From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 2078) id C8ED9385771D; Mon, 19 Jun 2023 01:34:47 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org C8ED9385771D DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org; s=default; t=1687138487; bh=N13ulN4hLBFaqo9ldGhmSYyUlYp+YwqatBMsG01jmUQ=; h=From:To:Subject:Date:From; b=xf7sA5C614QHg+caqMzdxsGF4LTNhtEdF1Ik3DZbbLbLWBhsaouXQZfxD6hxzaEMs IeqO4NEA6U3C8x+7lMBZflQavm2mChVe/9s+xhKtd3wTBNPV7zZsgyxECTe80auZB4 2xL0gO5JmikIjzC7TSAbodotxP+LBoKjNmLlI+64= MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="utf-8" From: hongtao Liu To: gcc-cvs@gcc.gnu.org Subject: [gcc r14-1917] Refined 256/512-bit vpacksswb/vpackssdw patterns. X-Act-Checkin: gcc X-Git-Author: liuhongt X-Git-Refname: refs/heads/master X-Git-Oldrev: 58e61a3ab1c13b6d5b07d86a30cf48a46e0345c8 X-Git-Newrev: f8e02702726d4514b8ff9f5481c9c1f5d34e1787 Message-Id: <20230619013447.C8ED9385771D@sourceware.org> Date: Mon, 19 Jun 2023 01:34:47 +0000 (GMT) List-Id: https://gcc.gnu.org/g:f8e02702726d4514b8ff9f5481c9c1f5d34e1787 commit r14-1917-gf8e02702726d4514b8ff9f5481c9c1f5d34e1787 Author: liuhongt Date: Thu Jun 15 16:46:14 2023 +0800 Refined 256/512-bit vpacksswb/vpackssdw patterns. The packing in vpacksswb/vpackssdw is not a simple concat, it's an interweave from src1 and src2 for every 128 bit(or 64-bit for the ss_truncate result). .i.e. dst[192-255] = ss_truncate (src2[128-255]) dst[128-191] = ss_truncate (src1[128-255]) dst[64-127] = ss_truncate (src2[0-127]) dst[0-63] = ss_truncate (src1[0-127] The patch refined those patterns with an extra vec_select for the interweave. gcc/ChangeLog: PR target/110235 * config/i386/sse.md (_packsswb): Substitute with .. (sse2_packsswb): .. this, .. (avx2_packsswb): .. this and .. (avx512bw_packsswb): .. this. (_packssdw): Substitute with .. (sse2_packssdw): .. this, .. (avx2_packssdw): .. this and .. (avx512bw_packssdw): .. this. gcc/testsuite/ChangeLog: * gcc.target/i386/avx512bw-vpackssdw-3.c: New test. * gcc.target/i386/avx512bw-vpacksswb-3.c: New test. Diff: --- gcc/config/i386/sse.md | 165 ++++++++++++++++++--- .../gcc.target/i386/avx512bw-vpackssdw-3.c | 55 +++++++ .../gcc.target/i386/avx512bw-vpacksswb-3.c | 50 +++++++ 3 files changed, 252 insertions(+), 18 deletions(-) diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md index 35b46a54130..87570357db6 100644 --- a/gcc/config/i386/sse.md +++ b/gcc/config/i386/sse.md @@ -17762,14 +17762,14 @@ DONE; }) -(define_insn "_packsswb" - [(set (match_operand:VI1_AVX512 0 "register_operand" "=x,") - (vec_concat:VI1_AVX512 - (ss_truncate: - (match_operand: 1 "register_operand" "0,")) - (ss_truncate: - (match_operand: 2 "vector_operand" "xBm,m"))))] - "TARGET_SSE2 && && " +(define_insn "sse2_packsswb" + [(set (match_operand:V16QI 0 "register_operand" "=x,Yw") + (vec_concat:V16QI + (ss_truncate:V8QI + (match_operand:V8HI 1 "register_operand" "0,Yw")) + (ss_truncate:V8QI + (match_operand:V8HI 2 "vector_operand" "xBm,Ywm"))))] + "TARGET_SSE2 && && " "@ packsswb\t{%2, %0|%0, %2} vpacksswb\t{%2, %1, %0|%0, %1, %2}" @@ -17777,16 +17777,93 @@ (set_attr "type" "sselog") (set_attr "prefix_data16" "1,*") (set_attr "prefix" "orig,") - (set_attr "mode" "")]) + (set_attr "mode" "TI")]) -(define_insn "_packssdw" - [(set (match_operand:VI2_AVX2 0 "register_operand" "=x,") - (vec_concat:VI2_AVX2 - (ss_truncate: - (match_operand: 1 "register_operand" "0,")) - (ss_truncate: - (match_operand: 2 "vector_operand" "xBm,m"))))] - "TARGET_SSE2 && && " +(define_insn "avx2_packsswb" + [(set (match_operand:V32QI 0 "register_operand" "=Yw") + (vec_select:V32QI + (vec_concat:V32QI + (ss_truncate:V16QI + (match_operand:V16HI 1 "register_operand" "Yw")) + (ss_truncate:V16QI + (match_operand:V16HI 