From: "Roger Sayle" <roger@nextmovesoftware.com>
To: <gcc-patches@gcc.gnu.org>
Cc: "'Uros Bizjak'" <ubizjak@gmail.com>
Subject: [x86 PATCH] Use QImode for offsets in zero_extract/sign_extract in i386.md
Date: Sat, 22 Jul 2023 16:37:39 +0100 [thread overview]
Message-ID: <00c701d9bcb2$733bcdc0$59b36940$@nextmovesoftware.com> (raw)
[-- Attachment #1: Type: text/plain, Size: 1773 bytes --]
As suggested by Uros, this patch changes the ZERO_EXTRACTs and SIGN_EXTRACTs
in i386.md to consistently use QImode for bit offsets (i.e. third and fourth
operands), matching the use of QImode for bit counts in shifts and rotates.
There's no change in functionality, and the new patterns simply ensure that
we continue to generate the same code (match revised patterns) as before.
This patch has been tested on x86_64-pc-linux-gnu with make bootstrap
and make -k check, both with and without --target_board=unix{-m32}
with no new failures. Ok for mainline?
2023-07-22 Roger Sayle <roger@nextmovesoftware.com>
gcc/ChangeLog
* config/i386/i386.md (extv<mode>): Use QImode for offsets.
(extzv<mode>): Likewise.
(insv<mode>): Likewise.
(*testqi_ext_3): Likewise.
(*btr<mode>_2): Likewise.
(define_split): Likewise.
(*btsq_imm): Likewise.
(*btrq_imm): Likewise.
(*btcq_imm): Likewise.
(define_peephole2 x3): Likewise.
(*bt<mode>): Likewise
(*bt<mode>_mask): New define_insn_and_split.
(*jcc_bt<mode>): Use QImode for offsets.
(*jcc_bt<mode>_1): Delete obsolete pattern.
(*jcc_bt<mode>_mask): Use QImode offsets.
(*jcc_bt<mode>_mask_1): Likewise.
(define_split): Likewise.
(*bt<mode>_setcqi): Likewise.
(*bt<mode>_setncqi): Likewise.
(*bt<mode>_setnc<mode>): Likewise.
(*bt<mode>_setncqi_2): Likewise.
(*bt<mode>_setc<mode>_mask): New define_insn_and_split.
(bmi2_bzhi_<mode>3): Use QImode offsets.
(*bmi2_bzhi_<mode>3): Likewise.
(*bmi2_bzhi_<mode>3_1): Likewise.
(*bmi2_bzhi_<mode>3_1_ccz): Likewise.
(@tbm_bextri_<mode>): Likewise.
Thanks,
Roger
--
[-- Attachment #2: patchqi.txt --]
[-- Type: text/plain, Size: 16789 bytes --]
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index 47ea050..de8c3a5 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -3312,8 +3312,8 @@
(define_expand "extv<mode>"
[(set (match_operand:SWI24 0 "register_operand")
(sign_extract:SWI24 (match_operand:SWI24 1 "register_operand")
- (match_operand:SI 2 "const_int_operand")
- (match_operand:SI 3 "const_int_operand")))]
+ (match_operand:QI 2 "const_int_operand")
+ (match_operand:QI 3 "const_int_operand")))]
""
{
/* Handle extractions from %ah et al. */
@@ -3340,8 +3340,8 @@
(define_expand "extzv<mode>"
[(set (match_operand:SWI248 0 "register_operand")
(zero_extract:SWI248 (match_operand:SWI248 1 "register_operand")
- (match_operand:SI 2 "const_int_operand")
- (match_operand:SI 3 "const_int_operand")))]
+ (match_operand:QI 2 "const_int_operand")
+ (match_operand:QI 3 "const_int_operand")))]
""
{
if (ix86_expand_pextr (operands))
