Don't test for sched_yield on Windows #48
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
TeBoring
pushed a commit
to TeBoring/protobuf
that referenced
this pull request
Jan 19, 2019
make googlepb compile with gcc < 4.5.1
copybara-service bot
pushed a commit
that referenced
this pull request
May 31, 2025
There is no need to keep these inline as they are not specialized to the function. We also don't need branch prediction slots for them as the `ret` will always be correctly predicted, and the "end of buffer" branch is rarely taken.
With these changes, the functions to parse fixed fields are about as short and optimal as they could be:
<details>
<summary>x86-64 Assembly</summary>
```
0000000000000000 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte>:
0: 45 84 c9 test r9b, r9b
3: 0f 85 00 00 00 00 jne 0x9 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x9>
0000000000000005: R_X86_64_PLT32 _upb_FastDecoder_DecodeGeneric-0x4
9: 4c 89 c8 mov rax, r9
c: 48 c1 e8 30 shr rax, 0x30
10: 45 89 ca mov r10d, r9d
13: 41 c1 ea 18 shr r10d, 0x18
17: 4d 0f ab d0 bts r8, r10
1b: 44 8b 56 01 mov r10d, dword ptr [rsi + 0x1]
1f: 44 89 14 02 mov dword ptr [rdx + rax], r10d
23: 48 83 c6 05 add rsi, 0x5
27: 48 3b 77 08 cmp rsi, qword ptr [rdi + 0x8]
2b: 0f 83 00 00 00 00 jae 0x31 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x31>
000000000000002d: R_X86_64_PLT32 upb_DecodeFast_MessageIsDoneFallback-0x4
31: 0f b7 06 movzx eax, word ptr [rsi]
34: 49 89 c9 mov r9, rcx
37: 49 c1 f9 08 sar r9, 0x8
3b: 41 89 c2 mov r10d, eax
3e: 41 21 ca and r10d, ecx
41: 41 81 e2 f8 00 00 00 and r10d, 0xf8
48: 4f 8b 5c 51 20 mov r11, qword ptr [r9 + 2*r10 + 0x20]
4d: 4f 8b 4c 51 18 mov r9, qword ptr [r9 + 2*r10 + 0x18]
52: 49 31 c1 xor r9, rax
55: 41 ff e3 jmp r11
```
</details>
<details>
<summary>ARM64 Assembly</summary>
```
0000000000000000 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte>:
0: f2401cbf tst x5, #0xff
4: 54000040 b.eq 0xc <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0xc>
8: 14000000 b 0x8 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x8>
0000000000000008: R_AARCH64_JUMP26 _upb_FastDecoder_DecodeGeneric
c: d370fca8 lsr x8, x5, #48
10: b840102 ldur w9, [x1, #0x1]
14: 5280002a mov w10, #0x1 // =1
18: 91001421 add x1, x1, #0x5
1c: b8286849 str w9, [x2, x8]
20: d358fca8 lsr x8, x5, #24
24: f9400409 ldr x9, [x0, #0x8]
28: 9ac82148 lsl x8, x10, x8
2c: eb01013f cmp x9, x1
30: 54000149 b.ls 0x58 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x58>
34: 79400029 ldrh w9, [x1]
38: 9348fc6a asr x10, x3, #8
3c: 12001c6b and w11, w3, #0xff
40: aa040104 orr x4, x8, x4
44: 8a09016b and x11, x11, x9
48: 8b0b054a add x10, x10, x11, lsl #1
4c: a941994b ldp x11, x6, [x10, #0x18]
50: ca090165 eor x5, x11, x9
54: d61f00c0 br x6
58: aa040104 orr x4, x8, x4
5c: 14000000 b 0x5c <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x5c>
000000000000005c: R_AARCH64_JUMP26 upb_DecodeFast_MessageIsDoneFallback
```
</details>
PiperOrigin-RevId: 765574494
copybara-service bot
pushed a commit
that referenced
this pull request
May 31, 2025
There is no need to keep these inline as they are not specialized to the function. We also don't need branch prediction slots for them as the `ret` will always be correctly predicted, and the "end of buffer" branch is rarely taken.
