Skip to content
This repository has been archived by the owner on Jan 10, 2023. It is now read-only.

Python API for Myelin #302

Merged
merged 2 commits into from
Dec 4, 2018
Merged

Python API for Myelin #302

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
901b140
Python API for Myelin
ringgaard Nov 30, 2018
79a5827
Fixes for #302
ringgaard Dec 4, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
166 changes: 166 additions & 0 deletions doc/guide/flowasm.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,166 @@
0000000000000000 <f>:
0: push rbp
1: mov rbp,rdi

0000000000000004 <f/MatMul:AVXFltVecMatMulAddRelu>:
4: mov rdi,rbp
7: movabs rsi,0x0 9: R_X86_64_64 f/W
11: movabs r9,0x0 13: R_X86_64_64 f/b
1b: lea r8,[rbp+0x100]
22: vxorps ymm13,ymm13,ymm13
27: xor rcx,rcx
2a: vmovaps ymm0,YMMWORD PTR [r9+rcx*1]
30: vmovaps ymm1,YMMWORD PTR [r9+rcx*1+0x20]
37: vmovaps ymm2,YMMWORD PTR [r9+rcx*1+0x40]
3e: vmovaps ymm3,YMMWORD PTR [r9+rcx*1+0x60]
45: vmovaps ymm4,YMMWORD PTR [r9+rcx*1+0x80]
4f: vmovaps ymm5,YMMWORD PTR [r9+rcx*1+0xa0]
59: vmovaps ymm6,YMMWORD PTR [r9+rcx*1+0xc0]
63: vmovaps ymm7,YMMWORD PTR [r9+rcx*1+0xe0]
6d: mov rdx,rsi
70: xor rax,rax
73: vbroadcastss ymm12,DWORD PTR [rdi+rax*1]
79: vfmadd231ps ymm0,ymm12,YMMWORD PTR [rdx]
7e: vfmadd231ps ymm1,ymm12,YMMWORD PTR [rdx+0x20]
84: vfmadd231ps ymm2,ymm12,YMMWORD PTR [rdx+0x40]
8a: vfmadd231ps ymm3,ymm12,YMMWORD PTR [rdx+0x60]
90: vfmadd231ps ymm4,ymm12,YMMWORD PTR [rdx+0x80]
99: vfmadd231ps ymm5,ymm12,YMMWORD PTR [rdx+0xa0]
a2: vfmadd231ps ymm6,ymm12,YMMWORD PTR [rdx+0xc0]
ab: vfmadd231ps ymm7,ymm12,YMMWORD PTR [rdx+0xe0]
b4: add rdx,0x400
bb: add rax,0x4
bf: cmp rax,0x100
c5: jl 73 <f/MatMul:AVXFltVecMatMulAddRelu+0x6f>
c7: vmaxps ymm0,ymm0,ymm13
cc: vmovaps YMMWORD PTR [r8+rcx*1],ymm0
d2: vmaxps ymm1,ymm1,ymm13
d7: vmovaps YMMWORD PTR [r8+rcx*1+0x20],ymm1
de: vmaxps ymm2,ymm2,ymm13
e3: vmovaps YMMWORD PTR [r8+rcx*1+0x40],ymm2
ea: vmaxps ymm3,ymm3,ymm13
ef: vmovaps YMMWORD PTR [r8+rcx*1+0x60],ymm3
f6: vmaxps ymm4,ymm4,ymm13
fb: vmovaps YMMWORD PTR [r8+rcx*1+0x80],ymm4
105: vmaxps ymm5,ymm5,ymm13
10a: vmovaps YMMWORD PTR [r8+rcx*1+0xa0],ymm5
114: vmaxps ymm6,ymm6,ymm13
119: vmovaps YMMWORD PTR [r8+rcx*1+0xc0],ymm6
123: vmaxps ymm7,ymm7,ymm13
128: vmovaps YMMWORD PTR [r8+rcx*1+0xe0],ymm7
132: add rsi,0x100
139: add rcx,0x100
140: cmp rcx,0x400
147: jl 2a <f/MatMul:AVXFltVecMatMulAddRelu+0x26>

000000000000014d <f/Max:MaxExpr>:
14d: lea rcx,[rbp+0x100]
154: vmovaps ymm0,YMMWORD PTR [rip+0x1a4] # 300 <f_data+0xd>
15c: xor rax,rax
15f: vmaxps ymm0,ymm0,YMMWORD PTR [rcx+rax*1]
164: add rax,0x20
168: cmp rax,0x400
16e: jl 15f <f/Max:MaxExpr+0x12>
170: vperm2f128 ymm1,ymm0,ymm0,0x1
176: vmaxps ymm0,ymm0,ymm1
17a: vpermilps ymm1,ymm0,0xe
180: vmaxps ymm0,ymm0,ymm1
184: vpermilps ymm1,ymm0,0x1
18a: vmaxps ymm0,ymm0,ymm1
18e: vmovss DWORD PTR [rbp+0x500],xmm0

