[compiler-rt][RISCV] Implement __riscv_ifunc_select #85790

BeMg · 2024-03-19T14:02:53Z

This patch implement the __riscv_ifunc_select base on Hwprobe.

unsigned __riscv_ifunc_select(struct riscv_hwprobe *ReqirePreKey, unsigned Length);

The __riscv_ifunc_select function checks if all required keys and their corresponding values from the caller match the hwprobe result. If all features are available in the current runtime environment, it returns 1. Otherwise, it returns 0.

This patch implement the `__riscv_ifunc_select` base on `Hwprobe/cpuinfo`. ``` unsigned __riscv_ifunc_select(char *FeatStrs); ``` The __riscv_ifunc_select function checks if the features specified in FeatStrs are supported. If all features are available in the current runtime environment, it returns 1. Otherwise, it returns 0.

BeMg · 2024-03-19T14:03:56Z

This patch make #85786 could run some real test.

wangpc-pp · 2024-03-21T11:02:01Z

compiler-rt/lib/builtins/riscv/ifunc_select.c

+#define RISCV_HWPROBE_MISALIGNED_FAST (3 << 0)
+#define RISCV_HWPROBE_MISALIGNED_UNSUPPORTED (4 << 0)
+#define RISCV_HWPROBE_MISALIGNED_MASK (7 << 0)
+


Can we align to the newest interface: https://docs.kernel.org/arch/riscv/hwprobe.html?

BeMg · 2024-03-30T12:13:37Z

Align with latest sys_riscv_hwprobe
Update __riscv_ifunc_select, from __riscv_ifunc_select(char *) to __riscv_ifunc_select(unsigned long long, unsigned long long).
Remove the cpuinfo relate code and string process relate code
Use the bitset method to determine whether a set of extension is available for current environment.

compiler-rt/lib/builtins/riscv/ifunc_select.c

github-actions · 2024-04-01T02:34:26Z

✅ With the latest revision this PR passed the C/C++ code formatter.

wangpc-pp · 2024-04-01T02:52:21Z

Are there any processes for GNU/GCC implementation? If we want to port glibc, I think it should be required.

BeMg · 2024-04-08T04:47:25Z

Let the caller to manage and construct the necessary key/value pairs for hwprobe, eliminating the need for the runtime site to sync with the hwprobe key table.
Modify __riscv_ifunc_select to accept a pointer to riscv_hwprobe and its length, so the prototype does not need to be updated when the hwprobe keys increase or change.

topperc · 2024-04-09T06:42:35Z

compiler-rt/lib/builtins/riscv/ifunc_select.c

+
+#endif // defined(__linux__)
+
+unsigned __riscv_ifunc_select(struct riscv_hwprobe *ReqireKeys,


Reqire -> Require

topperc · 2024-04-09T06:43:29Z

compiler-rt/lib/builtins/riscv/ifunc_select.c

+  }
+
+  // hwprobe not success
+  if (initHwProbe(HwprobePairs, 2))


What is the 2 here?

It should be replaced with Length from the argument. Fixed.

topperc · 2024-04-09T06:46:57Z

compiler-rt/lib/builtins/riscv/ifunc_select.c

+                              unsigned Length) {
+#if defined(__linux__)
+  // Init Hwprobe
+  struct riscv_hwprobe HwprobePairs[64];


Where does 64 come from?

It need a big enough buffer to place the key from __riscv_ifunc_select argument, and pass it into sys_hwprobe.

For now, the RISCV_HWPROBE_MAX_KEY is 6.

https://github.com/torvalds/linux/blob/20cb38a7af88dc40095da7c2c9094da3873fea23/arch/riscv/include/asm/hwprobe.h#L11

wangpc-pp · 2024-04-09T08:21:28Z

compiler-rt/lib/builtins/riscv/ifunc_select.c

+                              unsigned Length) {
+#if defined(__linux__)
+  // Init Hwprobe
+  struct riscv_hwprobe HwprobePairs[Length];


Just a question: are we allowed to use VLA here?

Can we use malloc?

I don't know if this is the convention of compiler-rt development, I think we should avoid using standard libraries here.
VLA is OK here, I just don't know if we can use this feature.

I suspect use VLA here may introduce vulnerabilities since we don't have any length check and this could trigger stack overflow.

Or could we define a MAX_PAIR_LENGTH constant here and discard content exceeding that limit before invoking the system call, or simply return false in this situation?

For example:

Something like

#define MAX_PAIR_LENGTH 6 struct riscv_hwprobe HwprobePairs[MAX_PAIR_LENGTH]; if (Length > MAX_PAIR_LENGTH) Length = MAX_PAIR_LENGTH;

Does this mean that we limit the number of features in target_clones?

No. That limit come from hwprobe system call's key. I expect the compiler will transform the feature inside target_clones into minimum hwprobe system call pair.

The pair from the arguments of __riscv_ifunc_select could be mapped into a struct riscv_hwprobe.

