1. update clientset, deepcopy using code-generator
2. add a dummy file tools.go to force "go mod vendor" to see code-generator as dependencies 3. add a script to update CRD 4. add a README to document CRD updating steps run go mod tidy update README
This commit is contained in:
xiangqian
151
vendor/gonum.org/v1/gonum/internal/asm/c64/axpyinc_amd64.s
generated
vendored
Normal file
151
vendor/gonum.org/v1/gonum/internal/asm/c64/axpyinc_amd64.s
generated
vendored
Normal file
@@ -0,0 +1,151 @@
|
||||
// Copyright ©2016 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
//+build !noasm,!appengine,!safe
|
||||
|
||||
#include "textflag.h"
|
||||
|
||||
// MOVSHDUP X3, X2
|
||||
#define MOVSHDUP_X3_X2 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xD3
|
||||
// MOVSLDUP X3, X3
|
||||
#define MOVSLDUP_X3_X3 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xDB
|
||||
// ADDSUBPS X2, X3
|
||||
#define ADDSUBPS_X2_X3 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xDA
|
||||
|
||||
// MOVSHDUP X5, X4
|
||||
#define MOVSHDUP_X5_X4 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xE5
|
||||
// MOVSLDUP X5, X5
|
||||
#define MOVSLDUP_X5_X5 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xED
|
||||
// ADDSUBPS X4, X5
|
||||
#define ADDSUBPS_X4_X5 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xEC
|
||||
|
||||
// MOVSHDUP X7, X6
|
||||
#define MOVSHDUP_X7_X6 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xF7
|
||||
// MOVSLDUP X7, X7
|
||||
#define MOVSLDUP_X7_X7 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xFF
|
||||
// ADDSUBPS X6, X7
|
||||
#define ADDSUBPS_X6_X7 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xFE
|
||||
|
||||
// MOVSHDUP X9, X8
|
||||
#define MOVSHDUP_X9_X8 BYTE $0xF3; BYTE $0x45; BYTE $0x0F; BYTE $0x16; BYTE $0xC1
|
||||
// MOVSLDUP X9, X9
|
||||
#define MOVSLDUP_X9_X9 BYTE $0xF3; BYTE $0x45; BYTE $0x0F; BYTE $0x12; BYTE $0xC9
|
||||
// ADDSUBPS X8, X9
|
||||
#define ADDSUBPS_X8_X9 BYTE $0xF2; BYTE $0x45; BYTE $0x0F; BYTE $0xD0; BYTE $0xC8
|
||||
|
||||
// func AxpyInc(alpha complex64, x, y []complex64, n, incX, incY, ix, iy uintptr)
|
||||
TEXT ·AxpyInc(SB), NOSPLIT, $0
|
||||
MOVQ x_base+8(FP), SI // SI = &x
|
||||
MOVQ y_base+32(FP), DI // DI = &y
|
||||
MOVQ n+56(FP), CX // CX = n
|
||||
CMPQ CX, $0 // if n==0 { return }
|
||||
JE axpyi_end
|
||||
MOVQ ix+80(FP), R8 // R8 = ix
|
||||
MOVQ iy+88(FP), R9 // R9 = iy
|
||||
LEAQ (SI)(R8*8), SI // SI = &(x[ix])
|
||||
LEAQ (DI)(R9*8), DI // DI = &(y[iy])
|
||||
MOVQ DI, DX // DX = DI // Read/Write pointers
|
||||
MOVQ incX+64(FP), R8 // R8 = incX
|
||||
SHLQ $3, R8 // R8 *= sizeof(complex64)
|
||||
MOVQ incY+72(FP), R9 // R9 = incY
|
||||
SHLQ $3, R9 // R9 *= sizeof(complex64)
|
||||
MOVSD alpha+0(FP), X0 // X0 = { 0, 0, imag(a), real(a) }
|
||||
MOVAPS X0, X1
|
||||
SHUFPS $0x11, X1, X1 // X1 = { 0, 0, real(a), imag(a) }
|
||||
MOVAPS X0, X10 // Copy X0 and X1 for pipelining
|
||||
MOVAPS X1, X11
|
||||
MOVQ CX, BX
|
||||
ANDQ $3, CX // CX = n % 4
|
||||
SHRQ $2, BX // BX = floor( n / 4 )
|
||||
JZ axpyi_tail // if BX == 0 { goto axpyi_tail }
|
||||
|
||||
axpyi_loop: // do {
|
||||
MOVSD (SI), X3 // X_i = { imag(x[i+1]), real(x[i+1]) }
|
||||
MOVSD (SI)(R8*1), X5
|
||||
LEAQ (SI)(R8*2), SI // SI = &(SI[incX*2])
|
||||
MOVSD (SI), X7
|
||||
MOVSD (SI)(R8*1), X9
|
||||
|
||||
// X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSHDUP_X3_X2
|
||||
MOVSHDUP_X5_X4
|
||||
MOVSHDUP_X7_X6
|
||||
MOVSHDUP_X9_X8
|
||||
|
||||
// X_i = { real(x[i]), real(x[i]) }
|
||||
MOVSLDUP_X3_X3
|
||||
MOVSLDUP_X5_X5
|
||||
MOVSLDUP_X7_X7
|
||||
MOVSLDUP_X9_X9
|
||||
|
||||
// X_(i-1) = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
|
||||
// X_i = { imag(a) * real(x[i]), real(a) * real(x[i]) }
|
||||
MULPS X1, X2
|
||||
MULPS X0, X3
|
||||
MULPS X11, X4
|
||||
MULPS X10, X5
|
||||
MULPS X1, X6
|
||||
MULPS X0, X7
|
||||
MULPS X11, X8
|
||||
MULPS X10, X9
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(a)*real(x[i]) + real(a)*imag(x[i]),
|
||||
// real(result[i]): real(a)*real(x[i]) - imag(a)*imag(x[i]),
|
||||
// }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDSUBPS_X4_X5
|
||||
ADDSUBPS_X6_X7
|
||||
ADDSUBPS_X8_X9
|
||||
|
||||
// X_i = { imag(result[i]) + imag(y[i]), real(result[i]) + real(y[i]) }
|
||||
MOVSD (DX), X2
|
||||
MOVSD (DX)(R9*1), X4
|
||||
LEAQ (DX)(R9*2), DX // DX = &(DX[incY*2])
|
||||
MOVSD (DX), X6
|
||||
MOVSD (DX)(R9*1), X8
|
||||
ADDPS X2, X3
|
||||
ADDPS X4, X5
|
||||
ADDPS X6, X7
|
||||
ADDPS X8, X9
|
||||
|
||||
MOVSD X3, (DI) // y[i] = X_i
|
||||
MOVSD X5, (DI)(R9*1)
|
||||
LEAQ (DI)(R9*2), DI // DI = &(DI[incDst])
|
||||
MOVSD X7, (DI)
|
||||
MOVSD X9, (DI)(R9*1)
|
||||
LEAQ (SI)(R8*2), SI // SI = &(SI[incX*2])
|
||||
LEAQ (DX)(R9*2), DX // DX = &(DX[incY*2])
|
||||
LEAQ (DI)(R9*2), DI // DI = &(DI[incDst])
|
||||
DECQ BX
|
||||
JNZ axpyi_loop // } while --BX > 0
|
||||
CMPQ CX, $0 // if CX == 0 { return }
|
||||
JE axpyi_end
|
||||
|
||||
axpyi_tail: // do {
|
||||
MOVSD (SI), X3 // X_i = { imag(x[i+1]), real(x[i+1]) }
|
||||
MOVSHDUP_X3_X2 // X_(i-1) = { real(x[i]), real(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X_i = { imag(x[i]), imag(x[i]) }
|
||||
|
||||
// X_i = { imag(a) * real(x[i]), real(a) * real(x[i]) }
|
||||
// X_(i-1) = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
|
||||
MULPS X1, X2
|
||||
MULPS X0, X3
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(a)*real(x[i]) + real(a)*imag(x[i]),
|
||||
// real(result[i]): real(a)*real(x[i]) - imag(a)*imag(x[i]),
|
||||
// }
|
||||
ADDSUBPS_X2_X3 // (ai*x1r+ar*x1i, ar*x1r-ai*x1i)
|
||||
|
||||
// X_i = { imag(result[i]) + imag(y[i]), real(result[i]) + real(y[i]) }
|
||||
MOVSD (DI), X4
|
||||
ADDPS X4, X3
|
||||
MOVSD X3, (DI) // y[i] = X_i
|
||||
ADDQ R8, SI // SI += incX
|
||||
ADDQ R9, DI // DI += incY
|
||||
LOOP axpyi_tail // } while --CX > 0
|
||||
|
||||
axpyi_end:
|
||||
RET
|
||||
156
vendor/gonum.org/v1/gonum/internal/asm/c64/axpyincto_amd64.s
generated
vendored
Normal file
156
vendor/gonum.org/v1/gonum/internal/asm/c64/axpyincto_amd64.s
generated
vendored
Normal file
@@ -0,0 +1,156 @@
|
||||
// Copyright ©2016 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
//+build !noasm,!appengine,!safe
|
||||
|
||||
#include "textflag.h"
|
||||
|
||||
// MOVSHDUP X3, X2
|
||||
#define MOVSHDUP_X3_X2 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xD3
|
||||
// MOVSLDUP X3, X3
|
||||
#define MOVSLDUP_X3_X3 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xDB
|
||||
// ADDSUBPS X2, X3
|
||||
#define ADDSUBPS_X2_X3 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xDA
|
||||
|
||||
// MOVSHDUP X5, X4
|
||||
#define MOVSHDUP_X5_X4 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xE5
|
||||
// MOVSLDUP X5, X5
|
||||
#define MOVSLDUP_X5_X5 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xED
|
||||
// ADDSUBPS X4, X5
|
||||
#define ADDSUBPS_X4_X5 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xEC
|
||||
|
||||
// MOVSHDUP X7, X6
|
||||
#define MOVSHDUP_X7_X6 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xF7
|
||||
// MOVSLDUP X7, X7
|
||||
#define MOVSLDUP_X7_X7 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xFF
|
||||
// ADDSUBPS X6, X7
|
||||
#define ADDSUBPS_X6_X7 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xFE
|
||||
|
||||
// MOVSHDUP X9, X8
|
||||
#define MOVSHDUP_X9_X8 BYTE $0xF3; BYTE $0x45; BYTE $0x0F; BYTE $0x16; BYTE $0xC1
|
||||
// MOVSLDUP X9, X9
|
||||
#define MOVSLDUP_X9_X9 BYTE $0xF3; BYTE $0x45; BYTE $0x0F; BYTE $0x12; BYTE $0xC9
|
||||
// ADDSUBPS X8, X9
|
||||
#define ADDSUBPS_X8_X9 BYTE $0xF2; BYTE $0x45; BYTE $0x0F; BYTE $0xD0; BYTE $0xC8
|
||||
|
||||
// func AxpyIncTo(dst []complex64, incDst, idst uintptr, alpha complex64, x, y []complex64, n, incX, incY, ix, iy uintptr)
|
||||
TEXT ·AxpyIncTo(SB), NOSPLIT, $0
|
||||
MOVQ dst_base+0(FP), DI // DI = &dst
|
||||
MOVQ x_base+48(FP), SI // SI = &x
|
||||
MOVQ y_base+72(FP), DX // DX = &y
|
||||
MOVQ n+96(FP), CX // CX = n
|
||||
CMPQ CX, $0 // if n==0 { return }
|
||||
JE axpyi_end
|
||||
MOVQ ix+120(FP), R8 // Load the first index
|
||||
MOVQ iy+128(FP), R9
|
||||
MOVQ idst+32(FP), R10
|
||||
LEAQ (SI)(R8*8), SI // SI = &(x[ix])
|
||||
LEAQ (DX)(R9*8), DX // DX = &(y[iy])
|
||||
LEAQ (DI)(R10*8), DI // DI = &(dst[idst])
|
||||