@@ -3428,8 +3428,8 @@
(define_expand "insv<mode>"
[(set (zero_extract:SWI248 (match_operand:SWI248 0 "register_operand")
- (match_operand:SI 1 "const_int_operand")
- (match_operand:SI 2 "const_int_operand"))
+ (match_operand:QI 1 "const_int_operand")
+ (match_operand:QI 2 "const_int_operand"))
(match_operand:SWI248 3 "register_operand"))]
""
{
@@ -10788,8 +10788,8 @@
(match_operator 1 "compare_operator"
[(zero_extract:SWI248
(match_operand 2 "int_nonimmediate_operand" "rm")
- (match_operand 3 "const_int_operand")
- (match_operand 4 "const_int_operand"))
+ (match_operand:QI 3 "const_int_operand")
+ (match_operand:QI 4 "const_int_operand"))
(const_int 0)]))]
"/* Ensure that resulting mask is zero or sign extended operand. */
INTVAL (operands[4]) >= 0
@@ -15904,7 +15904,7 @@
[(set (zero_extract:HI
(match_operand:SWI12 0 "nonimmediate_operand")
(const_int 1)
- (zero_extend:SI (match_operand:QI 1 "register_operand")))
+ (match_operand:QI 1 "register_operand"))
(const_int 0))
(clobber (reg:CC FLAGS_REG))]
"TARGET_USE_BT && ix86_pre_reload_split ()"
@@ -15928,7 +15928,7 @@
[(set (zero_extract:HI
(match_operand:SWI12 0 "register_operand")
(const_int 1)
- (zero_extend:SI (match_operand:QI 1 "register_operand")))
+ (match_operand:QI 1 "register_operand"))
(const_int 0))
(clobber (reg:CC FLAGS_REG))]
"TARGET_USE_BT && ix86_pre_reload_split ()"
@@ -15955,7 +15955,7 @@
(define_insn "*btsq_imm"
[(set (zero_extract:DI (match_operand:DI 0 "nonimmediate_operand" "+rm")
(const_int 1)
- (match_operand 1 "const_0_to_63_operand"))
+ (match_operand:QI 1 "const_0_to_63_operand"))
(const_int 1))
(clobber (reg:CC FLAGS_REG))]
"TARGET_64BIT && (TARGET_USE_BT || reload_completed)"
@@ -15968,7 +15968,7 @@
(define_insn "*btrq_imm"
[(set (zero_extract:DI (match_operand:DI 0 "nonimmediate_operand" "+rm")
(const_int 1)
- (match_operand 1 "const_0_to_63_operand"))
+ (match_operand:QI 1 "const_0_to_63_operand"))
(const_int 0))
(clobber (reg:CC FLAGS_REG))]
"TARGET_64BIT && (TARGET_USE_BT || reload_completed)"
@@ -15981,7 +15981,7 @@
(define_insn "*btcq_imm"
[(set (zero_extract:DI (match_operand:DI 0 "nonimmediate_operand" "+rm")
(const_int 1)
- (match_operand 1 "const_0_to_63_operand"))
+ (match_operand:QI 1 "const_0_to_63_operand"))
(not:DI (zero_extract:DI (match_dup 0) (const_int 1) (match_dup 1))))
(clobber (reg:CC FLAGS_REG))]
"TARGET_64BIT && (TARGET_USE_BT || reload_completed)"
@@ -15998,7 +15998,7 @@
(parallel [(set (zero_extract:DI
(match_operand:DI 0 "nonimmediate_operand")
(const_int 1)
- (match_operand 1 "const_0_to_63_operand"))
+ (match_operand:QI 1 "const_0_to_63_operand"))
(const_int 1))
(clobber (reg:CC FLAGS_REG))])]
"TARGET_64BIT && !TARGET_USE_BT"
@@ -16022,7 +16022,7 @@
(parallel [(set (zero_extract:DI
(match_operand:DI 0 "nonimmediate_operand")
(const_int 1)
- (match_operand 1 "const_0_to_63_operand"))
+ (match_operand:QI 1 "const_0_to_63_operand"))
(const_int 0))
