Transition nvptx backend to STORE_FLAG_VALUE = 1
Commit Message
This patch to the nvptx backend changes the backend's STORE_FLAG_VALUE
from -1 to 1, by using BImode predicates and selp instructions, instead
of set instructions (almost always followed by integer negation).
Historically, it was reasonable (through rare) for backends to use -1
for representing true during the RTL passes. However with tree-ssa,
GCC now emits lots of code that reads and writes _Bool values, requiring
STORE_FLAG_VALUE=-1 targets to frequently convert 0/-1 pseudos to 0/1
pseudos using integer negation. Unfortunately, this process prevents
or complicates many optimizations (negate isn't associative with logical
AND, OR and XOR, and interferes with range/vrp/nonzerobits bounds etc.).
The impact of this is that for a relatively simple logical expression
like "return (x==21) && (y==69);", the nvptx backend currently generates:
mov.u32 %r26, %ar0;
mov.u32 %r27, %ar1;
set.u32.eq.u32 %r30, %r26, 21;
neg.s32 %r31, %r30;
mov.u32 %r29, %r31;
set.u32.eq.u32 %r33, %r27, 69;
neg.s32 %r34, %r33;
mov.u32 %r32, %r34;
cvt.u16.u8 %r39, %r29;
mov.u16 %r36, %r39;
cvt.u16.u8 %r39, %r32;
mov.u16 %r37, %r39;
and.b16 %r35, %r36, %r37;
cvt.u32.u16 %r38, %r35;
cvt.u32.u8 %value, %r38;
This patch tweaks nvptx to generate 0/1 values instead, requiring the
same number of instructions, using (BImode) predicate registers and selp
instructions so as to now generate the almost identical:
mov.u32 %r26, %ar0;
mov.u32 %r27, %ar1;
setp.eq.u32 %r31, %r26, 21;
selp.u32 %r30, 1, 0, %r31;
mov.u32 %r29, %r30;
setp.eq.u32 %r34, %r27, 69;
selp.u32 %r33, 1, 0, %r34;
mov.u32 %r32, %r33;
cvt.u16.u8 %r39, %r29;
mov.u16 %r36, %r39;
cvt.u16.u8 %r39, %r32;
mov.u16 %r37, %r39;
and.b16 %r35, %r36, %r37;
cvt.u32.u16 %r38, %r35;
cvt.u32.u8 %value, %r38;
The hidden benefit is that this sequence can (in theory) be optimized
by the RTL passes to eventually generate a much shorter sequence using
an and.pred instruction (just like Nvidia's nvcc compiler).
This patch has been tested nvptx-none with a "make" and "make -k check"
(including newlib) hosted on x86_64-pc-linux-gnu with no new failures.
Ok for mainline?
2021-10-05 Roger Sayle <roger@nextmovesoftware.com>
gcc/ChangeLog
* config/nvptx/nvptx.h (STORE_FLAG_VALUE): Change to 1.
* config/nvptx/nvptx.md (movbi): Use P1 constraint for true.
(setcc_from_bi): Remove SImode specific pattern.
(setcc<mode>_from_bi): Provide more general HSDIM pattern.
(extendbi<mode>2, zeroextendbi<mode>2): Provide instructions
for sign- and zero-extending BImode predicates to integers.
(setcc_int<mode>): Remove previous (-1-based) instructions.
(cstorebi4): Remove BImode to SImode specific expander.
(cstore<mode>4): Fix indentation. Expand using setccsi_from_bi.
(cstore<mode>4): For both integer and floating point modes.
Thanks in advance,
Roger
--
Comments
On 10/5/21 19:48, Roger Sayle wrote:
>
> This patch to the nvptx backend changes the backend's STORE_FLAG_VALUE
> from -1 to 1, by using BImode predicates and selp instructions, instead
> of set instructions (almost always followed by integer negation).
>
> Historically, it was reasonable (through rare) for backends to use -1
> for representing true during the RTL passes. However with tree-ssa,
> GCC now emits lots of code that reads and writes _Bool values, requiring
> STORE_FLAG_VALUE=-1 targets to frequently convert 0/-1 pseudos to 0/1
> pseudos using integer negation. Unfortunately, this process prevents
> or complicates many optimizations (negate isn't associative with logical
> AND, OR and XOR, and interferes with range/vrp/nonzerobits bounds etc.).