0000000000000196 <f/Sub:Calculate>:
196: lea rcx,[rbp+0x100]
19d: lea rdx,[rbp+0x520]
1a4: vxorps ymm14,ymm14,ymm14
1a9: vmovaps ymm0,YMMWORD PTR [rip+0x16f] # 320 <f_data+0x2d>
1b1: vmovaps ymm1,YMMWORD PTR [rip+0x187] # 340 <f_data+0x4d>
1b9: vmovaps ymm2,YMMWORD PTR [rip+0x19f] # 360 <f_data+0x6d>
1c1: vmovaps ymm3,YMMWORD PTR [rip+0x1b7] # 380 <f_data+0x8d>
1c9: vmovaps ymm4,YMMWORD PTR [rip+0x1cf] # 3a0 <f_data+0xad>
1d1: vmovaps ymm5,YMMWORD PTR [rip+0x1e7] # 3c0 <f_data+0xcd>
1d9: vmovaps ymm6,YMMWORD PTR [rip+0x1ff] # 3e0 <f_data+0xed>
1e1: vmovaps ymm7,YMMWORD PTR [rip+0x217] # 400 <f_data+0x10d>
1e9: vbroadcastss ymm8,DWORD PTR [rbp+0x500]
1f2: xor rax,rax
1f5: vmovaps ymm9,YMMWORD PTR [rcx+rax*1]
1fa: vsubps ymm9,ymm9,ymm8
1ff: vminps ymm10,ymm9,ymm1
203: vmaxps ymm10,ymm10,ymm0
207: vmovaps ymm11,ymm10
20c: vfmadd213ps ymm11,ymm3,ymm2
211: vroundps ymm11,ymm11,0x1
217: vmovaps ymm12,ymm11
21c: vfmadd213ps ymm12,ymm4,ymm10
221: vmulps ymm10,ymm12,ymm12
226: vmovaps ymm13,ymm5
22a: vfmadd213ps ymm13,ymm12,ymm6
22f: vfmadd213ps ymm13,ymm12,ymm7
234: vfmadd213ps ymm13,ymm12,YMMWORD PTR [rip+0x1e3] # 420 <f_data+0x12d>
23d: vfmadd213ps ymm13,ymm12,YMMWORD PTR [rip+0x1fa] # 440 <f_data+0x14d>
246: vfmadd213ps ymm13,ymm12,YMMWORD PTR [rip+0x111] # 360 <f_data+0x6d>
24f: vfmadd213ps ymm13,ymm10,ymm12
254: vaddps ymm13,ymm13,YMMWORD PTR [rip+0x204] # 460 <f_data+0x16d>
25c: vaddps ymm11,ymm11,YMMWORD PTR [rip+0x21c] # 480 <f_data+0x18d>
264: vcvttps2dq ymm11,ymm11
269: vpslld ymm11,ymm11,0x17
26f: vmulps ymm13,ymm13,ymm11
274: vmaxps ymm13,ymm13,ymm9
279: vmovaps YMMWORD PTR [rdx+rax*1],ymm13
27e: vaddps ymm14,ymm14,ymm13
283: add rax,0x20
287: cmp rax,0x400
28d: jl 1f5 <f/Sub:Calculate+0x5f>
293: vperm2f128 ymm15,ymm14,ymm14,0x1
299: vhaddps ymm14,ymm14,ymm15
29e: vhaddps ymm14,ymm14,ymm14
2a3: vhaddps ymm14,ymm14,ymm14
2a8: vmovss DWORD PTR [rbp+0x500],xmm14

00000000000002b0 <f/Reciprocal:ReciprocalExpr>:
2b0: vmovss xmm0,DWORD PTR [rip+0x1e8] # 4a0 <f_data+0x1ad>
2b8: vdivss xmm1,xmm0,DWORD PTR [rbp+0x500]
2c0: vmovss DWORD PTR [rbp+0x500],xmm1

00000000000002c8 <f/y:MulExpr>:
2c8: lea rcx,[rbp+0x520]
2cf: vbroadcastss ymm0,DWORD PTR [rbp+0x500]
2d8: xor rax,rax
2db: vmulps ymm1,ymm0,YMMWORD PTR [rcx+rax*1]
2e0: vmovaps YMMWORD PTR [rcx+rax*1],ymm1
2e5: add rax,0x20
2e9: cmp rax,0x400
2ef: jl 2db <f/y:MulExpr+0x13>
2f1: pop rbp
2f2: ret

00000000000002f3 <f_data>:
...
2ff: ................
30f: ................
31f: ................
32f: ................
33f: ....B...B...B...
34f: B...B...B...B...
35f: B...?...?...?...
36f: ?...?...?...?...
37f: ?;..?;..?;..?;..
38f: ?;..?;..?;..?;..
39f: ?.r1..r1..r1..r1
3af: ..r1..r1..r1..r1
3bf: .giP9giP9giP9giP
3cf: 9giP9giP9giP9giP
3df: 9.C.:.C.:.C.:.C.
3ef: :.C.:.C.:.C.:.C.
3ff: :...<...<...<...
40f: <...<...<...<...
41f: <..*=..*=..*=..*
42f: =..*=..*=..*=..*
43f: =..*>..*>..*>..*
44f: >..*>..*>..*>..*
45f: >...?...?...?...
46f: ?...?...?...?...
47f: ?...B...B...B...
48f: B...B...B...B...
49f: B...?

Loading