(clobber (reg:CC FLAGS_REG))])]
"TARGET_64BIT && !TARGET_USE_BT"
@@ -16046,7 +16046,7 @@
(parallel [(set (zero_extract:DI
(match_operand:DI 0 "nonimmediate_operand")
(const_int 1)
- (match_operand 1 "const_0_to_63_operand"))
+ (match_operand:QI 1 "const_0_to_63_operand"))
(not:DI (zero_extract:DI
(match_dup 0) (const_int 1) (match_dup 1))))
(clobber (reg:CC FLAGS_REG))])]
@@ -16074,14 +16074,14 @@
(zero_extract:SWI48
(match_operand:SWI48 0 "nonimmediate_operand" "r,m")
(const_int 1)
- (match_operand:SI 1 "nonmemory_operand" "r<S>,<S>"))
+ (match_operand:QI 1 "nonmemory_operand" "q<S>,<S>"))
(const_int 0)))]
""
{
switch (get_attr_mode (insn))
{
case MODE_SI:
- return "bt{l}\t{%1, %k0|%k0, %1}";
+ return "bt{l}\t{%k1, %k0|%k0, %k1}";
case MODE_DI:
return "bt{q}\t{%q1, %0|%0, %q1}";
@@ -16099,13 +16099,36 @@
(const_string "SI")
(const_string "<MODE>")))])
+(define_insn_and_split "*bt<SWI48:mode>_mask"
+ [(set (reg:CCC FLAGS_REG)
+ (compare:CCC
+ (zero_extract:SWI48
+ (match_operand:SWI48 0 "nonimmediate_operand" "r,m")
+ (const_int 1)
+ (subreg:QI
+ (and:SWI248
+ (match_operand:SWI248 1 "register_operand")
+ (match_operand 2 "const_int_operand")) 0))
+ (const_int 0)))]
+ "TARGET_USE_BT
+ && (INTVAL (operands[2]) & (GET_MODE_BITSIZE (<SWI48:MODE>mode)-1))
+ == GET_MODE_BITSIZE (<SWI48:MODE>mode)-1
+ && ix86_pre_reload_split ()"
+ "#"
+ "&& 1"
+ [(set (reg:CCC FLAGS_REG)
+ (compare:CCC
+ (zero_extract:SWI48 (match_dup 0) (const_int 1) (match_dup 1))
+ (const_int 0)))]
+ "operands[1] = gen_lowpart (QImode, operands[1]);")
+
(define_insn_and_split "*jcc_bt<mode>"
[(set (pc)
(if_then_else (match_operator 0 "bt_comparison_operator"
[(zero_extract:SWI48
(match_operand:SWI48 1 "nonimmediate_operand")
(const_int 1)
- (match_operand:SI 2 "nonmemory_operand"))
+ (match_operand:QI 2 "nonmemory_operand"))
(const_int 0)])
(label_ref (match_operand 3))
(pc)))
@@ -16135,39 +16158,6 @@
PUT_CODE (operands[0], reverse_condition (GET_CODE (operands[0])));
})
-(define_insn_and_split "*jcc_bt<mode>_1"
- [(set (pc)
- (if_then_else (match_operator 0 "bt_comparison_operator"
- [(zero_extract:SWI48
- (match_operand:SWI48 1 "register_operand")
- (const_int 1)
- (zero_extend:SI
- (match_operand:QI 2 "register_operand")))
- (const_int 0)])
- (label_ref (match_operand 3))
- (pc)))
- (clobber (reg:CC FLAGS_REG))]
- "(TARGET_USE_BT || optimize_function_for_size_p (cfun))
- && ix86_pre_reload_split ()"
- "#"
- "&& 1"
- [(set (reg:CCC FLAGS_REG)
- (compare:CCC
- (zero_extract:SWI48
- (match_dup 1)
- (const_int 1)
- (match_dup 2))
- (const_int 0)))
- (set (pc)
- (if_then_else (match_op_dup 0 [(reg:CCC FLAGS_REG) (const_int 0)])
- (label_ref (match_dup 3))
- (pc)))]
-{
- operands[2] = lowpart_subreg (SImode, operands[2], QImode);
- operands[0] = shallow_copy_rtx (operands[0]);
- PUT_CODE (operands[0], reverse_condition (GET_CODE (operands[0])));
-})
-
;; Avoid useless masking of bit offset operand.