2 "vector_operand" "Ywm"))) + (parallel [(const_int 0) (const_int 1) + (const_int 2) (const_int 3) + (const_int 4) (const_int 5) + (const_int 6) (const_int 7) + (const_int 16) (const_int 17) + (const_int 18) (const_int 19) + (const_int 20) (const_int 21) + (const_int 22) (const_int 23) + (const_int 8) (const_int 9) + (const_int 10) (const_int 11) + (const_int 12) (const_int 13) + (const_int 14) (const_int 15) + (const_int 24) (const_int 25) + (const_int 26) (const_int 27) + (const_int 28) (const_int 29) + (const_int 30) (const_int 31)])))] + "TARGET_AVX2 && && " + "vpacksswb\t{%2, %1, %0|%0, %1, %2}" + [(set_attr "type" "sselog") + (set_attr "prefix" "") + (set_attr "mode" "OI")]) + +(define_insn "avx512bw_packsswb" + [(set (match_operand:V64QI 0 "register_operand" "=v") + (vec_select:V64QI + (vec_concat:V64QI + (ss_truncate:V32QI + (match_operand:V32HI 1 "register_operand" "v")) + (ss_truncate:V32QI + (match_operand:V32HI 2 "vector_operand" "vm"))) + (parallel [(const_int 0) (const_int 1) + (const_int 2) (const_int 3) + (const_int 4) (const_int 5) + (const_int 6) (const_int 7) + (const_int 32) (const_int 33) + (const_int 34) (const_int 35) + (const_int 36) (const_int 37) + (const_int 38) (const_int 39) + (const_int 8) (const_int 9) + (const_int 10) (const_int 11) + (const_int 12) (const_int 13) + (const_int 14) (const_int 15) + (const_int 40) (const_int 41) + (const_int 42) (const_int 43) + (const_int 44) (const_int 45) + (const_int 46) (const_int 47) + (const_int 16) (const_int 17) + (const_int 18) (const_int 19) + (const_int 20) (const_int 21) + (const_int 22) (const_int 23) + (const_int 48) (const_int 49) + (const_int 50) (const_int 51) + (const_int 52) (const_int 53) + (const_int 54) (const_int 55) + (const_int 24) (const_int 25) + (const_int 26) (const_int 27) + (const_int 28) (const_int 29) + (const_int 30) (const_int 31) + (const_int 56) (const_int 57) + (const_int 58) (const_int 59) + (const_int 60) (const_int 61) + (const_int 62) (const_int 63)])))] + + "TARGET_AVX512BW" + "vpacksswb\t{%2, %1, %0|%0, %1, %2}" + [(set_attr "type" "sselog") + (set_attr "prefix" "") + (set_attr "mode" "XI")]) + +(define_insn "sse2_packssdw" + [(set (match_operand:V8HI 0 "register_operand" "=x,Yw") + (vec_concat:V8HI + (ss_truncate:V4HI + (match_operand:V4SI 1 "register_operand" "0,Yw")) + (ss_truncate:V4HI + (match_operand:V4SI 2 "vector_operand" "xBm,Ywm"))))] + "TARGET_SSE2 && && " "@ packssdw\t{%2, %0|%0, %2} vpackssdw\t{%2, %1, %0|%0, %1, %2}" @@ -17794,7 +17871,59 @@ (set_attr "type" "sselog") (set_attr "prefix_data16" "1,*") (set_attr "prefix" "orig,") - (set_attr "mode" "")]) + (set_attr "mode" "TI")]) + +(define_insn "avx2_packssdw" + [(set (match_operand:V16HI 0 "register_operand" "=Yw") + (vec_select:V16HI + (vec_concat:V16HI + (ss_truncate:V8HI + (match_operand:V8SI 1 "register_operand" "Yw")) + (ss_truncate:V8HI + (match_operand:V8SI 2 "vector_operand" "Ywm"))) + (parallel [(const_int 0) (const_int 1) + (const_int 2) (const_int 3) + (const_int 8) (const_int 9) + (const_int 10) (const_int 11) + (const_int 4) (const_int 5) + (const_int 6) (const_int 7) + (const_int 12) (const_int 13) + (const_int 14) (const_int 15)])))] + "TARGET_AVX2 && && " + "vpackssdw\t{%2, %1, %0|%0, %1, %2}" + [(set_attr "type" "sselog") + (set_attr "prefix" "") + (set_attr "mode" "OI")]) + +(define_insn "avx512bw_packssdw" + [(set (match_operand:V32HI 0 "register_operand" "=v") + (vec_select:V32HI + (vec_concat:V32HI + (ss_truncate:V16HI + (match_operand:V16SI 1 "register_operand" "v")) + (ss_truncate:V16HI + (match_operand:V16SI 2 "vector_operand" "vm"))) + (parallel [(const_int 0) (const_int 1) + (const_int 2) (const_int 3) + (const_int 16) (const_int 17) + (const_int 18) (const_int 