MOVQ incX+104(FP), R8 // Incrementors*8 for easy iteration (ADDQ)
|
||||
SHLQ $3, R8
|
||||
MOVQ incY+112(FP), R9
|
||||
SHLQ $3, R9
|
||||
MOVQ incDst+24(FP), R10
|
||||
SHLQ $3, R10
|
||||
MOVSD alpha+40(FP), X0 // X0 = { 0, 0, imag(a), real(a) }
|
||||
MOVAPS X0, X1
|
||||
SHUFPS $0x11, X1, X1 // X1 = { 0, 0, real(a), imag(a) }
|
||||
MOVAPS X0, X10 // Copy X0 and X1 for pipelining
|
||||
MOVAPS X1, X11
|
||||
MOVQ CX, BX
|
||||
ANDQ $3, CX // CX = n % 4
|
||||
SHRQ $2, BX // BX = floor( n / 4 )
|
||||
JZ axpyi_tail // if BX == 0 { goto axpyi_tail }
|
||||
|
||||
axpyi_loop: // do {
|
||||
MOVSD (SI), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSD (SI)(R8*1), X5
|
||||
LEAQ (SI)(R8*2), SI // SI = &(SI[incX*2])
|
||||
MOVSD (SI), X7
|
||||
MOVSD (SI)(R8*1), X9
|
||||
|
||||
// X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSHDUP_X3_X2
|
||||
MOVSHDUP_X5_X4
|
||||
MOVSHDUP_X7_X6
|
||||
MOVSHDUP_X9_X8
|
||||
|
||||
// X_i = { real(x[i]), real(x[i]) }
|
||||
MOVSLDUP_X3_X3
|
||||
MOVSLDUP_X5_X5
|
||||
MOVSLDUP_X7_X7
|
||||
MOVSLDUP_X9_X9
|
||||
|
||||
// X_(i-1) = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
|
||||
// X_i = { imag(a) * real(x[i]), real(a) * real(x[i]) }
|
||||
MULPS X1, X2
|
||||
MULPS X0, X3
|
||||
MULPS X11, X4
|
||||
MULPS X10, X5
|
||||
MULPS X1, X6
|
||||
MULPS X0, X7
|
||||
MULPS X11, X8
|
||||
MULPS X10, X9
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(a)*real(x[i]) + real(a)*imag(x[i]),
|
||||
// real(result[i]): real(a)*real(x[i]) - imag(a)*imag(x[i]),
|
||||
// }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDSUBPS_X4_X5
|
||||
ADDSUBPS_X6_X7
|
||||
ADDSUBPS_X8_X9
|
||||
|
||||
// X_i = { imag(result[i]) + imag(y[i]), real(result[i]) + real(y[i]) }
|
||||
MOVSD (DX), X2
|
||||
MOVSD (DX)(R9*1), X4
|
||||
LEAQ (DX)(R9*2), DX // DX = &(DX[incY*2])
|
||||
MOVSD (DX), X6
|
||||
MOVSD (DX)(R9*1), X8
|
||||
ADDPS X2, X3
|
||||
ADDPS X4, X5
|
||||
ADDPS X6, X7
|
||||
ADDPS X8, X9
|
||||
|
||||
MOVSD X3, (DI) // y[i] = X_i
|
||||
MOVSD X5, (DI)(R10*1)
|
||||
LEAQ (DI)(R10*2), DI // DI = &(DI[incDst])
|
||||
MOVSD X7, (DI)
|
||||
MOVSD X9, (DI)(R10*1)
|
||||
LEAQ (SI)(R8*2), SI // SI = &(SI[incX*2])
|
||||
LEAQ (DX)(R9*2), DX // DX = &(DX[incY*2])
|
||||
LEAQ (DI)(R10*2), DI // DI = &(DI[incDst])
|
||||
DECQ BX
|
||||
JNZ axpyi_loop // } while --BX > 0
|
||||
CMPQ CX, $0 // if CX == 0 { return }
|
||||
JE axpyi_end
|
||||
|
||||
axpyi_tail:
|
||||
MOVSD (SI), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X_i = { real(x[i]), real(x[i]) }
|
||||
|
||||
// X_i = { imag(a) * real(x[i]), real(a) * real(x[i]) }
|
||||
// X_(i-1) = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
|
||||
MULPS X1, X2
|
||||
MULPS X0, X3
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(a)*real(x[i]) + real(a)*imag(x[i]),
|
||||
// real(result[i]): real(a)*real(x[i]) - imag(a)*imag(x[i]),
|
||||
// }
|
||||
ADDSUBPS_X2_X3
|
||||
|
||||
// X_i = { imag(result[i]) + imag(y[i]), real(result[i]) + real(y[i]) }
|
||||
MOVSD (DX), X4
|
||||
ADDPS X4, X3
|
||||
MOVSD X3, (DI) // y[i] = X_i
|
||||
ADDQ R8, SI // SI += incX
|
||||
ADDQ R9, DX // DX += incY
|
||||
ADDQ R10, DI // DI += incDst
|
||||
LOOP axpyi_tail // } while --CX > 0
|
||||
|
||||
axpyi_end:
|
||||
RET
|
||||
160
vendor/gonum.org/v1/gonum/internal/asm/c64/axpyunitary_amd64.s
generated
vendored
Normal file
160
vendor/gonum.org/v1/gonum/internal/asm/c64/axpyunitary_amd64.s
generated
vendored
Normal file
@@ -0,0 +1,160 @@
|
||||
// Copyright ©2016 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
//+build !noasm,!appengine,!safe
|
||||
|
||||
#include "textflag.h"
|
||||
|
||||
// MOVSHDUP X3, X2
|
||||
#define MOVSHDUP_X3_X2 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xD3
|
||||
// MOVSLDUP X3, X3
|
||||
#define MOVSLDUP_X3_X3 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xDB
|
||||
// ADDSUBPS X2, X3
|
||||
#define ADDSUBPS_X2_X3 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xDA
|
||||
|
||||
// MOVSHDUP X5, X4
|
||||
#define MOVSHDUP_X5_X4 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xE5
|
||||
// MOVSLDUP X5, X5
|
||||
#define MOVSLDUP_X5_X5 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xED
|
||||
// ADDSUBPS X4, X5
|
||||
#define ADDSUBPS_X4_X5 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xEC
|
||||
|
||||
// MOVSHDUP X7, X6
|
||||
#define MOVSHDUP_X7_X6 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xF7
|
||||
// MOVSLDUP X7, X7
|
||||
#define MOVSLDUP_X7_X7 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xFF
|
||||
// ADDSUBPS X6, X7
|
||||
#define ADDSUBPS_X6_X7 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xFE
|
||||
|
||||
// MOVSHDUP X9, X8
|
||||
#define MOVSHDUP_X9_X8 BYTE $0xF3; BYTE $0x45; BYTE $0x0F; BYTE $0x16; BYTE $0xC1
|
||||
// MOVSLDUP X9, X9
|
||||
#define MOVSLDUP_X9_X9 BYTE $0xF3; BYTE $0x45; BYTE $0x0F; BYTE $0x12; BYTE $0xC9
|
||||
// ADDSUBPS X8, X9
|
||||
#define ADDSUBPS_X8_X9 BYTE $0xF2; BYTE $0x45; BYTE $0x0F; BYTE $0xD0; BYTE $0xC8
|
||||
|
||||
// func AxpyUnitary(alpha complex64, x, y []complex64)
|
||||
TEXT ·AxpyUnitary(SB), NOSPLIT, $0
|
||||
MOVQ x_base+8(FP), SI // SI = &x
|
||||
MOVQ y_base+32(FP), DI // DI = &y
|
||||
MOVQ x_len+16(FP), CX // CX = min( len(x), len(y) )
|
||||
CMPQ y_len+40(FP), CX
|
||||
CMOVQLE y_len+40(FP), CX
|
||||
CMPQ CX, $0 // if CX == 0 { return }
|
||||
JE caxy_end
|
||||
PXOR X0, X0 // Clear work registers and cache-align loop
|
||||
PXOR X1, X1
|
||||
MOVSD alpha+0(FP), X0 // X0 = { 0, 0, imag(a), real(a) }
|
||||
SHUFPD $0, X0, X0 // X0 = { imag(a), real(a), imag(a), real(a) }
|
||||
MOVAPS X0, X1
|
||||
SHUFPS $0x11, X1, X1 // X1 = { real(a), imag(a), real(a), imag(a) }
|
||||
XORQ AX, AX // i = 0
|
||||
MOVQ DI, BX // Align on 16-byte boundary for ADDPS
|
||||
ANDQ $15, BX // BX = &y & 15
|
||||
JZ caxy_no_trim // if BX == 0 { goto caxy_no_trim }
|
||||
|
||||
// Trim first value in unaligned buffer
|
||||
XORPS X2, X2 // Clear work registers and cache-align loop
|
||||
XORPS X3, X3
|
||||
XORPS X4, X4
|
||||
MOVSD (SI)(AX*8), X3 // X3 = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X2 = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X3 = { real(x[i]), real(x[i]) }
|
||||
MULPS X1, X2 // X2 = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
|
||||
MULPS X0, X3 // X3 = { imag(a) * real(x[i]), real(a) * real(x[i]) }
|
||||
|
||||
// X3 = { imag(a)*real(x[i]) + real(a)*imag(x[i]), real(a)*real(x[i]) - imag(a)*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
MOVSD (DI)(AX*8), X4 // X3 += y[i]
|
||||
ADDPS X4, X3
|
||||
MOVSD X3, (DI)(AX*8) // y[i] = X3
|
||||
INCQ AX // i++
|
||||
DECQ CX // --CX
|
||||
JZ caxy_end // if CX == 0 { return }
|
||||
|
||||
caxy_no_trim:
|
||||
MOVAPS X0, X10 // Copy X0 and X1 for pipelineing
|
||||
MOVAPS X1, X11
|
||||
MOVQ CX, BX
|
||||
ANDQ $7, CX // CX = n % 8
|
||||
SHRQ $3, BX // BX = floor( n / 8 )
|
||||
JZ caxy_tail // if BX == 0 { goto caxy_tail }
|
||||
|
||||
caxy_loop: // do {
|
||||
// X_i = { imag(x[i]), real(x[i]), imag(x[i+1]), real(x[i+1]) }
|
||||
MOVUPS (SI)(AX*8), X3
|
||||
MOVUPS 16(SI)(AX*8), X5
|
||||
MOVUPS 32(SI)(AX*8), X7
|
||||
MOVUPS 48(SI)(AX*8), X9
|
||||
|
||||
// X_(i-1) = { imag(x[i]), imag(x[i]), imag(x[i]+1), imag(x[i]+1) }
|
||||
MOVSHDUP_X3_X2
|
||||
MOVSHDUP_X5_X4
|
||||
MOVSHDUP_X7_X6
|
||||
MOVSHDUP_X9_X8
|
||||
|
||||
// X_i = { real(x[i]), real(x[i]), real(x[i+1]), real(x[i+1]) }
|
||||
MOVSLDUP_X3_X3
|
||||
MOVSLDUP_X5_X5
|
||||
MOVSLDUP_X7_X7
|
||||
MOVSLDUP_X9_X9
|
||||
|
||||
// X_i = { imag(a) * real(x[i]), real(a) * real(x[i]),
|
||||
// imag(a) * real(x[i+1]), real(a) * real(x[i+1]) }
|
||||
// X_(i-1) = { real(a) * imag(x[i]), imag(a) * imag(x[i]),
|
||||
// real(a) * imag(x[i+1]), imag(a) * imag(x[i+1]) }
|
||||
MULPS X1, X2
|
||||
MULPS X0, X3
|
||||
MULPS X11, X4
|
||||
MULPS X10, X5
|
||||
MULPS X1, X6
|
||||
MULPS X0, X7
|
||||
MULPS X11, X8
|
||||
MULPS X10, X9
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(a)*real(x[i]) + real(a)*imag(x[i]),
|
||||
// real(result[i]): real(a)*real(x[i]) - imag(a)*imag(x[i]),
|
||||
// imag(result[i+1]): imag(a)*real(x[i+1]) + real(a)*imag(x[i+1]),
|
||||
// real(result[i+1]): real(a)*real(x[i+1]) - imag(a)*imag(x[i+1]),
|
||||
// }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDSUBPS_X4_X5