(define_insn_and_split "*jcc_bt<mode>_mask"
[(set (pc)
@@ -16175,8 +16165,8 @@
[(zero_extract:SWI48
(match_operand:SWI48 1 "register_operand")
(const_int 1)
- (and:SI
- (match_operand:SI 2 "register_operand")
+ (and:QI
+ (match_operand:QI 2 "register_operand")
(match_operand 3 "const_int_operand")))])
(label_ref (match_operand 4))
(pc)))
@@ -16203,23 +16193,23 @@
PUT_CODE (operands[0], reverse_condition (GET_CODE (operands[0])));
})
-(define_insn_and_split "*jcc_bt<mode>_mask_1"
+;; Avoid useless masking of bit offset operand.
+(define_insn_and_split "*jcc_bt<SWI48:mode>_mask_1"
[(set (pc)
- (if_then_else (match_operator 0 "bt_comparison_operator"
+ (if_then_else (match_operator 0 "bt_comparison_operator"
[(zero_extract:SWI48
(match_operand:SWI48 1 "register_operand")
(const_int 1)
- (zero_extend:SI
- (subreg:QI
- (and
- (match_operand 2 "int248_register_operand")
- (match_operand 3 "const_int_operand")) 0)))])
+ (subreg:QI
+ (and:SWI248
+ (match_operand:SWI248 2 "register_operand")
+ (match_operand 3 "const_int_operand")) 0))])
(label_ref (match_operand 4))
(pc)))
(clobber (reg:CC FLAGS_REG))]
"(TARGET_USE_BT || optimize_function_for_size_p (cfun))
- && (INTVAL (operands[3]) & (GET_MODE_BITSIZE (<MODE>mode)-1))
- == GET_MODE_BITSIZE (<MODE>mode)-1
+ && (INTVAL (operands[3]) & (GET_MODE_BITSIZE (<SWI48:MODE>mode)-1))
+ == GET_MODE_BITSIZE (<SWI48:MODE>mode)-1
&& ix86_pre_reload_split ()"
"#"
"&& 1"
@@ -16235,10 +16225,9 @@
(label_ref (match_dup 4))
(pc)))]
{
- operands[2] = force_reg (GET_MODE (operands[2]), operands[2]);
- operands[2] = gen_lowpart (SImode, operands[2]);
operands[0] = shallow_copy_rtx (operands[0]);
PUT_CODE (operands[0], reverse_condition (GET_CODE (operands[0])));
+ operands[2] = gen_lowpart (QImode, operands[2]);
})
;; Help combine recognize bt followed by cmov
@@ -16249,7 +16238,7 @@
[(zero_extract:SWI48
(match_operand:SWI48 1 "register_operand")
(const_int 1)
- (zero_extend:SI (match_operand:QI 2 "register_operand")))
+ (match_operand:QI 2 "register_operand"))
(const_int 0)])
(match_operand:SWI248 3 "nonimmediate_operand")
(match_operand:SWI248 4 "nonimmediate_operand")))]
@@ -16267,7 +16256,6 @@
{
if (GET_CODE (operands[5]) == EQ)
std::swap (operands[3], operands[4]);
- operands[2] = lowpart_subreg (SImode, operands[2], QImode);
})
;; Help combine recognize bt followed by setc
@@ -16276,7 +16264,7 @@
(zero_extract:SWI48
(match_operand:SWI48 1 "register_operand")
(const_int 1)
- (zero_extend:SI (match_operand:QI 2 "register_operand"))))
+ (match_operand:QI 2 "register_operand")))
(clobber (reg:CC FLAGS_REG))]
"TARGET_USE_BT && ix86_pre_reload_split ()"
"#"
