ +/* { dg-final { scan-assembler-not "sext.w" } } */ ^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2023-06-07 20:00 UTC | newest] Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2023-05-24 23:14 [RFC] RISC-V: Eliminate extension after for *w instructions Jivan Hakobyan 2023-05-31 0:49 ` Jeff Law 2023-06-04 3:59 ` Jeff Law 2023-06-07 20:00 ` Jeff Law
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).