|
||||
ADDSUBPS_X6_X7
|
||||
ADDSUBPS_X8_X9
|
||||
|
||||
// X_i = { imag(result[i]) + imag(y[i]), real(result[i]) + real(y[i]),
|
||||
// imag(result[i+1]) + imag(y[i+1]), real(result[i+1]) + real(y[i+1]) }
|
||||
ADDPS (DI)(AX*8), X3
|
||||
ADDPS 16(DI)(AX*8), X5
|
||||
ADDPS 32(DI)(AX*8), X7
|
||||
ADDPS 48(DI)(AX*8), X9
|
||||
MOVUPS X3, (DI)(AX*8) // y[i:i+1] = X_i
|
||||
MOVUPS X5, 16(DI)(AX*8)
|
||||
MOVUPS X7, 32(DI)(AX*8)
|
||||
MOVUPS X9, 48(DI)(AX*8)
|
||||
ADDQ $8, AX // i += 8
|
||||
DECQ BX // --BX
|
||||
JNZ caxy_loop // } while BX > 0
|
||||
CMPQ CX, $0 // if CX == 0 { return }
|
||||
JE caxy_end
|
||||
|
||||
caxy_tail: // do {
|
||||
MOVSD (SI)(AX*8), X3 // X3 = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X2 = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X3 = { real(x[i]), real(x[i]) }
|
||||
MULPS X1, X2 // X2 = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
|
||||
MULPS X0, X3 // X3 = { imag(a) * real(x[i]), real(a) * real(x[i]) }
|
||||
|
||||
// X3 = { imag(a)*real(x[i]) + real(a)*imag(x[i]),
|
||||
// real(a)*real(x[i]) - imag(a)*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
MOVSD (DI)(AX*8), X4 // X3 += y[i]
|
||||
ADDPS X4, X3
|
||||
MOVSD X3, (DI)(AX*8) // y[i] = X3
|
||||
INCQ AX // ++i
|
||||
LOOP caxy_tail // } while --CX > 0
|
||||
|
||||
caxy_end:
|
||||
RET
|
||||
157
vendor/gonum.org/v1/gonum/internal/asm/c64/axpyunitaryto_amd64.s
generated
vendored
Normal file
157
vendor/gonum.org/v1/gonum/internal/asm/c64/axpyunitaryto_amd64.s
generated
vendored
Normal file
@@ -0,0 +1,157 @@
|
||||
// Copyright ©2016 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
//+build !noasm,!appengine,!safe
|
||||
|
||||
#include "textflag.h"
|
||||
|
||||
// MOVSHDUP X3, X2
|
||||
#define MOVSHDUP_X3_X2 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xD3
|
||||
// MOVSLDUP X3, X3
|
||||
#define MOVSLDUP_X3_X3 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xDB
|
||||
// ADDSUBPS X2, X3
|
||||
#define ADDSUBPS_X2_X3 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xDA
|
||||
|
||||
// MOVSHDUP X5, X4
|
||||
#define MOVSHDUP_X5_X4 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xE5
|
||||
// MOVSLDUP X5, X5
|
||||
#define MOVSLDUP_X5_X5 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xED
|
||||
// ADDSUBPS X4, X5
|
||||
#define ADDSUBPS_X4_X5 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xEC
|
||||
|
||||
// MOVSHDUP X7, X6
|
||||
#define MOVSHDUP_X7_X6 BYTE $0xF3; BYTE $0x0F; BYTE $0x16; BYTE $0xF7
|
||||
// MOVSLDUP X7, X7
|
||||
#define MOVSLDUP_X7_X7 BYTE $0xF3; BYTE $0x0F; BYTE $0x12; BYTE $0xFF
|
||||
// ADDSUBPS X6, X7
|
||||
#define ADDSUBPS_X6_X7 BYTE $0xF2; BYTE $0x0F; BYTE $0xD0; BYTE $0xFE
|
||||
|
||||
// MOVSHDUP X9, X8
|
||||
#define MOVSHDUP_X9_X8 BYTE $0xF3; BYTE $0x45; BYTE $0x0F; BYTE $0x16; BYTE $0xC1
|
||||
// MOVSLDUP X9, X9
|
||||
#define MOVSLDUP_X9_X9 BYTE $0xF3; BYTE $0x45; BYTE $0x0F; BYTE $0x12; BYTE $0xC9
|
||||
// ADDSUBPS X8, X9
|
||||
#define ADDSUBPS_X8_X9 BYTE $0xF2; BYTE $0x45; BYTE $0x0F; BYTE $0xD0; BYTE $0xC8
|
||||
|
||||
// func AxpyUnitaryTo(dst []complex64, alpha complex64, x, y []complex64)
|
||||
TEXT ·AxpyUnitaryTo(SB), NOSPLIT, $0
|
||||
MOVQ dst_base+0(FP), DI // DI = &dst
|
||||
MOVQ x_base+32(FP), SI // SI = &x
|
||||
MOVQ y_base+56(FP), DX // DX = &y
|
||||
MOVQ x_len+40(FP), CX
|
||||
CMPQ y_len+64(FP), CX // CX = min( len(x), len(y), len(dst) )
|
||||
CMOVQLE y_len+64(FP), CX
|
||||
CMPQ dst_len+8(FP), CX
|
||||
CMOVQLE dst_len+8(FP), CX
|
||||
CMPQ CX, $0 // if CX == 0 { return }
|
||||
JE caxy_end
|
||||
MOVSD alpha+24(FP), X0 // X0 = { 0, 0, imag(a), real(a) }
|
||||
SHUFPD $0, X0, X0 // X0 = { imag(a), real(a), imag(a), real(a) }
|
||||
MOVAPS X0, X1
|
||||
SHUFPS $0x11, X1, X1 // X1 = { real(a), imag(a), real(a), imag(a) }
|
||||
XORQ AX, AX // i = 0
|
||||
MOVQ DX, BX // Align on 16-byte boundary for ADDPS
|
||||
ANDQ $15, BX // BX = &y & 15
|
||||
JZ caxy_no_trim // if BX == 0 { goto caxy_no_trim }
|
||||
|
||||
MOVSD (SI)(AX*8), X3 // X3 = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X2 = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X3 = { real(x[i]), real(x[i]) }
|
||||
MULPS X1, X2 // X2 = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
|
||||
MULPS X0, X3 // X3 = { imag(a) * real(x[i]), real(a) * real(x[i]) }
|
||||
|
||||
// X3 = { imag(a)*real(x[i]) + real(a)*imag(x[i]), real(a)*real(x[i]) - imag(a)*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
MOVSD (DX)(AX*8), X4 // X3 += y[i]
|
||||
ADDPS X4, X3
|
||||
MOVSD X3, (DI)(AX*8) // dst[i] = X3
|
||||
INCQ AX // i++
|
||||
DECQ CX // --CX
|
||||
JZ caxy_tail // if BX == 0 { goto caxy_tail }
|
||||
|
||||
caxy_no_trim:
|
||||
MOVAPS X0, X10 // Copy X0 and X1 for pipelineing
|
||||
MOVAPS X1, X11
|
||||
MOVQ CX, BX
|
||||
ANDQ $7, CX // CX = n % 8
|
||||
SHRQ $3, BX // BX = floor( n / 8 )
|
||||
JZ caxy_tail // if BX == 0 { goto caxy_tail }
|
||||
|
||||
caxy_loop:
|
||||
// X_i = { imag(x[i]), real(x[i]), imag(x[i+1]), real(x[i+1]) }
|
||||
MOVUPS (SI)(AX*8), X3
|
||||
MOVUPS 16(SI)(AX*8), X5
|
||||
MOVUPS 32(SI)(AX*8), X7
|
||||
MOVUPS 48(SI)(AX*8), X9
|
||||
|
||||
// X_(i-1) = { imag(x[i]), imag(x[i]), imag(x[i]+1), imag(x[i]+1) }
|
||||
MOVSHDUP_X3_X2
|
||||
MOVSHDUP_X5_X4
|
||||
MOVSHDUP_X7_X6
|
||||
MOVSHDUP_X9_X8
|
||||
|
||||
// X_i = { real(x[i]), real(x[i]), real(x[i+1]), real(x[i+1]) }
|
||||
MOVSLDUP_X3_X3
|
||||
MOVSLDUP_X5_X5
|
||||
MOVSLDUP_X7_X7
|
||||
MOVSLDUP_X9_X9
|
||||
|
||||
// X_i = { imag(a) * real(x[i]), real(a) * real(x[i]),
|
||||
// imag(a) * real(x[i+1]), real(a) * real(x[i+1]) }
|
||||
// X_(i-1) = { real(a) * imag(x[i]), imag(a) * imag(x[i]),
|
||||
// real(a) * imag(x[i+1]), imag(a) * imag(x[i+1]) }
|
||||
MULPS X1, X2
|
||||
MULPS X0, X3
|
||||
MULPS X11, X4
|
||||
MULPS X10, X5
|
||||
MULPS X1, X6
|
||||
MULPS X0, X7
|
||||
MULPS X11, X8
|
||||
MULPS X10, X9
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(a)*real(x[i]) + real(a)*imag(x[i]),
|
||||
// real(result[i]): real(a)*real(x[i]) - imag(a)*imag(x[i]),
|
||||
// imag(result[i+1]): imag(a)*real(x[i+1]) + real(a)*imag(x[i+1]),
|
||||
// real(result[i+1]): real(a)*real(x[i+1]) - imag(a)*imag(x[i+1]),
|
||||
// }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDSUBPS_X4_X5
|
||||
ADDSUBPS_X6_X7
|
||||
ADDSUBPS_X8_X9
|
||||
|
||||
// X_i = { imag(result[i]) + imag(y[i]), real(result[i]) + real(y[i]),
|
||||
// imag(result[i+1]) + imag(y[i+1]), real(result[i+1]) + real(y[i+1]) }
|
||||
ADDPS (DX)(AX*8), X3
|
||||
ADDPS 16(DX)(AX*8), X5
|
||||
ADDPS 32(DX)(AX*8), X7
|
||||
ADDPS 48(DX)(AX*8), X9
|
||||
MOVUPS X3, (DI)(AX*8) // y[i:i+1] = X_i
|
||||
MOVUPS X5, 16(DI)(AX*8)
|
||||
MOVUPS X7, 32(DI)(AX*8)
|
||||
MOVUPS X9, 48(DI)(AX*8)
|
||||
ADDQ $8, AX // i += 8
|
||||
DECQ BX // --BX
|
||||
JNZ caxy_loop // } while BX > 0
|
||||
CMPQ CX, $0 // if CX == 0 { return }
|
||||
JE caxy_end
|
||||
|
||||
caxy_tail: // do {
|
||||
MOVSD (SI)(AX*8), X3 // X3 = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X2 = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X3 = { real(x[i]), real(x[i]) }
|
||||
MULPS X1, X2 // X2 = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
|
||||
MULPS X0, X3 // X3 = { imag(a) * real(x[i]), real(a) * real(x[i]) }
|
||||
|
||||
// X3 = { imag(a)*real(x[i]) + real(a)*imag(x[i]),
|
||||
// real(a)*real(x[i]) - imag(a)*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
MOVSD (DX)(AX*8), X4 // X3 += y[i]
|
||||
ADDPS X4, X3
|
||||
MOVSD X3, (DI)(AX*8) // y[i] = X3
|
||||
INCQ AX // ++i
|
||||
LOOP caxy_tail // } while --CX > 0
|
||||
|
||||
caxy_end:
|
||||
RET
|
||||
7
vendor/gonum.org/v1/gonum/internal/asm/c64/conj.go
generated
vendored
Normal file
7
vendor/gonum.org/v1/gonum/internal/asm/c64/conj.go
generated
vendored
Normal file
@@ -0,0 +1,7 @@
|
||||
// Copyright ©2015 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
package c64
|
||||
|
||||
func conj(c complex64) complex64 { return complex(real(c), -imag(c)) }
|
||||
6
vendor/gonum.org/v1/gonum/internal/asm/c64/doc.go
generated
vendored
Normal file
6
vendor/gonum.org/v1/gonum/internal/asm/c64/doc.go
generated
vendored
Normal file
@@ -0,0 +1,6 @@
|
||||
// Copyright ©2017 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
// Package c64 provides complex64 vector primitives.