@@ -16286,8 +16274,7 @@
(zero_extract:SWI48 (match_dup 1) (const_int 1) (match_dup 2))
(const_int 0)))
(set (match_dup 0)
- (eq:QI (reg:CCC FLAGS_REG) (const_int 0)))]
- "operands[2] = lowpart_subreg (SImode, operands[2], QImode);")
+ (eq:QI (reg:CCC FLAGS_REG) (const_int 0)))])
;; Help combine recognize bt followed by setnc
(define_insn_and_split "*bt<mode>_setncqi"
@@ -16307,8 +16294,7 @@
(zero_extract:SWI48 (match_dup 1) (const_int 1) (match_dup 2))
(const_int 0)))
(set (match_dup 0)
- (ne:QI (reg:CCC FLAGS_REG) (const_int 0)))]
- "operands[2] = lowpart_subreg (SImode, operands[2], QImode);")
+ (ne:QI (reg:CCC FLAGS_REG) (const_int 0)))])
(define_insn_and_split "*bt<mode>_setnc<mode>"
[(set (match_operand:SWI48 0 "register_operand")
@@ -16328,10 +16314,7 @@
(set (match_dup 3)
(ne:QI (reg:CCC FLAGS_REG) (const_int 0)))
(set (match_dup 0) (zero_extend:SWI48 (match_dup 3)))]
-{
- operands[2] = lowpart_subreg (SImode, operands[2], QImode);
- operands[3] = gen_reg_rtx (QImode);
-})
+ "operands[3] = gen_reg_rtx (QImode);")
;; Help combine recognize bt followed by setnc (PR target/110588)
(define_insn_and_split "*bt<mode>_setncqi_2"
@@ -16340,7 +16323,7 @@
(zero_extract:SWI48
(match_operand:SWI48 1 "register_operand")
(const_int 1)
- (zero_extend:SI (match_operand:QI 2 "register_operand")))
+ (match_operand:QI 2 "register_operand"))
(const_int 0)))
(clobber (reg:CC FLAGS_REG))]
"TARGET_USE_BT && ix86_pre_reload_split ()"
@@ -16351,8 +16334,36 @@
(zero_extract:SWI48 (match_dup 1) (const_int 1) (match_dup 2))
(const_int 0)))
(set (match_dup 0)
- (ne:QI (reg:CCC FLAGS_REG) (const_int 0)))]
- "operands[2] = lowpart_subreg (SImode, operands[2], QImode);")
+ (ne:QI (reg:CCC FLAGS_REG) (const_int 0)))])
+
+;; Help combine recognize bt followed by setc
+(define_insn_and_split "*bt<mode>_setc<mode>_mask"
+ [(set (match_operand:SWI48 0 "register_operand")
+ (zero_extract:SWI48
+ (match_operand:SWI48 1 "register_operand")
+ (const_int 1)
+ (subreg:QI
+ (and:SWI48
+ (match_operand:SWI48 2 "register_operand")
+ (match_operand 3 "const_int_operand")) 0)))
+ (clobber (reg:CC FLAGS_REG))]
+ "TARGET_USE_BT
+ && (INTVAL (operands[3]) & (GET_MODE_BITSIZE (<MODE>mode)-1))
+ == GET_MODE_BITSIZE (<MODE>mode)-1
+ && ix86_pre_reload_split ()"
+ "#"
+ "&& 1"
+ [(set (reg:CCC FLAGS_REG)
+ (compare:CCC
+ (zero_extract:SWI48 (match_dup 1) (const_int 1) (match_dup 2))
+ (const_int 0)))
+ (set (match_dup 3)
+ (ne:QI (reg:CCC FLAGS_REG) (const_int 0)))
+ (set (match_dup 0) (zero_extend:SWI48 (match_dup 3)))]
+{
+ operands[2] = gen_lowpart (QImode, operands[2]);
+ operands[3] = gen_reg_rtx (QImode);
+})
\f
;; Store-flag instructions.