With these changes, the functions to parse fixed fields are about as short and optimal as they could be:
<details>
<summary>x86-64 Assembly</summary>
```
0000000000000000 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte>:
0: 45 84 c9 test r9b, r9b
3: 0f 85 00 00 00 00 jne 0x9 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x9>
0000000000000005: R_X86_64_PLT32 _upb_FastDecoder_DecodeGeneric-0x4
9: 4c 89 c8 mov rax, r9
c: 48 c1 e8 30 shr rax, 0x30
10: 45 89 ca mov r10d, r9d
13: 41 c1 ea 18 shr r10d, 0x18
17: 4d 0f ab d0 bts r8, r10
1b: 44 8b 56 01 mov r10d, dword ptr [rsi + 0x1]
1f: 44 89 14 02 mov dword ptr [rdx + rax], r10d
23: 48 83 c6 05 add rsi, 0x5
27: 48 3b 77 08 cmp rsi, qword ptr [rdi + 0x8]
2b: 0f 83 00 00 00 00 jae 0x31 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x31>
000000000000002d: R_X86_64_PLT32 upb_DecodeFast_MessageIsDoneFallback-0x4
31: 0f b7 06 movzx eax, word ptr [rsi]
34: 49 89 c9 mov r9, rcx
37: 49 c1 f9 08 sar r9, 0x8
3b: 41 89 c2 mov r10d, eax
3e: 41 21 ca and r10d, ecx
41: 41 81 e2 f8 00 00 00 and r10d, 0xf8
48: 4f 8b 5c 51 20 mov r11, qword ptr [r9 + 2*r10 + 0x20]
4d: 4f 8b 4c 51 18 mov r9, qword ptr [r9 + 2*r10 + 0x18]
52: 49 31 c1 xor r9, rax
55: 41 ff e3 jmp r11
```
</details>
<details>
<summary>ARM64 Assembly</summary>
```
0000000000000000 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte>:
0: f2401cbf tst x5, #0xff
4: 54000040 b.eq 0xc <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0xc>
8: 14000000 b 0x8 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x8>
0000000000000008: R_AARCH64_JUMP26 _upb_FastDecoder_DecodeGeneric
c: d370fca8 lsr x8, x5, #48
10: b840102 ldur w9, [x1, #0x1]
14: 5280002a mov w10, #0x1 // =1
18: 91001421 add x1, x1, #0x5
1c: b8286849 str w9, [x2, x8]
20: d358fca8 lsr x8, x5, #24
24: f9400409 ldr x9, [x0, #0x8]
28: 9ac82148 lsl x8, x10, x8
2c: eb01013f cmp x9, x1
30: 54000149 b.ls 0x58 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x58>
34: 79400029 ldrh w9, [x1]
38: 9348fc6a asr x10, x3, #8
3c: 12001c6b and w11, w3, #0xff
40: aa040104 orr x4, x8, x4
44: 8a09016b and x11, x11, x9
48: 8b0b054a add x10, x10, x11, lsl #1
4c: a941994b ldp x11, x6, [x10, #0x18]
50: ca090165 eor x5, x11, x9
54: d61f00c0 br x6
58: aa040104 orr x4, x8, x4
5c: 14000000 b 0x5c <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x5c>
000000000000005c: R_AARCH64_JUMP26 upb_DecodeFast_MessageIsDoneFallback
```
</details>
PiperOrigin-RevId: 765574494
copybara-service bot
pushed a commit
that referenced
this pull request
May 31, 2025
There is no need to keep these inline as they are not specialized to the function. We also don't need branch prediction slots for them as the `ret` will always be correctly predicted, and the "end of buffer" branch is rarely taken.
With these changes, the functions to parse fixed fields are about as short and optimal as they could be:
<details>
<summary>x86-64 Assembly</summary>
```
0000000000000000 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte>:
0: 45 84 c9 test r9b, r9b
3: 0f 85 00 00 00 00 jne 0x9 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x9>
0000000000000005: R_X86_64_PLT32 _upb_FastDecoder_DecodeGeneric-0x4
9: 4c 89 c8 mov rax, r9
c: 48 c1 e8 30 shr rax, 0x30
10: 45 89 ca mov r10d, r9d
13: 41 c1 ea 18 shr r10d, 0x18
17: 4d 0f ab d0 bts r8, r10
1b: 44 8b 56 01 mov r10d, dword ptr [rsi + 0x1]
1f: 44 89 14 02 mov dword ptr [rdx + rax], r10d
23: 48 83 c6 05 add rsi, 0x5
27: 48 3b 77 08 cmp rsi, qword ptr [rdi + 0x8]
2b: 0f 83 00 00 00 00 jae 0x31 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x31>
000000000000002d: R_X86_64_PLT32 upb_DecodeFast_MessageIsDoneFallback-0x4
31: 0f b7 06 movzx eax, word ptr [rsi]
34: 49 89 c9 mov r9, rcx
37: 49 c1 f9 08 sar r9, 0x8
3b: 41 89 c2 mov r10d, eax
3e: 41 21 ca and r10d, ecx
41: 41 81 e2 f8 00 00 00 and r10d, 0xf8
48: 4f 8b 5c 51 20 mov r11, qword ptr [r9 + 2*r10 + 0x20]
4d: 4f 8b 4c 51 18 mov r9, qword ptr [r9 + 2*r10 + 0x18]
52: 49 31 c1 xor r9, rax
55: 41 ff e3 jmp r11
```
</details>
<details>
<summary>ARM64 Assembly</summary>
```
0000000000000000 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte>:
0: f2401cbf tst x5, #0xff
4: 54000040 b.eq 0xc <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0xc>
8: 14000000 b 0x8 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x8>
0000000000000008: R_AARCH64_JUMP26 _upb_FastDecoder_DecodeGeneric
c: d370fca8 lsr x8, x5, #48
10: b840102 ldur w9, [x1, #0x1]
14: 5280002a mov w10, #0x1 // =1
18: 91001421 add x1, x1, #0x5
1c: b8286849 str w9, [x2, x8]
20: d358fca8 lsr x8, x5, #24
24: f9400409 ldr x9, [x0, #0x8]
28: 9ac82148 lsl x8, x10, x8
2c: eb01013f cmp x9, x1
30: 54000149 b.ls 0x58 <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x58>
34: 79400029 ldrh w9, [x1]
38: 9348fc6a asr x10, x3, #8
3c: 12001c6b and w11, w3, #0xff
40: aa040104 orr x4, x8, x4
44: 8a09016b and x11, x11, x9
48: 8b0b054a add x10, x10, x11, lsl #1
4c: a941994b ldp x11, x6, [x10, #0x18]
50: ca090165 eor x5, x11, x9
54: d61f00c0 br x6
58: aa040104 orr x4, x8, x4
5c: 14000000 b 0x5c <upb_DecodeFast_Fixed32_Scalar_Tag1Byte+0x5c>
000000000000005c: R_AARCH64_JUMP26 upb_DecodeFast_MessageIsDoneFallback
```
</details>
PiperOrigin-RevId: 765632464
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Only test for sched_yield on non-Windows platforms, since it breaks under MingW. sched_yield ought to be tested for on platforms other than Solaris, since it is required. Fixes #45.