|
||||
package c64 // import "gonum.org/v1/gonum/internal/asm/c64"
|
||||
160
vendor/gonum.org/v1/gonum/internal/asm/c64/dotcinc_amd64.s
generated
vendored
Normal file
160
vendor/gonum.org/v1/gonum/internal/asm/c64/dotcinc_amd64.s
generated
vendored
Normal file
@@ -0,0 +1,160 @@
|
||||
// Copyright ©2016 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
//+build !noasm,!appengine,!safe
|
||||
|
||||
#include "textflag.h"
|
||||
|
||||
#define MOVSHDUP_X3_X2 LONG $0xD3160FF3 // MOVSHDUP X3, X2
|
||||
#define MOVSHDUP_X5_X4 LONG $0xE5160FF3 // MOVSHDUP X5, X4
|
||||
#define MOVSHDUP_X7_X6 LONG $0xF7160FF3 // MOVSHDUP X7, X6
|
||||
#define MOVSHDUP_X9_X8 LONG $0x160F45F3; BYTE $0xC1 // MOVSHDUP X9, X8
|
||||
|
||||
#define MOVSLDUP_X3_X3 LONG $0xDB120FF3 // MOVSLDUP X3, X3
|
||||
#define MOVSLDUP_X5_X5 LONG $0xED120FF3 // MOVSLDUP X5, X5
|
||||
#define MOVSLDUP_X7_X7 LONG $0xFF120FF3 // MOVSLDUP X7, X7
|
||||
#define MOVSLDUP_X9_X9 LONG $0x120F45F3; BYTE $0xC9 // MOVSLDUP X9, X9
|
||||
|
||||
#define ADDSUBPS_X2_X3 LONG $0xDAD00FF2 // ADDSUBPS X2, X3
|
||||
#define ADDSUBPS_X4_X5 LONG $0xECD00FF2 // ADDSUBPS X4, X5
|
||||
#define ADDSUBPS_X6_X7 LONG $0xFED00FF2 // ADDSUBPS X6, X7
|
||||
#define ADDSUBPS_X8_X9 LONG $0xD00F45F2; BYTE $0xC8 // ADDSUBPS X8, X9
|
||||
|
||||
#define X_PTR SI
|
||||
#define Y_PTR DI
|
||||
#define LEN CX
|
||||
#define TAIL BX
|
||||
#define SUM X0
|
||||
#define P_SUM X1
|
||||
#define INC_X R8
|
||||
#define INCx3_X R9
|
||||
#define INC_Y R10
|
||||
#define INCx3_Y R11
|
||||
#define NEG1 X15
|
||||
#define P_NEG1 X14
|
||||
|
||||
// func DotcInc(x, y []complex64, n, incX, incY, ix, iy uintptr) (sum complex64)
|
||||
TEXT ·DotcInc(SB), NOSPLIT, $0
|
||||
MOVQ x_base+0(FP), X_PTR // X_PTR = &x
|
||||
MOVQ y_base+24(FP), Y_PTR // Y_PTR = &y
|
||||
PXOR SUM, SUM // SUM = 0
|
||||
PXOR P_SUM, P_SUM // P_SUM = 0
|
||||
MOVQ n+48(FP), LEN // LEN = n
|
||||
CMPQ LEN, $0 // if LEN == 0 { return }
|
||||
JE dotc_end
|
||||
MOVQ ix+72(FP), INC_X
|
||||
MOVQ iy+80(FP), INC_Y
|
||||
LEAQ (X_PTR)(INC_X*8), X_PTR // X_PTR = &(X_PTR[ix])
|
||||
LEAQ (Y_PTR)(INC_Y*8), Y_PTR // Y_PTR = &(Y_PTR[iy])
|
||||
MOVQ incX+56(FP), INC_X // INC_X = incX * sizeof(complex64)
|
||||
SHLQ $3, INC_X
|
||||
MOVQ incY+64(FP), INC_Y // INC_Y = incY * sizeof(complex64)
|
||||
SHLQ $3, INC_Y
|
||||
MOVSS $(-1.0), NEG1
|
||||
SHUFPS $0, NEG1, NEG1 // { -1, -1, -1, -1 }
|
||||
|
||||
MOVQ LEN, TAIL
|
||||
ANDQ $3, TAIL // TAIL = LEN % 4
|
||||
SHRQ $2, LEN // LEN = floor( LEN / 4 )
|
||||
JZ dotc_tail // if LEN == 0 { goto dotc_tail }
|
||||
|
||||
MOVUPS NEG1, P_NEG1 // Copy NEG1 for pipelining
|
||||
LEAQ (INC_X)(INC_X*2), INCx3_X // INCx3_X = INC_X * 3
|
||||
LEAQ (INC_Y)(INC_Y*2), INCx3_Y // INCx3_Y = INC_Y * 3
|
||||
|
||||
dotc_loop: // do {
|
||||
MOVSD (X_PTR), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSD (X_PTR)(INC_X*1), X5
|
||||
MOVSD (X_PTR)(INC_X*2), X7
|
||||
MOVSD (X_PTR)(INCx3_X*1), X9
|
||||
|
||||
// X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSHDUP_X3_X2
|
||||
MOVSHDUP_X5_X4
|
||||
MOVSHDUP_X7_X6
|
||||
MOVSHDUP_X9_X8
|
||||
|
||||
// X_i = { real(x[i]), real(x[i]) }
|
||||
MOVSLDUP_X3_X3
|
||||
MOVSLDUP_X5_X5
|
||||
MOVSLDUP_X7_X7
|
||||
MOVSLDUP_X9_X9
|
||||
|
||||
// X_(i-1) = { -imag(x[i]), -imag(x[i]) }
|
||||
MULPS NEG1, X2
|
||||
MULPS P_NEG1, X4
|
||||
MULPS NEG1, X6
|
||||
MULPS P_NEG1, X8
|
||||
|
||||
// X_j = { imag(y[i]), real(y[i]) }
|
||||
MOVSD (Y_PTR), X10
|
||||
MOVSD (Y_PTR)(INC_Y*1), X11
|
||||
MOVSD (Y_PTR)(INC_Y*2), X12
|
||||
MOVSD (Y_PTR)(INCx3_Y*1), X13
|
||||
|
||||
// X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
MULPS X10, X3
|
||||
MULPS X11, X5
|
||||
MULPS X12, X7
|
||||
MULPS X13, X9
|
||||
|
||||
// X_j = { real(y[i]), imag(y[i]) }
|
||||
SHUFPS $0xB1, X10, X10
|
||||
SHUFPS $0xB1, X11, X11
|
||||
SHUFPS $0xB1, X12, X12
|
||||
SHUFPS $0xB1, X13, X13
|
||||
|
||||
// X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
MULPS X10, X2
|
||||
MULPS X11, X4
|
||||
MULPS X12, X6
|
||||
MULPS X13, X8
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i]) * real(x[i]) + real(y[i]) * imag(x[i]),
|
||||
// real(result[i]): real(y[i]) * real(x[i]) - imag(y[i]) * imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDSUBPS_X4_X5
|
||||
ADDSUBPS_X6_X7
|
||||
ADDSUBPS_X8_X9
|
||||
|
||||
// SUM += X_i
|
||||
ADDPS X3, SUM
|
||||
ADDPS X5, P_SUM
|
||||
ADDPS X7, SUM
|
||||
ADDPS X9, P_SUM
|
||||
|
||||
LEAQ (X_PTR)(INC_X*4), X_PTR // X_PTR = &(X_PTR[INC_X*4])
|
||||
LEAQ (Y_PTR)(INC_Y*4), Y_PTR // Y_PTR = &(Y_PTR[INC_Y*4])
|
||||
|
||||
DECQ LEN
|
||||
JNZ dotc_loop // } while --LEN > 0
|
||||
|
||||
ADDPS P_SUM, SUM // SUM = { P_SUM + SUM }
|
||||
CMPQ TAIL, $0 // if TAIL == 0 { return }
|
||||
JE dotc_end
|
||||
|
||||
dotc_tail: // do {
|
||||
MOVSD (X_PTR), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X_i = { real(x[i]), real(x[i]) }
|
||||
MULPS NEG1, X2 // X_(i-1) = { -imag(x[i]), imag(x[i]) }
|
||||
MOVUPS (Y_PTR), X10 // X_j = { imag(y[i]), real(y[i]) }
|
||||
MULPS X10, X3 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
SHUFPS $0x1, X10, X10 // X_j = { real(y[i]), imag(y[i]) }
|
||||
MULPS X10, X2 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i])*real(x[i]) + real(y[i])*imag(x[i]),
|
||||
// real(result[i]): real(y[i])*real(x[i]) - imag(y[i])*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDPS X3, SUM // SUM += X_i
|
||||
ADDQ INC_X, X_PTR // X_PTR += INC_X
|
||||
ADDQ INC_Y, Y_PTR // Y_PTR += INC_Y
|
||||
DECQ TAIL
|
||||
JNZ dotc_tail // } while --TAIL > 0
|
||||
|
||||
dotc_end:
|
||||
MOVSD SUM, sum+88(FP) // return SUM
|
||||
RET
|
||||
208
vendor/gonum.org/v1/gonum/internal/asm/c64/dotcunitary_amd64.s
generated
vendored
Normal file
208
vendor/gonum.org/v1/gonum/internal/asm/c64/dotcunitary_amd64.s
generated
vendored
Normal file
@@ -0,0 +1,208 @@
|
||||
// Copyright ©2017 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
//+build !noasm,!appengine,!safe
|
||||
|
||||
#include "textflag.h"
|
||||
|
||||
#define MOVSLDUP_XPTR_IDX_8__X3 LONG $0x1C120FF3; BYTE $0xC6 // MOVSLDUP (SI)(AX*8), X3
|
||||
#define MOVSLDUP_16_XPTR_IDX_8__X5 LONG $0x6C120FF3; WORD $0x10C6 // MOVSLDUP 16(SI)(AX*8), X5
|
||||
#define MOVSLDUP_32_XPTR_IDX_8__X7 LONG $0x7C120FF3; WORD $0x20C6 // MOVSLDUP 32(SI)(AX*8), X7
|
||||
#define MOVSLDUP_48_XPTR_IDX_8__X9 LONG $0x120F44F3; WORD $0xC64C; BYTE $0x30 // MOVSLDUP 48(SI)(AX*8), X9
|
||||
|
||||
#define MOVSHDUP_XPTR_IDX_8__X2 LONG $0x14160FF3; BYTE $0xC6 // MOVSHDUP (SI)(AX*8), X2
|
||||
#define MOVSHDUP_16_XPTR_IDX_8__X4 LONG $0x64160FF3; WORD $0x10C6 // MOVSHDUP 16(SI)(AX*8), X4
|
||||
#define MOVSHDUP_32_XPTR_IDX_8__X6 LONG $0x74160FF3; WORD $0x20C6 // MOVSHDUP 32(SI)(AX*8), X6
|
||||
#define MOVSHDUP_48_XPTR_IDX_8__X8 LONG $0x160F44F3; WORD $0xC644; BYTE $0x30 // MOVSHDUP 48(SI)(AX*8), X8