@@ -18647,46 +18658,29 @@
[(parallel
[(set (match_operand:SWI48 0 "register_operand")
(if_then_else:SWI48
- (ne:QI (and:SWI48 (match_operand:SWI48 2 "register_operand")
- (const_int 255))
+ (ne:QI (match_operand:QI 2 "register_operand")
(const_int 0))
(zero_extract:SWI48
(match_operand:SWI48 1 "nonimmediate_operand")
- (umin:SWI48 (and:SWI48 (match_dup 2) (const_int 255))
- (match_dup 3))
+ (umin:QI (match_dup 2) (match_dup 3))
(const_int 0))
(const_int 0)))
(clobber (reg:CC FLAGS_REG))])]
"TARGET_BMI2"
- "operands[3] = GEN_INT (<MODE_SIZE> * BITS_PER_UNIT);")
+{
+ operands[2] = gen_lowpart (QImode, operands[2]);
+ operands[3] = GEN_INT (<MODE_SIZE> * BITS_PER_UNIT);
+})
(define_insn "*bmi2_bzhi_<mode>3"
[(set (match_operand:SWI48 0 "register_operand" "=r")
(if_then_else:SWI48
- (ne:QI (and:SWI48 (match_operand:SWI48 2 "register_operand" "r")
- (const_int 255))
+ (ne:QI (match_operand:QI 2 "register_operand" "q")
(const_int 0))
(zero_extract:SWI48
(match_operand:SWI48 1 "nonimmediate_operand" "rm")
- (umin:SWI48 (and:SWI48 (match_dup 2) (const_int 255))
- (match_operand:SWI48 3 "const_int_operand"))
- (const_int 0))
- (const_int 0)))
- (clobber (reg:CC FLAGS_REG))]
- "TARGET_BMI2 && INTVAL (operands[3]) == <MODE_SIZE> * BITS_PER_UNIT"
- "bzhi\t{%2, %1, %0|%0, %1, %2}"
- [(set_attr "type" "bitmanip")
- (set_attr "prefix" "vex")
- (set_attr "mode" "<MODE>")])
-
-(define_insn "*bmi2_bzhi_<mode>3_1"
- [(set (match_operand:SWI48 0 "register_operand" "=r")
- (if_then_else:SWI48
- (ne:QI (match_operand:QI 2 "register_operand" "r") (const_int 0))
- (zero_extract:SWI48
- (match_operand:SWI48 1 "nonimmediate_operand" "rm")
- (umin:SWI48 (zero_extend:SWI48 (match_dup 2))
- (match_operand:SWI48 3 "const_int_operand"))
+ (umin:QI (match_dup 2)
+ (match_operand:QI 3 "const_int_operand"))
(const_int 0))
(const_int 0)))
(clobber (reg:CC FLAGS_REG))]
@@ -18703,8 +18697,8 @@
(ne:QI (match_operand:QI 2 "register_operand" "r") (const_int 0))
(zero_extract:SWI48
(match_operand:SWI48 1 "nonimmediate_operand" "rm")
- (umin:SWI48 (zero_extend:SWI48 (match_dup 2))
- (match_operand:SWI48 3 "const_int_operand"))
+ (umin:QI (match_dup 2)
+ (match_operand:QI 3 "const_int_operand"))
(const_int 0))
(const_int 0))
(const_int 0)))
@@ -18803,8 +18797,8 @@
[(set (match_operand:SWI48 0 "register_operand" "=r")
(zero_extract:SWI48
(match_operand:SWI48 1 "nonimmediate_operand" "rm")
- (match_operand 2 "const_0_to_255_operand")
- (match_operand 3 "const_0_to_255_operand")))
+ (match_operand:QI 2 "const_0_to_255_operand")
+ (match_operand:QI 3 "const_0_to_255_operand")))
(clobber (reg:CC FLAGS_REG))]
"TARGET_TBM"
{
next reply other threads:[~2023-07-22 15:37 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-07-22 15:37 Roger Sayle [this message]
2023-07-22 17:16 ` Uros Bizjak
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='00c701d9bcb2$733bcdc0$59b36940$@nextmovesoftware.com' \
--to=roger@nextmovesoftware.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=ubizjak@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).