OpenBLAS/kernel/loongarch64/loongarch64_asm.S

431 lines
12 KiB
ArmAsm

/*******************************************************************************
Copyright (c) 2023, The OpenBLAS Project
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
3. Neither the name of the OpenBLAS project nor the names of
its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*******************************************************************************/
#if __loongarch_grlen == 64
#define LA_REG int64_t
#define REG_SIZE 8
#define REG_LOG 3
#define PTR_ADDI addi.d
#define PTR_ADD add.d
#define PTR_SUB sub.d
#define PTR_LD ld.d
#define PTR_ST st.d
#define PTR_SLLI slli.d
#define PTR_SRLI srli.d
#define PTR_SRAI srai.d
#define PTR_MUL mul.d
#define PTR_ALSL alsl.d
#elif __loongarch_grlen == 32
#define LA_REG int32_t
#define REG_SIZE 4
#define REG_LOG 2
#define PTR_ADDI addi.w
#define PTR_ADD add.w
#define PTR_SUB sub.w
#define PTR_LD ld.w
#define PTR_ST st.w
#define PTR_SLLI slli.w
#define PTR_SRLI srli.w
#define PTR_SRAI srai.w
#define PTR_MUL mul.w
#define PTR_ALSL alsl.w
#else
// If neither of the above two conditions is supported, it means this is an early
// internal toolchain. To ensure maximum compatibility, the following approach is taken:
#define LA_REG int64_t
#define REG_SIZE 8
#define REG_LOG 3
#define PTR_ADDI addi.d
#define PTR_ADD add.d
#define PTR_SUB sub.d
#define PTR_LD ld.d
#define PTR_ST st.d
#define PTR_SLLI slli.d
#define PTR_SRLI srli.d
#define PTR_SRAI srai.d
#define PTR_MUL mul.d
#define PTR_ALSL alsl.d
#endif
#if __loongarch_frlen == 64
#define FREG_SIZE 8
#define FREG_LOG 3
#define PTR_FLD fld.d
#define PTR_FST fst.d
#elif __loongarch_frlen == 32
#define FREG_SIZE 4
#define FREG_LOG 2
#define PTR_FLD fld.s
#define PTR_FST fst.s
#else
// If neither of the above two conditions is supported, it means this is an early
// internal toolchain. To ensure maximum compatibility, the following approach is taken:
#define FREG_SIZE 8
#define FREG_LOG 3
#define PTR_FLD fld.d
#define PTR_FST fst.d
#endif
// The max registers available to the user which
// do not need to be preserved across calls.
// Ref: https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-CN.html
#define MAX_INT_CALLER_SAVED 17
#define MAX_FP_CALLER_SAVED 24
.altmacro // Enable alternate macro mode
.macro push_if_used regs, fregs
.if \regs > MAX_INT_CALLER_SAVED
PTR_ADDI $sp, $sp, -((\regs - MAX_INT_CALLER_SAVED) << REG_LOG)
push_regs 0, \regs - MAX_INT_CALLER_SAVED - 1
.endif
.if \fregs > MAX_FP_CALLER_SAVED
PTR_ADDI $sp, $sp, -((\fregs - MAX_FP_CALLER_SAVED) << FREG_LOG)
push_fregs 0, \fregs - MAX_FP_CALLER_SAVED - 1
.endif
.endm // End push_if_used
.macro pop_if_used regs, fregs
.if \fregs > MAX_FP_CALLER_SAVED
pop_fregs 0, \fregs - MAX_FP_CALLER_SAVED - 1
PTR_ADDI $sp, $sp, (\fregs - MAX_FP_CALLER_SAVED) << FREG_LOG
.endif
.if \regs > MAX_INT_CALLER_SAVED
pop_regs 0, \regs - MAX_INT_CALLER_SAVED - 1
PTR_ADDI $sp, $sp, (\regs - MAX_INT_CALLER_SAVED) << REG_LOG
.endif
.endm // End pop_if_used
.macro push_regs from, to
PTR_ST $s\()\from, $sp, \from << REG_LOG
.if \to - \from
push_regs %from + 1, \to
.endif
.endm // End push_regs
.macro pop_regs from, to
PTR_LD $s\()\from, $sp, \from << REG_LOG
.if \to - \from
pop_regs %from + 1, \to
.endif
.endm // End pop_regs
.macro push_fregs from, to
PTR_FST $fs\()\from, $sp, \from << FREG_LOG