|
||||
|
||||
#define MOVSHDUP_X3_X2 LONG $0xD3160FF3 // MOVSHDUP X3, X2
|
||||
#define MOVSLDUP_X3_X3 LONG $0xDB120FF3 // MOVSLDUP X3, X3
|
||||
|
||||
#define ADDSUBPS_X2_X3 LONG $0xDAD00FF2 // ADDSUBPS X2, X3
|
||||
#define ADDSUBPS_X4_X5 LONG $0xECD00FF2 // ADDSUBPS X4, X5
|
||||
#define ADDSUBPS_X6_X7 LONG $0xFED00FF2 // ADDSUBPS X6, X7
|
||||
#define ADDSUBPS_X8_X9 LONG $0xD00F45F2; BYTE $0xC8 // ADDSUBPS X8, X9
|
||||
|
||||
#define X_PTR SI
|
||||
#define Y_PTR DI
|
||||
#define LEN CX
|
||||
#define TAIL BX
|
||||
#define SUM X0
|
||||
#define P_SUM X1
|
||||
#define IDX AX
|
||||
#define I_IDX DX
|
||||
#define NEG1 X15
|
||||
#define P_NEG1 X14
|
||||
|
||||
// func DotcUnitary(x, y []complex64) (sum complex64)
|
||||
TEXT ·DotcUnitary(SB), NOSPLIT, $0
|
||||
MOVQ x_base+0(FP), X_PTR // X_PTR = &x
|
||||
MOVQ y_base+24(FP), Y_PTR // Y_PTR = &y
|
||||
PXOR SUM, SUM // SUM = 0
|
||||
PXOR P_SUM, P_SUM // P_SUM = 0
|
||||
MOVQ x_len+8(FP), LEN // LEN = min( len(x), len(y) )
|
||||
CMPQ y_len+32(FP), LEN
|
||||
CMOVQLE y_len+32(FP), LEN
|
||||
CMPQ LEN, $0 // if LEN == 0 { return }
|
||||
JE dotc_end
|
||||
XORQ IDX, IDX // i = 0
|
||||
MOVSS $(-1.0), NEG1
|
||||
SHUFPS $0, NEG1, NEG1 // { -1, -1, -1, -1 }
|
||||
|
||||
MOVQ X_PTR, DX
|
||||
ANDQ $15, DX // DX = &x & 15
|
||||
JZ dotc_aligned // if DX == 0 { goto dotc_aligned }
|
||||
|
||||
MOVSD (X_PTR)(IDX*8), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X_i = { real(x[i]), real(x[i]) }
|
||||
MOVSD (Y_PTR)(IDX*8), X10 // X_j = { imag(y[i]), real(y[i]) }
|
||||
MULPS NEG1, X2 // X_(i-1) = { -imag(x[i]), imag(x[i]) }
|
||||
MULPS X10, X3 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
SHUFPS $0x1, X10, X10 // X_j = { real(y[i]), imag(y[i]) }
|
||||
MULPS X10, X2 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i])*real(x[i]) + real(y[i])*imag(x[i]),
|
||||
// real(result[i]): real(y[i])*real(x[i]) - imag(y[i])*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
|
||||
MOVAPS X3, SUM // SUM = X_i
|
||||
INCQ IDX // IDX++
|
||||
DECQ LEN // LEN--
|
||||
JZ dotc_ret // if LEN == 0 { goto dotc_ret }
|
||||
|
||||
dotc_aligned:
|
||||
MOVQ LEN, TAIL
|
||||
ANDQ $7, TAIL // TAIL = LEN % 8
|
||||
SHRQ $3, LEN // LEN = floor( LEN / 8 )
|
||||
JZ dotc_tail // if LEN == 0 { return }
|
||||
MOVUPS NEG1, P_NEG1 // Copy NEG1 for pipelining
|
||||
|
||||
dotc_loop: // do {
|
||||
MOVSLDUP_XPTR_IDX_8__X3 // X_i = { real(x[i]), real(x[i]), real(x[i+1]), real(x[i+1]) }
|
||||
MOVSLDUP_16_XPTR_IDX_8__X5
|
||||
MOVSLDUP_32_XPTR_IDX_8__X7
|
||||
MOVSLDUP_48_XPTR_IDX_8__X9
|
||||
|
||||
MOVSHDUP_XPTR_IDX_8__X2 // X_(i-1) = { imag(x[i]), imag(x[i]), imag(x[i+1]), imag(x[i+1]) }
|
||||
MOVSHDUP_16_XPTR_IDX_8__X4
|
||||
MOVSHDUP_32_XPTR_IDX_8__X6
|
||||
MOVSHDUP_48_XPTR_IDX_8__X8
|
||||
|
||||
// X_j = { imag(y[i]), real(y[i]), imag(y[i+1]), real(y[i+1]) }
|
||||
MOVUPS (Y_PTR)(IDX*8), X10
|
||||
MOVUPS 16(Y_PTR)(IDX*8), X11
|
||||
MOVUPS 32(Y_PTR)(IDX*8), X12
|
||||
MOVUPS 48(Y_PTR)(IDX*8), X13
|
||||
|
||||
// X_(i-1) = { -imag(x[i]), -imag(x[i]), -imag(x[i]+1), -imag(x[i]+1) }
|
||||
MULPS NEG1, X2
|
||||
MULPS P_NEG1, X4
|
||||
MULPS NEG1, X6
|
||||
MULPS P_NEG1, X8
|
||||
|
||||
// X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]),
|
||||
// imag(y[i+1]) * real(x[i+1]), real(y[i+1]) * real(x[i+1]) }
|
||||
MULPS X10, X3
|
||||
MULPS X11, X5
|
||||
MULPS X12, X7
|
||||
MULPS X13, X9
|
||||
|
||||
// X_j = { real(y[i]), imag(y[i]), real(y[i+1]), imag(y[i+1]) }
|
||||
SHUFPS $0xB1, X10, X10
|
||||
SHUFPS $0xB1, X11, X11
|
||||
SHUFPS $0xB1, X12, X12
|
||||
SHUFPS $0xB1, X13, X13
|
||||
|
||||
// X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]),
|
||||
// real(y[i+1]) * imag(x[i+1]), imag(y[i+1]) * imag(x[i+1]) }
|
||||
MULPS X10, X2
|
||||
MULPS X11, X4
|
||||
MULPS X12, X6
|
||||
MULPS X13, X8
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i]) * real(x[i]) + real(y[i]) * imag(x[i]),
|
||||
// real(result[i]): real(y[i]) * real(x[i]) - imag(y[i]) * imag(x[i]),
|
||||
// imag(result[i+1]): imag(y[i+1]) * real(x[i+1]) + real(y[i+1]) * imag(x[i+1]),
|
||||
// real(result[i+1]): real(y[i+1]) * real(x[i+1]) - imag(y[i+1]) * imag(x[i+1]),
|
||||
// }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDSUBPS_X4_X5
|
||||
ADDSUBPS_X6_X7
|
||||
ADDSUBPS_X8_X9
|
||||
|
||||
// SUM += X_i
|
||||
ADDPS X3, SUM
|
||||
ADDPS X5, P_SUM
|
||||
ADDPS X7, SUM
|
||||
ADDPS X9, P_SUM
|
||||
|
||||
ADDQ $8, IDX // IDX += 8
|
||||
DECQ LEN
|
||||
JNZ dotc_loop // } while --LEN > 0
|
||||
|
||||
ADDPS SUM, P_SUM // P_SUM = { P_SUM[1] + SUM[1], P_SUM[0] + SUM[0] }
|
||||
XORPS SUM, SUM // SUM = 0
|
||||
|
||||
CMPQ TAIL, $0 // if TAIL == 0 { return }
|
||||
JE dotc_end
|
||||
|
||||
dotc_tail:
|
||||
MOVQ TAIL, LEN
|
||||
SHRQ $1, LEN // LEN = floor( LEN / 2 )
|
||||
JZ dotc_tail_one // if LEN == 0 { goto dotc_tail_one }
|
||||
|
||||
dotc_tail_two: // do {
|
||||
MOVSLDUP_XPTR_IDX_8__X3 // X_i = { real(x[i]), real(x[i]), real(x[i+1]), real(x[i+1]) }
|
||||
MOVSHDUP_XPTR_IDX_8__X2 // X_(i-1) = { imag(x[i]), imag(x[i]), imag(x[i]+1), imag(x[i]+1) }
|
||||
MOVUPS (Y_PTR)(IDX*8), X10 // X_j = { imag(y[i]), real(y[i]) }
|
||||
MULPS NEG1, X2 // X_(i-1) = { -imag(x[i]), imag(x[i]) }
|
||||
MULPS X10, X3 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
SHUFPS $0xB1, X10, X10 // X_j = { real(y[i]), imag(y[i]) }
|
||||
MULPS X10, X2 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i])*real(x[i]) + real(y[i])*imag(x[i]),
|
||||
// real(result[i]): real(y[i])*real(x[i]) - imag(y[i])*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
|
||||
ADDPS X3, SUM // SUM += X_i
|
||||
|
||||
ADDQ $2, IDX // IDX += 2
|
||||
DECQ LEN
|
||||
JNZ dotc_tail_two // } while --LEN > 0
|
||||
|
||||
ADDPS SUM, P_SUM // P_SUM = { P_SUM[1] + SUM[1], P_SUM[0] + SUM[0] }
|
||||
XORPS SUM, SUM // SUM = 0
|
||||
|
||||
ANDQ $1, TAIL
|
||||
JZ dotc_end
|
||||
|
||||
dotc_tail_one:
|
||||
MOVSD (X_PTR)(IDX*8), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X_i = { real(x[i]), real(x[i]) }
|
||||
MOVSD (Y_PTR)(IDX*8), X10 // X_j = { imag(y[i]), real(y[i]) }
|
||||
MULPS NEG1, X2 // X_(i-1) = { -imag(x[i]), imag(x[i]) }
|
||||
MULPS X10, X3 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
SHUFPS $0x1, X10, X10 // X_j = { real(y[i]), imag(y[i]) }
|
||||
MULPS X10, X2 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i])*real(x[i]) + real(y[i])*imag(x[i]),
|
||||
// real(result[i]): real(y[i])*real(x[i]) - imag(y[i])*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
|
||||
ADDPS X3, SUM // SUM += X_i
|
||||
|
||||
dotc_end:
|
||||
ADDPS P_SUM, SUM // SUM = { P_SUM[0] + SUM[0] }
|
||||
MOVHLPS P_SUM, P_SUM // P_SUM = { P_SUM[1], P_SUM[1] }
|
||||
ADDPS P_SUM, SUM // SUM = { P_SUM[1] + SUM[0] }
|
||||
|
||||
dotc_ret:
|
||||
MOVSD SUM, sum+48(FP) // return SUM
|
||||
RET
|
||||
