106 lines
14 KiB
Plaintext
106 lines
14 KiB
Plaintext
Open Source Architecture Code Analyzer (OSACA) - 0.4.12
|
|
Analyzed file: force_lj_icx_avx2_markers.s
|
|
Architecture: ZEN3
|
|
Timestamp: 2022-12-12 12:47:07
|
|
|
|
|
|
P - Throughput of LOAD operation can be hidden behind a past or future STORE instruction
|
|
* - Instruction micro-ops not bound to a port
|
|
X - No throughput/latency information for this instruction in data file
|
|
|
|
|
|
Combined Analysis Report
|
|
------------------------
|
|
Port pressure in cycles
|
|
| 0 | 1 | 2 | 3 | DV0 | DV1 | 4 | 5 | 6 | 7 | 8 - 8DV | 9 | 10 | 11 | 12 | 13 || CP | LCD |
|
|
---------------------------------------------------------------------------------------------------------------------------------------------
|
|
172 | | | | | | | | | | | | | | | | || | | .LBB0_9: #
|
|
173 | | | | | | | | | | | | | | | | || | | # Parent Loop BB0_6 Depth=1
|
|
174 | | | | | | | | | | | | | | | | || | | # => This Inner Loop Header: Depth=2
|
|
175 | | 0.250 | 0.75 | | | | | | | | | | | 0.50 | 0.50 | || 1.0 | | vpbroadcastd .LCPI0_1(%rip), %xmm1 # xmm1 = [3,3,3,3]
|
|
176 | 0.00 | | | 1.00 | | | | | | | | | | 0.50 | 0.50 | || 3.0 | | vpmulld (%r11,%rbp,4), %xmm1, %xmm11
|
|
177 | 0.00 | 1.010 | 0.25 | 0.74 | | | | | | | | | | | | || 4.0 | | vpmovsxdq %xmm11, %ymm1
|
|
178 | | 0.000 | 1.00 | | | | | | | | | | | | | || 1.0 | | vpsllq $3, %ymm1, %ymm1
|
|
179 | 0.00 | 0.000 | 0.49 | 0.51 | | | | | | | | | | | | || 1.0 | | vpaddq %ymm1, %ymm3, %ymm1
|
|
180 | 0.00 | 0.000 | 0.51 | 0.49 | | | | | | | | | | | | || | | vmovq %xmm1, %r14
|
|
181 | | 1.000 | | | | | | | | | | | | | | || | | vpextrq $1, %xmm1, %r9
|
|
182 | | 1.000 | | | | | | | | | | | | | | || 3.0 | | vextracti128 $1, %ymm1, %xmm1
|
|
183 | | | | | | | | | | | | | | 0.50 | 0.50 | || | | vmovsd (%r14), %xmm2 # xmm2 = mem[0],zero
|
|
184 | 0.00 | 0.000 | 0.49 | 0.51 | | | | | | | | | | | | || | | vpsubd .LCPI0_5, %xmm11, %xmm6
|
|
185 | 0.00 | 0.750 | 0.38 | 0.87 | | | | | | | | | | | | || | | vpmovsxdq %xmm6, %ymm6
|
|
186 | | 0.000 | 1.00 | | | | | | | | | | | | | || | | vpsllq $3, %ymm6, %ymm6
|
|
187 | 0.00 | 0.000 | 0.50 | 0.50 | | | | | | | | | | | | || | | vmovq %xmm1, %rdi
|
|
188 | 0.00 | 0.000 | 0.50 | 0.50 | | | | | | | | | | | | || | | vpaddq %ymm6, %ymm3, %ymm6
|
|
189 | 0.00 | 0.000 | 0.50 | 0.50 | | | | | | | | | | | | || | | vmovq %xmm6, %rcx
|
|
190 | | 1.000 | | | | | | | | | | | | | | || 6.0 | | vpextrq $1, %xmm1, %rbx
|
|
191 | | 1.000 | | | | | | | | | | | | | | || | | vpextrq $1, %xmm6, %rax
|
|
192 | | 1.000 | | | | | | | | | | | | | | || | | vextracti128 $1, %ymm6, %xmm1
|
|
193 | | | | | | | | | | | | | | 0.50 | 0.50 | || | | vmovsd (%rdi), %xmm6 # xmm6 = mem[0],zero
|
|
194 | 0.00 | 0.000 | 0.00 | 1.00 | | | | | | | | | | | | || | | vmovq %xmm1, %rdi
|
|