.if \to - \from
push_fregs %from + 1, \to
.endif
.endm // End push_fregs
.macro pop_fregs from, to
PTR_FLD $fs\()\from, $sp, \from << FREG_LOG
.if \to - \from
pop_fregs %from + 1, \to
.endif
.endm // End pop_fregs
//
// Instruction Related Macros
//
// GLD
//
.macro GLD pre_op:req, suf_op=0, out:req, src:req, offset:req/* imm */, more:vararg
.ifeqs "\suf_op", "0"
\pre_op\()ld \out, \src, \offset
.else
\pre_op\()ld.\suf_op \out, \src, \offset
.endif
.ifnb \more
GLD \pre_op, \suf_op, \more
.endif
.endm
//
// GLD_INC
//
.macro GLD_INC pre_op:req, suf_op=0, inc:req, out:req, src:req, offset:req/* imm */, more:vararg
.ifeqs "\suf_op", "0"
\pre_op\()ld \out, \src, \offset
.else
\pre_op\()ld.\suf_op \out, \src, \offset
.endif
PTR_ADDI \src, \src, \inc
.ifnb \more
GLD_INC \pre_op, \suf_op, \inc, \more
.endif
.endm
//
// GLDX is same as GLD except the stride is a register
//
.macro GLDX pre_op:req, suf_op=0, out:req, src:req, offset:req/* reg */, more:vararg
.ifeqs "\suf_op", "0"
\pre_op\()ldx \out, \src, \offset
.else
\pre_op\()ldx.\suf_op \out, \src, \offset
.endif
.ifnb \more
GLDX \pre_op, \suf_op, \more
.endif
.endm
//
// GLDREPL
//
.macro GLDREPL pre_op:req, suf_op:req, out:req, src:req, offset:req/* imm */, more:vararg
\pre_op\()ldrepl.\suf_op \out, \src, \offset
.ifnb \more
GLDREPL \pre_op, \suf_op, \more
.endif
.endm
//
// GST
//
.macro GST pre_op:req, suf_op=0, src:req, dst:req, offset:req/* imm */, more:vararg
.ifeqs "\suf_op", "0"
\pre_op\()st \src, \dst, \offset
.else
\pre_op\()st.\suf_op \src, \dst, \offset
.endif
.ifnb \more
GST \pre_op, \suf_op, \more
.endif
.endm
//
// GMUL
//
.macro GMUL pre_op, suf_op:req, out:req, in0:req, in1:req, more:vararg
\pre_op\()mul.\suf_op \out, \in0, \in1
.ifnb \more
GMUL \pre_op, \suf_op, \more
.endif
.endm
//
// GMADD
//
.macro GMADD pre_op, suf_op:req, out:req, in0:req, in1:req, in2:req, more:vararg
\pre_op\()madd.\suf_op \out, \in0, \in1, \in2
.ifnb \more
GMADD \pre_op, \suf_op, \more
.endif
.endm
//
// GADD
//
.macro GADD pre_op, suf_op:req, out:req, in0:req, in1:req, more:vararg
\pre_op\()add.\suf_op \out, \in0, \in1
.ifnb \more
GADD \pre_op, \suf_op, \more
.endif
.endm
//
// GADDI
//
.macro GADDI pre_op, suf_op:req, out:req, in0:req, in1:req, more:vararg
\pre_op\()addi.\suf_op \out, \in0, \in1
.ifnb \more
GADDI \pre_op, \suf_op, \more
.endif
.endm
//
// GSUB
//
.macro GSUB pre_op, suf_op:req, out:req, in0:req, in1:req, more:vararg
\pre_op\()sub.\suf_op \out, \in0, \in1
.ifnb \more
GSUB \pre_op, \suf_op, \more
.endif
.endm
//
// GSLLI
//
.macro GSLLI pre_op, suf_op:req, out:req, in0:req, in1:req, more:vararg
\pre_op\()slli.\suf_op \out, \in0, \in1
.ifnb \more
GSLLI \pre_op, \suf_op, \more
.endif
.endm
//
// GINSVE0
//
.macro GINSVE0 pre_op:req, suf_op:req, out:req, in0:req, in1:req, more:vararg
\pre_op\()insve0.\suf_op \out, \in0, \in1
.ifnb \more
GINSVE0 \pre_op, \suf_op, \more
.endif
.endm
//
// GXOR
//
.macro GXOR pre_op:req, suf_op:req, out:req, in0:req, in1:req, more:vararg
\pre_op\()xor.\suf_op \out, \in0, \in1
.ifnb \more
GXOR \pre_op, \suf_op, \more
.endif
.endm
//
// GPERMI
//
.macro GPERMI pre_op:req, suf_op:req, out:req, in0:req, in1:req, more:vararg
\pre_op\()permi.\suf_op \out, \in0, \in1
.ifnb \more
GPERMI \pre_op, \suf_op, \more
.endif
.endm
//
// GNMSUB
//
.macro GNMSUB pre_op:req, suf_op:req, out:req, in0:req, in1:req, in2:req, more:vararg