148
vendor/gonum.org/v1/gonum/internal/asm/c64/dotuinc_amd64.s
generated
vendored
Normal file
148
vendor/gonum.org/v1/gonum/internal/asm/c64/dotuinc_amd64.s
generated
vendored
Normal file
@@ -0,0 +1,148 @@
|
||||
// Copyright ©2016 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
//+build !noasm,!appengine,!safe
|
||||
|
||||
#include "textflag.h"
|
||||
|
||||
#define MOVSHDUP_X3_X2 LONG $0xD3160FF3 // MOVSHDUP X3, X2
|
||||
#define MOVSHDUP_X5_X4 LONG $0xE5160FF3 // MOVSHDUP X5, X4
|
||||
#define MOVSHDUP_X7_X6 LONG $0xF7160FF3 // MOVSHDUP X7, X6
|
||||
#define MOVSHDUP_X9_X8 LONG $0x160F45F3; BYTE $0xC1 // MOVSHDUP X9, X8
|
||||
|
||||
#define MOVSLDUP_X3_X3 LONG $0xDB120FF3 // MOVSLDUP X3, X3
|
||||
#define MOVSLDUP_X5_X5 LONG $0xED120FF3 // MOVSLDUP X5, X5
|
||||
#define MOVSLDUP_X7_X7 LONG $0xFF120FF3 // MOVSLDUP X7, X7
|
||||
#define MOVSLDUP_X9_X9 LONG $0x120F45F3; BYTE $0xC9 // MOVSLDUP X9, X9
|
||||
|
||||
#define ADDSUBPS_X2_X3 LONG $0xDAD00FF2 // ADDSUBPS X2, X3
|
||||
#define ADDSUBPS_X4_X5 LONG $0xECD00FF2 // ADDSUBPS X4, X5
|
||||
#define ADDSUBPS_X6_X7 LONG $0xFED00FF2 // ADDSUBPS X6, X7
|
||||
#define ADDSUBPS_X8_X9 LONG $0xD00F45F2; BYTE $0xC8 // ADDSUBPS X8, X9
|
||||
|
||||
#define X_PTR SI
|
||||
#define Y_PTR DI
|
||||
#define LEN CX
|
||||
#define TAIL BX
|
||||
#define SUM X0
|
||||
#define P_SUM X1
|
||||
#define INC_X R8
|
||||
#define INCx3_X R9
|
||||
#define INC_Y R10
|
||||
#define INCx3_Y R11
|
||||
|
||||
// func DotuInc(x, y []complex64, n, incX, incY, ix, iy uintptr) (sum complex64)
|
||||
TEXT ·DotuInc(SB), NOSPLIT, $0
|
||||
MOVQ x_base+0(FP), X_PTR // X_PTR = &x
|
||||
MOVQ y_base+24(FP), Y_PTR // Y_PTR = &y
|
||||
PXOR SUM, SUM // SUM = 0
|
||||
PXOR P_SUM, P_SUM // P_SUM = 0
|
||||
MOVQ n+48(FP), LEN // LEN = n
|
||||
CMPQ LEN, $0 // if LEN == 0 { return }
|
||||
JE dotu_end
|
||||
MOVQ ix+72(FP), INC_X
|
||||
MOVQ iy+80(FP), INC_Y
|
||||
LEAQ (X_PTR)(INC_X*8), X_PTR // X_PTR = &(X_PTR[ix])
|
||||
LEAQ (Y_PTR)(INC_Y*8), Y_PTR // Y_PTR = &(Y_PTR[iy])
|
||||
MOVQ incX+56(FP), INC_X // INC_X = incX * sizeof(complex64)
|
||||
SHLQ $3, INC_X
|
||||
MOVQ incY+64(FP), INC_Y // INC_Y = incY * sizeof(complex64)
|
||||
SHLQ $3, INC_Y
|
||||
|
||||
MOVQ LEN, TAIL
|
||||
ANDQ $3, TAIL // TAIL = LEN % 4
|
||||
SHRQ $2, LEN // LEN = floor( LEN / 4 )
|
||||
JZ dotu_tail // if TAIL == 0 { goto dotu_tail }
|
||||
|
||||
LEAQ (INC_X)(INC_X*2), INCx3_X // INCx3_X = INC_X * 3
|
||||
LEAQ (INC_Y)(INC_Y*2), INCx3_Y // INCx3_Y = INC_Y * 3
|
||||
|
||||
dotu_loop: // do {
|
||||
MOVSD (X_PTR), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSD (X_PTR)(INC_X*1), X5
|
||||
MOVSD (X_PTR)(INC_X*2), X7
|
||||
MOVSD (X_PTR)(INCx3_X*1), X9
|
||||
|
||||
// X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSHDUP_X3_X2
|
||||
MOVSHDUP_X5_X4
|
||||
MOVSHDUP_X7_X6
|
||||
MOVSHDUP_X9_X8
|
||||
|
||||
// X_i = { real(x[i]), real(x[i]) }
|
||||
MOVSLDUP_X3_X3
|
||||
MOVSLDUP_X5_X5
|
||||
MOVSLDUP_X7_X7
|
||||
MOVSLDUP_X9_X9
|
||||
|
||||
// X_j = { imag(y[i]), real(y[i]) }
|
||||
MOVSD (Y_PTR), X10
|
||||
MOVSD (Y_PTR)(INC_Y*1), X11
|
||||
MOVSD (Y_PTR)(INC_Y*2), X12
|
||||
MOVSD (Y_PTR)(INCx3_Y*1), X13
|
||||
|
||||
// X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
MULPS X10, X3
|
||||
MULPS X11, X5
|
||||
MULPS X12, X7
|
||||
MULPS X13, X9
|
||||
|
||||
// X_j = { real(y[i]), imag(y[i]) }
|
||||
SHUFPS $0xB1, X10, X10
|
||||
SHUFPS $0xB1, X11, X11
|
||||
SHUFPS $0xB1, X12, X12
|
||||
SHUFPS $0xB1, X13, X13
|
||||
|
||||
// X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
MULPS X10, X2
|
||||
MULPS X11, X4
|
||||
MULPS X12, X6
|
||||
MULPS X13, X8
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i]) * real(x[i]) + real(y[i]) * imag(x[i]),
|
||||
// real(result[i]): real(y[i]) * real(x[i]) - imag(y[i]) * imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDSUBPS_X4_X5
|
||||
ADDSUBPS_X6_X7
|
||||
ADDSUBPS_X8_X9
|
||||
|
||||
// SUM += X_i
|
||||
ADDPS X3, SUM
|
||||
ADDPS X5, P_SUM
|
||||
ADDPS X7, SUM
|
||||
ADDPS X9, P_SUM
|
||||
|
||||
LEAQ (X_PTR)(INC_X*4), X_PTR // X_PTR = &(X_PTR[INC_X*4])
|
||||
LEAQ (Y_PTR)(INC_Y*4), Y_PTR // Y_PTR = &(Y_PTR[INC_Y*4])
|
||||
|
||||
DECQ LEN
|
||||
JNZ dotu_loop // } while --LEN > 0
|
||||
|
||||
ADDPS P_SUM, SUM // SUM = { P_SUM + SUM }
|
||||
CMPQ TAIL, $0 // if TAIL == 0 { return }
|
||||
JE dotu_end
|
||||
|
||||
dotu_tail: // do {
|
||||
MOVSD (X_PTR), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X_i = { real(x[i]), real(x[i]) }
|
||||
MOVUPS (Y_PTR), X10 // X_j = { imag(y[i]), real(y[i]) }
|
||||
MULPS X10, X3 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
SHUFPS $0x1, X10, X10 // X_j = { real(y[i]), imag(y[i]) }
|
||||
MULPS X10, X2 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i])*real(x[i]) + real(y[i])*imag(x[i]),
|
||||
// real(result[i]): real(y[i])*real(x[i]) - imag(y[i])*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDPS X3, SUM // SUM += X_i
|
||||
ADDQ INC_X, X_PTR // X_PTR += INC_X
|
||||
ADDQ INC_Y, Y_PTR // Y_PTR += INC_Y
|
||||
DECQ TAIL
|
||||
JNZ dotu_tail // } while --TAIL > 0
|
||||
|
||||
dotu_end:
|
||||
MOVSD SUM, sum+88(FP) // return SUM
|
||||
RET
|
||||
197
vendor/gonum.org/v1/gonum/internal/asm/c64/dotuunitary_amd64.s
generated
vendored
Normal file
197
vendor/gonum.org/v1/gonum/internal/asm/c64/dotuunitary_amd64.s
generated
vendored
Normal file
@@ -0,0 +1,197 @@
|
||||
// Copyright ©2017 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
//+build !noasm,!appengine,!safe
|
||||
|
||||
#include "textflag.h"
|
||||
|
||||
#define MOVSLDUP_XPTR_IDX_8__X3 LONG $0x1C120FF3; BYTE $0xC6 // MOVSLDUP (SI)(AX*8), X3
|
||||
#define MOVSLDUP_16_XPTR_IDX_8__X5 LONG $0x6C120FF3; WORD $0x10C6 // MOVSLDUP 16(SI)(AX*8), X5
|
||||
#define MOVSLDUP_32_XPTR_IDX_8__X7 LONG $0x7C120FF3; WORD $0x20C6 // MOVSLDUP 32(SI)(AX*8), X7
|
||||
#define MOVSLDUP_48_XPTR_IDX_8__X9 LONG $0x120F44F3; WORD $0xC64C; BYTE $0x30 // MOVSLDUP 48(SI)(AX*8), X9
|
||||
|
||||
#define MOVSHDUP_XPTR_IDX_8__X2 LONG $0x14160FF3; BYTE $0xC6 // MOVSHDUP (SI)(AX*8), X2
|
||||
#define MOVSHDUP_16_XPTR_IDX_8__X4 LONG $0x64160FF3; WORD $0x10C6 // MOVSHDUP 16(SI)(AX*8), X4
|
||||
#define MOVSHDUP_32_XPTR_IDX_8__X6 LONG $0x74160FF3; WORD $0x20C6 // MOVSHDUP 32(SI)(AX*8), X6
|
||||
#define MOVSHDUP_48_XPTR_IDX_8__X8 LONG $0x160F44F3; WORD $0xC644; BYTE $0x30 // MOVSHDUP 48(SI)(AX*8), X8
|
||||
|
||||
#define MOVSHDUP_X3_X2 LONG $0xD3160FF3 // MOVSHDUP X3, X2
|
||||
#define MOVSLDUP_X3_X3 LONG $0xDB120FF3 // MOVSLDUP X3, X3
|
||||
|
||||
#define ADDSUBPS_X2_X3 LONG $0xDAD00FF2 // ADDSUBPS X2, X3
|
||||
#define ADDSUBPS_X4_X5 LONG $0xECD00FF2 // ADDSUBPS X4, X5
|
||||
#define ADDSUBPS_X6_X7 LONG $0xFED00FF2 // ADDSUBPS X6, X7
|
||||
#define ADDSUBPS_X8_X9 LONG $0xD00F45F2; BYTE $0xC8 // ADDSUBPS X8, X9
|
||||
|
||||
#define X_PTR SI
|
||||