19) + (const_int 4) (const_int 5) + (const_int 6) (const_int 7) + (const_int 20) (const_int 21) + (const_int 22) (const_int 23) + (const_int 8) (const_int 9) + (const_int 10) (const_int 11) + (const_int 24) (const_int 25) + (const_int 26) (const_int 27) + (const_int 12) (const_int 13) + (const_int 14) (const_int 15) + (const_int 28) (const_int 29) + (const_int 30) (const_int 31)])))] + "TARGET_AVX512BW" + "vpackssdw\t{%2, %1, %0|%0, %1, %2}" + [(set_attr "type" "sselog") + (set_attr "prefix" "") + (set_attr "mode" "XI")]) ;; This instruction does unsigned saturation of signed source ;; and is different from generic us_truncate RTX. diff --git a/gcc/testsuite/gcc.target/i386/avx512bw-vpackssdw-3.c b/gcc/testsuite/gcc.target/i386/avx512bw-vpackssdw-3.c new file mode 100644 index 00000000000..ae839e8d2c6 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512bw-vpackssdw-3.c @@ -0,0 +1,55 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -mavx512bw" } */ +/* { dg-require-effective-target avx512bw } */ + +#define AVX512BW +#include "avx512f-helper.h" + +#define DST_SIZE (AVX512F_LEN / 16) +#define SRC_SIZE (AVX512F_LEN / 32) + +#include "limits.h" + +#include "avx512f-mask-type.h" + +static short +int_to_short (int iVal) +{ + short sVal; + + if (iVal < -32768) + sVal = -32768; + else if (iVal > 32767) + sVal = 32767; + else + sVal = iVal; + + return sVal; +} + +void +TEST (void) +{ + union512i_d s1, s2; + union512i_w res1; + short dst_ref[32]; + int i; + + s1.x = _mm512_set_epi32 (1, 2, 3, 4, 65000, 20, 30, 90, 88, 44, 33, 22, 11, 98, 76, -65000); + s2.x = _mm512_set_epi32 (80, 40, 31, 21, 10, 99, 74, -65000, 2, 3, 4, 5, 65010, 21, 31, 91); + res1.x = _mm512_packs_epi32 (s1.x, s2.x); + for (int i = 0; i != 4; i++) + { + dst_ref[i] = int_to_short (s1.a[i]); + dst_ref[i + 4] = int_to_short (s2.a[i]); + dst_ref[i + 8] = int_to_short (s1.a[i + 4]); + dst_ref[i + 12] = int_to_short (s2.a[i + 4]); + dst_ref[i + 16] = int_to_short (s1.a[i + 8]); + dst_ref[i + 20] = int_to_short (s2.a[i + 8]); + dst_ref[i + 24] = int_to_short (s1.a[i + 12]); + dst_ref[i + 28] = int_to_short (s2.a[i + 12]); + } + + if (check_union512i_w (res1, dst_ref)) + abort (); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512bw-vpacksswb-3.c b/gcc/testsuite/gcc.target/i386/avx512bw-vpacksswb-3.c new file mode 100644 index 00000000000..056c735ae0e --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512bw-vpacksswb-3.c @@ -0,0 +1,50 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -mavx512bw" } */ +/* { dg-require-effective-target avx512bw } */ + +#define AVX512BW +#include "avx512f-helper.h" + +static char +short_to_byte (short iVal) +{ + short sVal; + + if (iVal < -128) + sVal = -128; + else if (iVal > 127) + sVal = 127; + else + sVal = iVal; + + return sVal; +} + +void +TEST (void) +{ + union512i_w s1, s2; + union512i_b res1; + char dst_ref[64]; + int i; + + s1.x = _mm512_set_epi16 (1, 2, 3, 4, 650, 20, 30, 90, 88, 44, 33, 22, 11, 98, 76, -650, + 128, 230, -112, -128, -3, -4, -7, 9, 10, 11, 12, 13, -223, 10, 8, 11); + s2.x = _mm512_set_epi16 (80, 40, 31, 21, 10, 99, 74, -650, 2, 3, 4, 5, 650, 21, 31, 91, + 280, -140, 310, 20, 9, 98, 73, -651, 3, 4, 5, 6, 651, 22, 32, 92); + res1.x = _mm512_packs_epi16 (s1.x, s2.x); + for (int i = 0; i != 8; i++) + { + dst_ref[i] = short_to_byte (s1.a[i]); + dst_ref[i + 8] = short_to_byte (s2.a[i]); + dst_ref[i + 16] = short_to_byte (s1.a[i + 8]); + dst_ref[i + 24] = short_to_byte (s2.a[i + 8]); + dst_ref[i + 32] = short_to_byte (s1.a[i + 16]); + dst_ref[i + 40] = short_to_byte (s2.a[i + 16]); + dst_ref[i + 48] = short_to_byte (s1.a[i + 24]); + dst_ref[i + 56] = short_to_byte (s2.a[i + 24]); + } + + if (check_union512i_b (res1, dst_ref)) + abort (); +}