\pre_op\()nmsub.\suf_op \out, \in0, \in1, \in2
.ifnb \more
GNMSUB \pre_op, \suf_op, \more
.endif
.endm
//
// GPRELD
//
.macro GPRELD in0:req, in1:req, in2:req, more:vararg
preld \in0, \in1, \in2
.ifnb \more
GPRELD \more
.endif
.endm
//
// Compound instructions
//
// GACC: Accumulate the values of vector registers
//
.macro GACC pre_op:req, suf_op:req, out:req, in:req, more:vararg
.ifeqs "\pre_op", "xvf"
xvpermi.q \out, \in, 0x01
\pre_op\()add.\suf_op \in, \out, \in
xvpackod.d \out, \in, \in
\pre_op\()add.\suf_op \out, \out, \in
.ifeqs "\suf_op", "s"
xvpackod.w \in, \out, \out
\pre_op\()add.\suf_op \out, \out, \in
.endif
.endif
.ifeqs "\pre_op", "vf"
vpackod.d \out, \in, \in
\pre_op\()add.\suf_op \out, \out, \in
.ifeqs "\suf_op", "s"
vpackod.w \in, \out, \out
\pre_op\()add.\suf_op \out, \out, \in
.endif
.endif
.ifeqs "\pre_op", "xv"
xvpermi.q \out, \in, 0x01
\pre_op\()add.\suf_op \in, \out, \in
xvpackod.d \out, \in, \in
\pre_op\()add.\suf_op \out, \out, \in
.ifnc "\suf_op", "d"
xvpackod.w \in, \out, \out
\pre_op\()add.\suf_op \out, \out, \in
.ifnc "\suf_op", "w"
xvpackod.h \in, \out, \out
\pre_op\()add.\suf_op \out, \out, \in
.ifnc "\suf_op", "h"
xvpackod.b \in, \out, \out
\pre_op\()add.\suf_op \out, \out, \in
.endif
.endif
.endif
.endif
.ifeqs "\pre_op", "v"
vpackod.d \out, \in, \in
\pre_op\()add.\suf_op \out, \out, \in
.ifnc "\suf_op", "d"
vpackod.w \in, \out, \out
\pre_op\()add.\suf_op \out, \out, \in
.ifnc "\suf_op", "w"
vpackod.h \in, \out, \out
\pre_op\()add.\suf_op \out, \out, \in
.ifnc "\suf_op", "h"
vpackod.b \in, \out, \out
\pre_op\()add.\suf_op \out, \out, \in
.endif
.endif
.endif
.endif
.ifnb \more
GACC \pre_op, \suf_op, \more
.endif
.endm
//
// GMOV
//
.macro GMOV pre_op:req, out:req, in:req, more:vararg
\pre_op\()or.v \out, \in, \in
.ifnb \more
GMOV \pre_op, \more
.endif
.endm
//
// Media Related Macros
//
.macro GSBUTTERFLY pre_op, suf_op, out0, out1, in0, in1
\pre_op\()ilvl.\suf_op \out0, \in0, \in1
\pre_op\()ilvh.\suf_op \out1, \in0, \in1
.endm
.macro GINTERLACE pre_op, suf_op, out0, out1, in0, in1
\pre_op\()pickev.\suf_op \out0, \in0, \in1
\pre_op\()pickod.\suf_op \out1, \in0, \in1
.endm
//
// TRANSPOSE4x4_D: Transpose 4x4 block with double-word elements in vectors,
// has no pre_op param. 128-bit vector instructions are not supported.
//
.macro GTRANSPOSE4x4_D in0, in1, in2, in3, out0, out1, out2, out3, \
vt0, vt1
GSBUTTERFLY xv, d, \vt0, \out1, \in1, \in0
GSBUTTERFLY xv, d, \vt1, \out3, \in3, \in2
GMOV xv, \out0, \vt0, \out2, \vt1, \vt1, \out3
GPERMI xv, q, \out0, \out2, 0x02, \out2, \vt0, 0x31, \out3, \out1, 0x31, \out1, \vt1, 0x02
.endm
.macro GTRANSPOSE8x8_W out0, out1, out2, out3, out4, out5, out6, out7, \
in0, in1, in2, in3, in4, in5, in6, in7, \
tmp0, tmp1, tmp2, tmp3
GSBUTTERFLY xv, w, \tmp0, \tmp2, \in2, \in0
GSBUTTERFLY xv, w, \tmp1, \tmp3, \in3, \in1
GSBUTTERFLY xv, w, \out0, \out1, \tmp1, \tmp0
GSBUTTERFLY xv, w, \out2, \out3, \tmp3, \tmp2
GSBUTTERFLY xv, w, \tmp0, \tmp2, \in6, \in4
GSBUTTERFLY xv, w, \tmp1, \tmp3, \in7, \in5
GSBUTTERFLY xv, w, \out4, \out5, \tmp1, \tmp0
GSBUTTERFLY xv, w, \out6, \out7, \tmp3, \tmp2
GMOV xv, \tmp0, \out0, \tmp1, \out1, \tmp2, \out2, \tmp3, \out3
GPERMI xv, q, \out0, \out4, 0x02, \out1, \out5, 0x02, \
\out2, \out6, 0x02, \out3, \out7, 0x02, \
\out4, \tmp0, 0x31, \out5, \tmp1, 0x31, \
\out6, \tmp2, 0x31, \out7, \tmp3, 0x31
.endm