>
> The impact of this is that for a relatively simple logical expression
> like "return (x==21) && (y==69);", the nvptx backend currently generates:
>
> mov.u32 %r26, %ar0;
> mov.u32 %r27, %ar1;
> set.u32.eq.u32 %r30, %r26, 21;
> neg.s32 %r31, %r30;
> mov.u32 %r29, %r31;
> set.u32.eq.u32 %r33, %r27, 69;
> neg.s32 %r34, %r33;
> mov.u32 %r32, %r34;
> cvt.u16.u8 %r39, %r29;
> mov.u16 %r36, %r39;
> cvt.u16.u8 %r39, %r32;
> mov.u16 %r37, %r39;
> and.b16 %r35, %r36, %r37;
> cvt.u32.u16 %r38, %r35;
> cvt.u32.u8 %value, %r38;
>
> This patch tweaks nvptx to generate 0/1 values instead, requiring the
> same number of instructions, using (BImode) predicate registers and selp
> instructions so as to now generate the almost identical:
>
> mov.u32 %r26, %ar0;
> mov.u32 %r27, %ar1;
> setp.eq.u32 %r31, %r26, 21;
> selp.u32 %r30, 1, 0, %r31;
> mov.u32 %r29, %r30;
> setp.eq.u32 %r34, %r27, 69;
> selp.u32 %r33, 1, 0, %r34;
> mov.u32 %r32, %r33;
> cvt.u16.u8 %r39, %r29;
> mov.u16 %r36, %r39;
> cvt.u16.u8 %r39, %r32;
> mov.u16 %r37, %r39;
> and.b16 %r35, %r36, %r37;
> cvt.u32.u16 %r38, %r35;
> cvt.u32.u8 %value, %r38;
>
> The hidden benefit is that this sequence can (in theory) be optimized
> by the RTL passes to eventually generate a much shorter sequence using
> an and.pred instruction (just like Nvidia's nvcc compiler).
>
> This patch has been tested nvptx-none with a "make" and "make -k check"
> (including newlib) hosted on x86_64-pc-linux-gnu with no new failures.
> Ok for mainline?
>
>
Thanks for the patch, sounds reasonable.
Committed.
Thanks,
- Tom
> 2021-10-05 Roger Sayle <roger@nextmovesoftware.com>
>
> gcc/ChangeLog
> * config/nvptx/nvptx.h (STORE_FLAG_VALUE): Change to 1.
> * config/nvptx/nvptx.md (movbi): Use P1 constraint for true.
> (setcc_from_bi): Remove SImode specific pattern.
> (setcc<mode>_from_bi): Provide more general HSDIM pattern.
> (extendbi<mode>2, zeroextendbi<mode>2): Provide instructions
> for sign- and zero-extending BImode predicates to integers.
> (setcc_int<mode>): Remove previous (-1-based) instructions.
> (cstorebi4): Remove BImode to SImode specific expander.
> (cstore<mode>4): Fix indentation. Expand using setccsi_from_bi.
> (cstore<mode>4): For both integer and floating point modes.
>
>
> Thanks in advance,
> Roger
> --
>
@@ -315,7 +315,7 @@ struct GTY(()) machine_function
#define NO_DOT_IN_LABEL
#define ASM_COMMENT_START "//"
-#define STORE_FLAG_VALUE -1
+#define STORE_FLAG_VALUE 1
#define FLOAT_STORE_FLAG_VALUE(MODE) REAL_VALUE_ATOF("1.0", (MODE))
#define CASE_VECTOR_MODE SImode
@@ -213,7 +213,7 @@
;; get variables in this mode and pseudos are never spilled.