#define Y_PTR DI
|
||||
#define LEN CX
|
||||
#define TAIL BX
|
||||
#define SUM X0
|
||||
#define P_SUM X1
|
||||
#define IDX AX
|
||||
#define I_IDX DX
|
||||
#define NEG1 X15
|
||||
#define P_NEG1 X14
|
||||
|
||||
// func DotuUnitary(x, y []complex64) (sum complex64)
|
||||
TEXT ·DotuUnitary(SB), NOSPLIT, $0
|
||||
MOVQ x_base+0(FP), X_PTR // X_PTR = &x
|
||||
MOVQ y_base+24(FP), Y_PTR // Y_PTR = &y
|
||||
PXOR SUM, SUM // SUM = 0
|
||||
PXOR P_SUM, P_SUM // P_SUM = 0
|
||||
MOVQ x_len+8(FP), LEN // LEN = min( len(x), len(y) )
|
||||
CMPQ y_len+32(FP), LEN
|
||||
CMOVQLE y_len+32(FP), LEN
|
||||
CMPQ LEN, $0 // if LEN == 0 { return }
|
||||
JE dotu_end
|
||||
XORQ IDX, IDX // IDX = 0
|
||||
|
||||
MOVQ X_PTR, DX
|
||||
ANDQ $15, DX // DX = &x & 15
|
||||
JZ dotu_aligned // if DX == 0 { goto dotu_aligned }
|
||||
|
||||
MOVSD (X_PTR)(IDX*8), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X_i = { real(x[i]), real(x[i]) }
|
||||
MOVSD (Y_PTR)(IDX*8), X10 // X_j = { imag(y[i]), real(y[i]) }
|
||||
MULPS X10, X3 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
SHUFPS $0x1, X10, X10 // X_j = { real(y[i]), imag(y[i]) }
|
||||
MULPS X10, X2 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i])*real(x[i]) + real(y[i])*imag(x[i]),
|
||||
// real(result[i]): real(y[i])*real(x[i]) - imag(y[i])*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
|
||||
MOVAPS X3, SUM // SUM = X_i
|
||||
INCQ IDX // IDX++
|
||||
DECQ LEN // LEN--
|
||||
JZ dotu_end // if LEN == 0 { goto dotu_end }
|
||||
|
||||
dotu_aligned:
|
||||
MOVQ LEN, TAIL
|
||||
ANDQ $7, TAIL // TAIL = LEN % 8
|
||||
SHRQ $3, LEN // LEN = floor( LEN / 8 )
|
||||
JZ dotu_tail // if LEN == 0 { goto dotu_tail }
|
||||
PXOR P_SUM, P_SUM
|
||||
|
||||
dotu_loop: // do {
|
||||
MOVSLDUP_XPTR_IDX_8__X3 // X_i = { real(x[i]), real(x[i]), real(x[i+1]), real(x[i+1]) }
|
||||
MOVSLDUP_16_XPTR_IDX_8__X5
|
||||
MOVSLDUP_32_XPTR_IDX_8__X7
|
||||
MOVSLDUP_48_XPTR_IDX_8__X9
|
||||
|
||||
MOVSHDUP_XPTR_IDX_8__X2 // X_(i-1) = { imag(x[i]), imag(x[i]), imag(x[i]+1), imag(x[i]+1) }
|
||||
MOVSHDUP_16_XPTR_IDX_8__X4
|
||||
MOVSHDUP_32_XPTR_IDX_8__X6
|
||||
MOVSHDUP_48_XPTR_IDX_8__X8
|
||||
|
||||
// X_j = { imag(y[i]), real(y[i]), imag(y[i+1]), real(y[i+1]) }
|
||||
MOVUPS (Y_PTR)(IDX*8), X10
|
||||
MOVUPS 16(Y_PTR)(IDX*8), X11
|
||||
MOVUPS 32(Y_PTR)(IDX*8), X12
|
||||
MOVUPS 48(Y_PTR)(IDX*8), X13
|
||||
|
||||
// X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]),
|
||||
// imag(y[i+1]) * real(x[i+1]), real(y[i+1]) * real(x[i+1]) }
|
||||
MULPS X10, X3
|
||||
MULPS X11, X5
|
||||
MULPS X12, X7
|
||||
MULPS X13, X9
|
||||
|
||||
// X_j = { real(y[i]), imag(y[i]), real(y[i+1]), imag(y[i+1]) }
|
||||
SHUFPS $0xB1, X10, X10
|
||||
SHUFPS $0xB1, X11, X11
|
||||
SHUFPS $0xB1, X12, X12
|
||||
SHUFPS $0xB1, X13, X13
|
||||
|
||||
// X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]),
|
||||
// real(y[i+1]) * imag(x[i+1]), imag(y[i+1]) * imag(x[i+1]) }
|
||||
MULPS X10, X2
|
||||
MULPS X11, X4
|
||||
MULPS X12, X6
|
||||
MULPS X13, X8
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i]) * real(x[i]) + real(y[i]) * imag(x[i]),
|
||||
// real(result[i]): real(y[i]) * real(x[i]) - imag(y[i]) * imag(x[i]),
|
||||
// imag(result[i+1]): imag(y[i+1]) * real(x[i+1]) + real(y[i+1]) * imag(x[i+1]),
|
||||
// real(result[i+1]): real(y[i+1]) * real(x[i+1]) - imag(y[i+1]) * imag(x[i+1]),
|
||||
// }
|
||||
ADDSUBPS_X2_X3
|
||||
ADDSUBPS_X4_X5
|
||||
ADDSUBPS_X6_X7
|
||||
ADDSUBPS_X8_X9
|
||||
|
||||
// SUM += X_i
|
||||
ADDPS X3, SUM
|
||||
ADDPS X5, P_SUM
|
||||
ADDPS X7, SUM
|
||||
ADDPS X9, P_SUM
|
||||
|
||||
ADDQ $8, IDX // IDX += 8
|
||||
DECQ LEN
|
||||
JNZ dotu_loop // } while --LEN > 0
|
||||
|
||||
ADDPS SUM, P_SUM // P_SUM = { P_SUM[1] + SUM[1], P_SUM[0] + SUM[0] }
|
||||
XORPS SUM, SUM // SUM = 0
|
||||
|
||||
CMPQ TAIL, $0 // if TAIL == 0 { return }
|
||||
JE dotu_end
|
||||
|
||||
dotu_tail:
|
||||
MOVQ TAIL, LEN
|
||||
SHRQ $1, LEN // LEN = floor( LEN / 2 )
|
||||
JZ dotu_tail_one // if LEN == 0 { goto dotc_tail_one }
|
||||
|
||||
dotu_tail_two: // do {
|
||||
MOVSLDUP_XPTR_IDX_8__X3 // X_i = { real(x[i]), real(x[i]), real(x[i+1]), real(x[i+1]) }
|
||||
MOVSHDUP_XPTR_IDX_8__X2 // X_(i-1) = { imag(x[i]), imag(x[i]), imag(x[i]+1), imag(x[i]+1) }
|
||||
MOVUPS (Y_PTR)(IDX*8), X10 // X_j = { imag(y[i]), real(y[i]) }
|
||||
MULPS X10, X3 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
SHUFPS $0xB1, X10, X10 // X_j = { real(y[i]), imag(y[i]) }
|
||||
MULPS X10, X2 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i])*real(x[i]) + real(y[i])*imag(x[i]),
|
||||
// real(result[i]): real(y[i])*real(x[i]) - imag(y[i])*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
|
||||
ADDPS X3, SUM // SUM += X_i
|
||||
|
||||
ADDQ $2, IDX // IDX += 2
|
||||
DECQ LEN
|
||||
JNZ dotu_tail_two // } while --LEN > 0
|
||||
|
||||
ADDPS SUM, P_SUM // P_SUM = { P_SUM[1] + SUM[1], P_SUM[0] + SUM[0] }
|
||||
XORPS SUM, SUM // SUM = 0
|
||||
|
||||
ANDQ $1, TAIL
|
||||
JZ dotu_end
|
||||
|
||||
dotu_tail_one:
|
||||
MOVSD (X_PTR)(IDX*8), X3 // X_i = { imag(x[i]), real(x[i]) }
|
||||
MOVSHDUP_X3_X2 // X_(i-1) = { imag(x[i]), imag(x[i]) }
|
||||
MOVSLDUP_X3_X3 // X_i = { real(x[i]), real(x[i]) }
|
||||
MOVSD (Y_PTR)(IDX*8), X10 // X_j = { imag(y[i]), real(y[i]) }
|
||||
MULPS X10, X3 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
|
||||
SHUFPS $0x1, X10, X10 // X_j = { real(y[i]), imag(y[i]) }
|
||||
MULPS X10, X2 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
|
||||
|
||||
// X_i = {
|
||||
// imag(result[i]): imag(y[i])*real(x[i]) + real(y[i])*imag(x[i]),
|
||||
// real(result[i]): real(y[i])*real(x[i]) - imag(y[i])*imag(x[i]) }
|
||||
ADDSUBPS_X2_X3
|
||||
|
||||
ADDPS X3, SUM // SUM += X_i
|
||||
|
||||
dotu_end:
|
||||
ADDPS P_SUM, SUM // SUM = { P_SUM[0] + SUM[0] }
|
||||
MOVHLPS P_SUM, P_SUM // P_SUM = { P_SUM[1], P_SUM[1] }
|
||||
ADDPS P_SUM, SUM // SUM = { P_SUM[1] + SUM[0] }
|
||||
|
||||
dotu_ret:
|
||||
MOVSD SUM, sum+48(FP) // return SUM
|
||||
RET
|
||||
79
vendor/gonum.org/v1/gonum/internal/asm/c64/scal.go
generated
vendored
Normal file
79
vendor/gonum.org/v1/gonum/internal/asm/c64/scal.go
generated
vendored
Normal file
@@ -0,0 +1,79 @@
|
||||
// Copyright ©2016 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
package c64
|
||||
|
||||
// ScalUnitary is
|
||||
// for i := range x {
|
||||
// x[i] *= alpha
|
||||
// }
|
||||
func ScalUnitary(alpha complex64, x []complex64) {
|
||||
for i := range x {
|
||||
x[i] *= alpha
|
||||
}
|
||||
}
|
||||
|
||||
// ScalUnitaryTo is
|
||||
// for i, v := range x {
|
||||
// dst[i] = alpha * v
|
||||
// }
|
||||
func ScalUnitaryTo(dst []complex64, alpha complex64, x []complex64) {
|
||||
for i, v := range x {
|
||||
dst[i] = alpha * v
|
||||
}
|
||||
}
|
||||
|
||||
// ScalInc is
|
||||
// var ix uintptr