195 | | 1.000 | | | | | | | | | | | | | | || | | vpextrq $1, %xmm1, %rsi
|
|
196 | | | | | | | | | | | | | | 0.50 | 0.50 | || | | vmovsd (%rdi), %xmm1 # xmm1 = mem[0],zero
|
|
197 | | | | | | | | | | | | | | 0.50 | 0.50 | || | | vmovsd (%rcx), %xmm7 # xmm7 = mem[0],zero
|
|
198 | | 0.000 | 1.00 | | | | | | | | | | | 0.50 | 0.50 | || | | vpbroadcastd .LCPI0_2(%rip), %xmm12 # xmm12 = [2,2,2,2]
|
|
199 | | 0.000 | 1.00 | | | | | | | | | | | 0.50 | 0.50 | || | | vmovhpd (%r9), %xmm2, %xmm2 # xmm2 = xmm2[0],mem[0]
|
|
200 | 0.00 | 0.000 | 0.62 | 0.38 | | | | | | | | | | | | || | | vpaddd %xmm12, %xmm11, %xmm4
|
|
201 | 0.00 | 0.750 | 0.00 | 1.25 | | | | | | | | | | | | || | | vpmovsxdq %xmm4, %ymm4
|
|
202 | | 0.000 | 1.00 | | | | | | | | | | | 0.50 | 0.50 | || | | vmovhpd (%rax), %xmm7, %xmm7 # xmm7 = xmm7[0],mem[0]
|
|
203 | | 0.000 | 1.00 | | | | | | | | | | | | | || | | vpsllq $3, %ymm4, %ymm4
|
|
204 | 0.00 | 0.000 | 0.00 | 1.00 | | | | | | | | | | | | || | | vpaddq %ymm4, %ymm3, %ymm4
|
|
205 | | 0.000 | 1.00 | | | | | | | | | | | 0.50 | 0.50 | || 5.0 | | vmovhpd (%rbx), %xmm6, %xmm6 # xmm6 = xmm6[0],mem[0]
|
|
206 | | 1.000 | | | | | | | | | | | | | | || | | vpextrq $1, %xmm4, %rax
|
|
207 | | 0.000 | 1.00 | | | | | | | | | | | 0.50 | 0.50 | || | | vmovhpd (%rsi), %xmm1, %xmm1 # xmm1 = xmm1[0],mem[0]
|
|
208 | 0.00 | 0.000 | 0.51 | 0.49 | | | | | | | | | | | | || | | vmovq %xmm4, %rcx
|
|
209 | | 1.000 | | | | | | | | | | | | | | || | | vextracti128 $1, %ymm4, %xmm4
|
|
210 | 0.00 | -0.01 | 0.00 | 1.00 | | | | | | | | | | | | || | | vmovq %xmm4, %rsi
|
|
211 | | 1.000 | | | | | | | | | | | | | | || 1.0 | | vinsertf128 $1, %xmm6, %ymm2, %ymm2
|
|
212 | | 1.000 | | | | | | | | | | | | | | || | | vpextrq $1, %xmm4, %rdi
|
|
213 | | | | | | | | | | | | | | 0.50 | 0.50 | || | | vmovsd (%rsi), %xmm4 # xmm4 = mem[0],zero
|
|
214 | | | 0.00 | 1.00 | | | | | | | | | | | | || 3.0 | | vsubpd %ymm2, %ymm14, %ymm2
|
|
215 | | 0.000 | 1.00 | | | | | | | | | | | 0.50 | 0.50 | || | | vmovhpd (%rdi), %xmm4, %xmm4 # xmm4 = xmm4[0],mem[0]
|
|
216 | | | | | | | | | | | | | | 0.50 | 0.50 | || | | vmovsd (%rcx), %xmm6 # xmm6 = mem[0],zero
|
|
217 | | 1.000 | | | | | | | | | | | | | | || | | vinsertf128 $1, %xmm1, %ymm7, %ymm1
|
|
218 | | 0.000 | 1.00 | | | | | | | | | | | 0.50 | 0.50 | || | | vmovhpd (%rax), %xmm6, %xmm6 # xmm6 = xmm6[0],mem[0]
|
|
219 | | 1.000 | | | | | | | | | | | | | | || | | vinsertf128 $1, %xmm4, %ymm6, %ymm4
|
|
220 | | | 0.00 | 1.00 | | | | | | | | | | | | || | | vsubpd %ymm1, %ymm5, %ymm1
|
|
221 | | | 0.00 | 1.00 | | | | | | | | | | | | || | | vsubpd %ymm4, %ymm10, %ymm4
|
|
222 | 1.00 | 0.000 | | | | | | | | | | | | | | || 3.0 | | vmulpd %ymm2, %ymm2, %ymm6
|
|
223 | 1.00 | 0.000 | | | | | | | | | | | | | | || 4.0 | | vfmadd231pd %ymm1, %ymm1, %ymm6 # ymm6 = (ymm1 * ymm1) + ymm6
|
|