(define_insn "movbi"
[(set (match_operand:BI 0 "nvptx_register_operand" "=R,R,R")
- (match_operand:BI 1 "nvptx_nonmemory_operand" "R,P0,Pn"))]
+ (match_operand:BI 1 "nvptx_nonmemory_operand" "R,P0,P1"))]
""
"@
%.\\tmov%t0\\t%0, %1;
@@ -789,12 +789,26 @@
;; Conditional stores
-(define_insn "setcc_from_bi"
- [(set (match_operand:SI 0 "nvptx_register_operand" "=R")
- (ne:SI (match_operand:BI 1 "nvptx_register_operand" "R")
- (const_int 0)))]
+(define_insn "setcc<mode>_from_bi"
+ [(set (match_operand:HSDIM 0 "nvptx_register_operand" "=R")
+ (ne:HSDIM (match_operand:BI 1 "nvptx_register_operand" "R")
+ (const_int 0)))]
+ ""
+ "%.\\tselp%t0\\t%0, 1, 0, %1;")
+
+(define_insn "extendbi<mode>2"
+ [(set (match_operand:HSDIM 0 "nvptx_register_operand" "=R")
+ (sign_extend:HSDIM
+ (match_operand:BI 1 "nvptx_register_operand" "R")))]
+ ""
+ "%.\\tselp%t0\\t%0, -1, 0, %1;")
+
+(define_insn "zero_extendbi<mode>2"
+ [(set (match_operand:HSDIM 0 "nvptx_register_operand" "=R")
+ (zero_extend:HSDIM
+ (match_operand:BI 1 "nvptx_register_operand" "R")))]
""
- "%.\\tselp%t0 %0,-1,0,%1;")
+ "%.\\tselp%t0\\t%0, 1, 0, %1;")
(define_insn "sel_true<mode>"
[(set (match_operand:HSDIM 0 "nvptx_register_operand" "=R")
@@ -832,22 +846,6 @@
""
"%.\\tselp%t0\\t%0, %3, %2, %1;")
-(define_insn "setcc_int<mode>"
- [(set (match_operand:SI 0 "nvptx_register_operand" "=R")
- (match_operator:SI 1 "nvptx_comparison_operator"
- [(match_operand:HSDIM 2 "nvptx_register_operand" "R")
- (match_operand:HSDIM 3 "nvptx_nonmemory_operand" "Ri")]))]
- ""
- "%.\\tset%t0%c1\\t%0, %2, %3;")
-
-(define_insn "setcc_int<mode>"
- [(set (match_operand:SI 0 "nvptx_register_operand" "=R")
- (match_operator:SI 1 "nvptx_float_comparison_operator"
- [(match_operand:SDFM 2 "nvptx_register_operand" "R")
- (match_operand:SDFM 3 "nvptx_nonmemory_operand" "RF")]))]
- ""
- "%.\\tset%t0%c1\\t%0, %2, %3;")
-
(define_insn "setcc_float<mode>"
[(set (match_operand:SF 0 "nvptx_register_operand" "=R")
(match_operator:SF 1 "nvptx_comparison_operator"
@@ -864,29 +862,35 @@
""
"%.\\tset%t0%c1\\t%0, %2, %3;")
-(define_expand "cstorebi4"
- [(set (match_operand:SI 0 "nvptx_register_operand")
- (match_operator:SI 1 "ne_operator"
- [(match_operand:BI 2 "nvptx_register_operand")
- (match_operand:BI 3 "const0_operand")]))]
- ""
- "")
-
(define_expand "cstore<mode>4"
[(set (match_operand:SI 0 "nvptx_register_operand")
(match_operator:SI 1 "nvptx_comparison_operator"
- [(match_operand:HSDIM 2 "nvptx_register_operand")
- (match_operand:HSDIM 3 "nvptx_nonmemory_operand")]))]
+ [(match_operand:HSDIM 2 "nvptx_register_operand")
+ (match_operand:HSDIM 3 "nvptx_nonmemory_operand")]))]
""
- "")
+{
+ rtx reg = gen_reg_rtx (BImode);
+ rtx cmp = gen_rtx_fmt_ee (GET_CODE (operands[1]), BImode,
+ operands[2], operands[3]);
+ emit_move_insn (reg, cmp);
+ emit_insn (gen_setccsi_from_bi (operands[0], reg));
+ DONE;
+})
(define_expand "cstore<mode>4"
[(set (match_operand:SI 0 "nvptx_register_operand")
(match_operator:SI 1 "nvptx_float_comparison_operator"
- [(match_operand:SDFM 2 "nvptx_register_operand")
- (match_operand:SDFM 3 "nvptx_nonmemory_operand")]))]
+ [(match_operand:SDFM 2 "nvptx_register_operand")
+ (match_operand:SDFM 3 "nvptx_nonmemory_operand")]))]
""
- "")
+{
+ rtx reg = gen_reg_rtx (BImode);
+ rtx cmp = gen_rtx_fmt_ee (GET_CODE (operands[1]), BImode,
+ operands[2], operands[3]);
+ emit_move_insn (reg, cmp);
+ emit_insn (gen_setccsi_from_bi (operands[0], reg));
+ DONE;
+})
;; Calls