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// x[ix] *= alpha
|
||||
// ix += incX
|
||||
// }
|
||||
func ScalInc(alpha complex64, x []complex64, n, incX uintptr) {
|
||||
var ix uintptr
|
||||
for i := 0; i < int(n); i++ {
|
||||
x[ix] *= alpha
|
||||
ix += incX
|
||||
}
|
||||
}
|
||||
|
||||
// ScalIncTo is
|
||||
// var idst, ix uintptr
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// dst[idst] = alpha * x[ix]
|
||||
// ix += incX
|
||||
// idst += incDst
|
||||
// }
|
||||
func ScalIncTo(dst []complex64, incDst uintptr, alpha complex64, x []complex64, n, incX uintptr) {
|
||||
var idst, ix uintptr
|
||||
for i := 0; i < int(n); i++ {
|
||||
dst[idst] = alpha * x[ix]
|
||||
ix += incX
|
||||
idst += incDst
|
||||
}
|
||||
}
|
||||
|
||||
// SscalUnitary is
|
||||
// for i, v := range x {
|
||||
// x[i] = complex(real(v)*alpha, imag(v)*alpha)
|
||||
// }
|
||||
func SscalUnitary(alpha float32, x []complex64) {
|
||||
for i, v := range x {
|
||||
x[i] = complex(real(v)*alpha, imag(v)*alpha)
|
||||
}
|
||||
}
|
||||
|
||||
// SscalInc is
|
||||
// var ix uintptr
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// x[ix] = complex(real(x[ix])*alpha, imag(x[ix])*alpha)
|
||||
// ix += inc
|
||||
// }
|
||||
func SscalInc(alpha float32, x []complex64, n, inc uintptr) {
|
||||
var ix uintptr
|
||||
for i := 0; i < int(n); i++ {
|
||||
x[ix] = complex(real(x[ix])*alpha, imag(x[ix])*alpha)
|
||||
ix += inc
|
||||
}
|
||||
}
|
||||
68
vendor/gonum.org/v1/gonum/internal/asm/c64/stubs_amd64.go
generated
vendored
Normal file
68
vendor/gonum.org/v1/gonum/internal/asm/c64/stubs_amd64.go
generated
vendored
Normal file
@@ -0,0 +1,68 @@
|
||||
// Copyright ©2016 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
// +build !noasm,!appengine,!safe
|
||||
|
||||
package c64
|
||||
|
||||
// AxpyUnitary is
|
||||
// for i, v := range x {
|
||||
// y[i] += alpha * v
|
||||
// }
|
||||
func AxpyUnitary(alpha complex64, x, y []complex64)
|
||||
|
||||
// AxpyUnitaryTo is
|
||||
// for i, v := range x {
|
||||
// dst[i] = alpha*v + y[i]
|
||||
// }
|
||||
func AxpyUnitaryTo(dst []complex64, alpha complex64, x, y []complex64)
|
||||
|
||||
// AxpyInc is
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// y[iy] += alpha * x[ix]
|
||||
// ix += incX
|
||||
// iy += incY
|
||||
// }
|
||||
func AxpyInc(alpha complex64, x, y []complex64, n, incX, incY, ix, iy uintptr)
|
||||
|
||||
// AxpyIncTo is
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// dst[idst] = alpha*x[ix] + y[iy]
|
||||
// ix += incX
|
||||
// iy += incY
|
||||
// idst += incDst
|
||||
// }
|
||||
func AxpyIncTo(dst []complex64, incDst, idst uintptr, alpha complex64, x, y []complex64, n, incX, incY, ix, iy uintptr)
|
||||
|
||||
// DotcUnitary is
|
||||
// for i, v := range x {
|
||||
// sum += y[i] * conj(v)
|
||||
// }
|
||||
// return sum
|
||||
func DotcUnitary(x, y []complex64) (sum complex64)
|
||||
|
||||
// DotcInc is
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// sum += y[iy] * conj(x[ix])
|
||||
// ix += incX
|
||||
// iy += incY
|
||||
// }
|
||||
// return sum
|
||||
func DotcInc(x, y []complex64, n, incX, incY, ix, iy uintptr) (sum complex64)
|
||||
|
||||
// DotuUnitary is
|
||||
// for i, v := range x {
|
||||
// sum += y[i] * v
|
||||
// }
|
||||
// return sum
|
||||
func DotuUnitary(x, y []complex64) (sum complex64)
|
||||
|
||||
// DotuInc is
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// sum += y[iy] * x[ix]
|
||||
// ix += incX
|
||||
// iy += incY
|
||||
// }
|
||||
// return sum
|
||||
func DotuInc(x, y []complex64, n, incX, incY, ix, iy uintptr) (sum complex64)
|
||||
113
vendor/gonum.org/v1/gonum/internal/asm/c64/stubs_noasm.go
generated
vendored
Normal file
113
vendor/gonum.org/v1/gonum/internal/asm/c64/stubs_noasm.go
generated
vendored
Normal file
@@ -0,0 +1,113 @@
|
||||
// Copyright ©2016 The Gonum Authors. All rights reserved.
|
||||
// Use of this source code is governed by a BSD-style
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
// +build !amd64 noasm appengine safe
|
||||
|
||||
package c64
|
||||
|
||||
// AxpyUnitary is
|
||||
// for i, v := range x {
|
||||
// y[i] += alpha * v
|
||||
// }
|
||||
func AxpyUnitary(alpha complex64, x, y []complex64) {
|
||||
for i, v := range x {
|
||||
y[i] += alpha * v
|
||||
}
|
||||
}
|
||||
|
||||
// AxpyUnitaryTo is
|
||||
// for i, v := range x {
|
||||
// dst[i] = alpha*v + y[i]
|
||||
// }
|
||||
func AxpyUnitaryTo(dst []complex64, alpha complex64, x, y []complex64) {
|
||||
for i, v := range x {
|
||||
dst[i] = alpha*v + y[i]
|
||||
}
|
||||
}
|
||||
|
||||
// AxpyInc is
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// y[iy] += alpha * x[ix]
|
||||
// ix += incX
|
||||
// iy += incY
|
||||
// }
|
||||
func AxpyInc(alpha complex64, x, y []complex64, n, incX, incY, ix, iy uintptr) {
|
||||
for i := 0; i < int(n); i++ {
|
||||
y[iy] += alpha * x[ix]
|
||||
ix += incX
|
||||
iy += incY
|
||||
}
|
||||
}
|
||||
|
||||
// AxpyIncTo is
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// dst[idst] = alpha*x[ix] + y[iy]
|
||||
// ix += incX
|
||||
// iy += incY
|
||||
// idst += incDst
|
||||
// }
|
||||
func AxpyIncTo(dst []complex64, incDst, idst uintptr, alpha complex64, x, y []complex64, n, incX, incY, ix, iy uintptr) {
|
||||
for i := 0; i < int(n); i++ {
|
||||
dst[idst] = alpha*x[ix] + y[iy]
|
||||
ix += incX
|
||||
iy += incY
|
||||
idst += incDst
|
||||
}
|
||||
}
|
||||
|
||||
// DotcUnitary is
|
||||
// for i, v := range x {
|
||||
// sum += y[i] * conj(v)
|
||||
// }
|
||||
// return sum
|
||||
func DotcUnitary(x, y []complex64) (sum complex64) {
|
||||
for i, v := range x {
|
||||
sum += y[i] * conj(v)
|
||||
}
|
||||
return sum
|
||||
}
|
||||
|
||||
// DotcInc is
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// sum += y[iy] * conj(x[ix])
|
||||
// ix += incX
|
||||
// iy += incY
|
||||
// }
|
||||
// return sum
|
||||
func DotcInc(x, y []complex64, n, incX, incY, ix, iy uintptr) (sum complex64) {
|
||||
for i := 0; i < int(n); i++ {
|
||||
sum += y[iy] * conj(x[ix])
|
||||
ix += incX
|
||||
iy += incY
|
||||
}
|
||||
return sum
|
||||
}
|
||||
|
||||
// DotuUnitary is
|
||||
// for i, v := range x {
|
||||
// sum += y[i] * v
|
||||
// }
|
||||
// return sum
|
||||
func DotuUnitary(x, y []complex64) (sum complex64) {
|
||||
for i, v := range x {
|
||||
sum += y[i] * v
|
||||
}
|
||||
return sum
|
||||
}
|
||||
|
||||
// DotuInc is
|
||||
// for i := 0; i < int(n); i++ {
|
||||
// sum += y[iy] * x[ix]
|
||||
// ix += incX
|
||||
// iy += incY
|
||||
// }
|
||||
// return sum
|
||||
func DotuInc(x, y []complex64, n, incX, incY, ix, iy uintptr) (sum complex64) {
|
||||
for i := 0; i < int(n); i++ {
|
||||
sum += y[iy] * x[ix]
|
||||
ix += incX
|
||||
iy += incY
|
||||
}
|
||||
return sum
|
||||
}
|
||||
Reference in New Issue
Block a user