224 | 1.00 | 0.000 | | | | | | | | | | | | | | || 4.0 | | vfmadd231pd %ymm4, %ymm4, %ymm6 # ymm6 = (ymm4 * ymm4) + ymm6
|
|
225 | 1.00 | | | | | | | | | | | | | 0.50 | 0.50 | || | | vbroadcastsd .LCPI0_3(%rip), %ymm7 # ymm7 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
|
|
226 | | | | | 4.50 | 4.50 | | | | | | | | | | || 13.0 | | vdivpd %ymm6, %ymm7, %ymm7
|
|
227 | 1.00 | 0.000 | | | | | | | | | | | | | | || 3.0 | | vmulpd %ymm7, %ymm7, %ymm11
|
|
228 | 1.00 | 0.000 | | | | | | | | | | | | | | || 3.0 | | vmulpd %ymm9, %ymm11, %ymm11
|
|
229 | 1.00 | | | | | | | | | | | | | 0.50 | 0.50 | || | | vbroadcastsd .LCPI0_4(%rip), %ymm12 # ymm12 = [-5.0E-1,-5.0E-1,-5.0E-1,-5.0E-1]
|
|
230 | 1.00 | 0.000 | | | | | | | | | | | | | | || 3.0 | | vmulpd %ymm7, %ymm11, %ymm11
|
|
231 | | | 0.00 | 1.00 | | | | | | | | | | | | || 3.0 | | vaddpd %ymm12, %ymm11, %ymm12
|
|
232 | 1.00 | 0.000 | | | | | | | | | | | | 0.50 | 0.50 | || | | vmulpd 128(%rsp), %ymm7, %ymm7 # 32-byte Folded Reload
|
|
233 | 1.00 | 0.000 | | | | | | | | | | | | | | || | | vmulpd %ymm7, %ymm11, %ymm7
|
|
234 | 1.00 | 0.000 | | | | | | | | | | | | | | || 3.0 | | vmulpd %ymm7, %ymm12, %ymm7
|
|
235 | | | 0.12 | 0.88 | | | | | | | | | | | | || | | vcmpltpd %ymm8, %ymm6, %ymm6
|
|
236 | 1.00 | 0.000 | | | | | | | | | | | | | | || 4.0 | | vfmadd213pd %ymm0, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm2) + ymm0
|
|
237 | 1.00 | 0.000 | | | | | | | | | | | | | | || 1.0 | | vblendvpd %ymm6, %ymm2, %ymm0, %ymm0
|
|
238 | 1.00 | 0.000 | | | | | | | | | | | | | | || | | vfmadd213pd %ymm15, %ymm7, %ymm1 # ymm1 = (ymm7 * ymm1) + ymm15
|
|
239 | 1.00 | 0.000 | | | | | | | | | | | | | | || | 4.0 | vfmadd213pd %ymm13, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm4) + ymm13
|
|
240 | 0.62 | 0.380 | | | | | | | | | | | | | | || | | vblendvpd %ymm6, %ymm1, %ymm15, %ymm15
|
|
241 | 0.50 | 0.500 | | | | | | | | | | | | | | || | 1.0 | vblendvpd %ymm6, %ymm4, %ymm13, %ymm13
|
|
242 | | | | | | | | | 0.25 | 0.25 | 0.25 | 0.25 | | | | || | | addq $4, %rbp
|
|
243 | | | | | | | | | 0.25 | 0.25 | 0.25 | 0.25 | | | | || | | cmpq %rdx, %rbp
|
|
244 | | | | | | | | | 0.00 | | | | 1.00 | | | || | | jb .LBB0_9
|
|
|
|
16.1 15.63 15.6 15.6 4.50 4.50 0.50 0.50 0.50 0.50 9.00 9.00 72 5.0
|
|
|
|
|
|
|
|
|
|
Loop-Carried Dependencies Analysis Report
|
|
-----------------------------------------
|
|
239 | 5.0 | vfmadd213pd %ymm13, %ymm7, %ymm4 # ymm4 = (ymm7 * ymm4) + ymm13| [239, 241]
|
|
238 | 5.0 | vfmadd213pd %ymm15, %ymm7, %ymm1 # ymm1 = (ymm7 * ymm1) + ymm15| [238, 240]
|
|
236 | 5.0 | vfmadd213pd %ymm0, %ymm7, %ymm2 # ymm2 = (ymm7 * ymm2) + ymm0| [236, 237]
|
|
242 | 1.0 | addq $4, %rbp | [242]
|
|
241 | 1.0 | vblendvpd %ymm6, %ymm4, %ymm13, %ymm13| [241]
|
|
240 | 1.0 | vblendvpd %ymm6, %ymm1, %ymm15, %ymm15| [240]
|
|
237 | 1.0 | vblendvpd %ymm6, %ymm2, %ymm0, %ymm0| [237]
|
|
|