On Mon, May 21, 2018 at 4:53 PM, Uros Bizjak <ubiz...@gmail.com> wrote:
> Hello!
>
> Attached patch implements scalar float->unsigned int truncations with AVX512F.
>
> 2018-05-21 Uros Bizjak <ubiz...@gmail.com>
>
> * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
> (fixuns_trunc<mode>si2_avx512f): Ditto.
> (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
> (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
> Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
>
> testsuite/ChangeLog:
>
> 2018-05-21 Uros Bizjak <ubiz...@gmail.com>
>
> * gcc.target/i386/cvt-2.c: New test.
>
> Patch was bootstrapped and regression tested on x86_64-linux-gnu {,-m32}.
>
> Unfortunately, I have to means to test the patch on AVX512 target, so
> to avoid some hidden issue, I'd like to ask someone to test it on live
> target.
Ops, ssemodesuffix handling was missing in the insn mnemonic. Fixed in
the attached v-2 patch.
Uros.
Index: config/i386/i386.md
===================================================================
--- config/i386/i386.md (revision 260441)
+++ config/i386/i386.md (working copy)
@@ -5017,6 +5017,18 @@
}
})
+;; Unsigned conversion to DImode
+
+(define_insn "fixuns_trunc<mode>di2"
+ [(set (match_operand:DI 0 "register_operand" "=r")
+ (unsigned_fix:DI
+ (match_operand:MODEF 1 "nonimmediate_operand" "vm")))]
+ "TARGET_64BIT && TARGET_AVX512F && TARGET_SSE_MATH"
+ "vcvtt<ssemodesuffix>2usi\t{%1, %0|%0, %1}"
+ [(set_attr "type" "sseicvt")
+ (set_attr "prefix" "evex")
+ (set_attr "mode" "DI")])
+
;; Unsigned conversion to SImode.
(define_expand "fixuns_trunc<mode>si2"
@@ -5027,7 +5039,7 @@
(use (match_dup 2))
(clobber (match_scratch:<ssevecmode> 3))
(clobber (match_scratch:<ssevecmode> 4))])]
- "!TARGET_64BIT && TARGET_SSE2 && TARGET_SSE_MATH"
+ "(!TARGET_64BIT || TARGET_AVX512F) && TARGET_SSE2 && TARGET_SSE_MATH"
{
machine_mode mode = <MODE>mode;
machine_mode vecmode = <ssevecmode>mode;
@@ -5034,6 +5046,12 @@
REAL_VALUE_TYPE TWO31r;
rtx two31;
+ if (TARGET_AVX512F)
+ {
+ emit_insn (gen_fixuns_trunc<mode>si2_avx512f (operands[0], operands[1]));
+ DONE;
+ }
+
if (optimize_insn_for_size_p ())
FAIL;
@@ -5043,6 +5061,27 @@
operands[2] = force_reg (vecmode, two31);
})
+(define_insn "fixuns_trunc<mode>si2_avx512f"
+ [(set (match_operand:SI 0 "register_operand" "=r")
+ (unsigned_fix:SI
+ (match_operand:MODEF 1 "nonimmediate_operand" "vm")))]
+ "TARGET_AVX512F && TARGET_SSE_MATH"
+ "vcvtt<ssemodesuffix>2usi\t{%1, %0|%0, %1}"
+ [(set_attr "type" "sseicvt")
+ (set_attr "prefix" "evex")
+ (set_attr "mode" "SI")])
+
+(define_insn "*fixuns_trunc<mode>si2_avx512f_zext"
+ [(set (match_operand:DI 0 "register_operand" "=r")
+ (zero_extend:DI
+ (unsigned_fix:SI
+ (match_operand:MODEF 1 "nonimmediate_operand" "vm"))))]
+ "TARGET_64BIT && TARGET_AVX512F"
+ "vcvtt<ssemodesuffix>2usi\t{%1, %k0|%k0, %1}"
+ [(set_attr "type" "sseicvt")
+ (set_attr "prefix" "evex")
+ (set_attr "mode" "SI")])
+
(define_insn_and_split "*fixuns_trunc<mode>_1"
[(set (match_operand:SI 0 "register_operand" "=&x,&x")
(unsigned_fix:SI
Index: testsuite/gcc.target/i386/cvt-2.c
===================================================================
--- testsuite/gcc.target/i386/cvt-2.c (nonexistent)
+++ testsuite/gcc.target/i386/cvt-2.c (working copy)
@@ -0,0 +1,15 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx512f -mfpmath=sse" } */
+
+unsigned int f2ui (float x) { return x; }
+unsigned int d2ui (double x) { return x; }
+
+#ifdef __x86_64__
+unsigned long f2ul (float x) { return x; }
+unsigned long d2ul (double x) { return x; }
+#endif
+
+/* { dg-final { scan-assembler-times "vcvttss2usi" 1 { target ia32 } } } */
+/* { dg-final { scan-assembler-times "vcvttsd2usi" 1 { target ia32 } } } */
+/* { dg-final { scan-assembler-times "vcvttss2usi" 2 { target { ! ia32 } } } }
*/
+/* { dg-final { scan-assembler-times "vcvttsd2usi" 2 { target { ! ia32 } } } }
*/