For example:

; query for version "arch=rv64im" // CHECK: @__riscv_hwprobe_args = internal global [2 x %riscv_hwprobe_pair] [%riscv_hwprobe_pair { i64 3, i64 1 }, %riscv_hwprobe_pair { i64 4, i64 0 }] ; query for version "arch=+zbb,+zba,+zbc,+zbkb,+zknd,+zknh,+zksed,+zvksh,+zfh,+zfa,+v" // CHECK: @__riscv_hwprobe_args.1 = internal global [2 x %riscv_hwprobe_pair.0] [%riscv_hwprobe_pair.0 { i64 3, i64 1 }, %riscv_hwprobe_pair.0 { i64 4, i64 4462766492 }] // CHECK-LABEL: define weak_odr ptr @foo1.resolver() comdat { // CHECK-NEXT: resolver_entry: // CHECK-NEXT: [[TMP0:%.*]] = call i1 @__riscv_ifunc_select(ptr @__riscv_hwprobe_args, i32 2) ... // CHECK-LABEL: define weak_odr ptr @foo2.resolver() comdat { // CHECK-NEXT: resolver_entry: // CHECK-NEXT: [[TMP0:%.*]] = call i1 @__riscv_ifunc_select(ptr @__riscv_hwprobe_args.1, i32 2) // CHECK-NEXT: br i1 [[TMP0]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]] ...

The "arch=+zbb,+zba,+zbc,+zbkb,+zknd,+zknh,+zksed,+zvksh,+zfh,+zfa,+v" version has more features than the "arch=rv64im" version, but only the pair value is updated during __riscv_ifunc_select invocation.

But we need to increase limit when hwprobe support more key for new feature.

Then I think the MAX_PAIR_LENGTH way is OK to me.

BeMg · 2024-04-22T04:30:54Z

Since this resolver function is expected to be available and interchangeable for both libgcc and compiler-rt, a formal specification for the resolver function interface is necessary.

I've create one for this PR riscv-non-isa/riscv-c-api-doc#74 and provide the three different candidate approach to achieve the same purpose.

BeMg mentioned this pull request Mar 21, 2024

Function multi-version proposal riscv-non-isa/riscv-c-api-doc#48

Open

lukel97 mentioned this pull request Mar 21, 2024

[RISCV][FMV] Support target_clones #85786

Draft

wangpc-pp reviewed Mar 21, 2024

View reviewed changes

Use bitset instead of string

60848d3

BeMg marked this pull request as ready for review March 30, 2024 12:13

BeMg requested a review from kito-cheng March 30, 2024 12:14

llvmbot added compiler-rt compiler-rt:builtins labels Mar 30, 2024

BeMg requested review from topperc, lukel97 and preames March 30, 2024 12:14

lukel97 reviewed Apr 1, 2024

View reviewed changes

compiler-rt/lib/builtins/riscv/ifunc_select.c Outdated Show resolved Hide resolved

compiler-rt/lib/builtins/riscv/ifunc_select.c Outdated Show resolved Hide resolved

BeMg added 4 commits March 31, 2024 19:20

Fixup comment

6022ee3

Make SYS_riscv_hwprobe closer to use

368789d

Update hwprobe relate comment

8b2aa7d

Pass NULL-like form for cpu_set_t *cpus

7e1dc72

BeMg requested review from lukel97 and wangpc-pp April 1, 2024 02:35

Fix format issue

8223580

BeMg added 4 commits April 3, 2024 03:13

[compiler-rt] Update __riscv_ifunc_select prototype

4a3f4b6

[NFC] Rename ReqirePreKey to ReqireKeys

305af7d

[NFC] Rename Pairs to HwprobePairs

f39b2cc

Move sys_hwprobe define key into caller site

a0ad273

Fix format issue

b790093

topperc reviewed Apr 9, 2024

View reviewed changes

BeMg added 3 commits April 8, 2024 23:55

Use Length as system call argument

bf7ca4c

Rename ReqireKeys to RequireKeys

e5a2d6c

Use Length to declare HwprobePairs

9b06f1b

wangpc-pp reviewed Apr 9, 2024

View reviewed changes

BeMg mentioned this pull request Apr 9, 2024

[RFC] Function multiversion resolver function implementation riscv-non-isa/riscv-c-api-doc#72

Open

BeMg requested review from wangpc-pp and topperc April 15, 2024 06:51

BeMg mentioned this pull request Apr 19, 2024

[FMV] Runtime Resolver Function riscv-non-isa/riscv-c-api-doc#74

Open

BeMg marked this pull request as draft May 23, 2024 05:34

wangpc-pp mentioned this pull request Jun 4, 2024

[RISCV] Add support for getHostCPUFeatures using hwprobe #94352

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[compiler-rt][RISCV] Implement __riscv_ifunc_select #85790

[compiler-rt][RISCV] Implement __riscv_ifunc_select #85790

BeMg commented Mar 19, 2024 •

edited

BeMg commented Mar 19, 2024

wangpc-pp Mar 21, 2024

BeMg Mar 26, 2024

BeMg commented Mar 30, 2024

github-actions bot commented Apr 1, 2024 •

edited

wangpc-pp commented Apr 1, 2024

BeMg commented Apr 8, 2024

topperc Apr 9, 2024

BeMg Apr 9, 2024

topperc Apr 9, 2024

BeMg Apr 9, 2024

topperc Apr 9, 2024

BeMg Apr 9, 2024

wangpc-pp Apr 9, 2024

topperc Apr 9, 2024

wangpc-pp Apr 10, 2024

kito-cheng Apr 10, 2024

BeMg Apr 11, 2024

wangpc-pp Apr 15, 2024

BeMg Apr 15, 2024

wangpc-pp Apr 16, 2024

BeMg commented Apr 22, 2024


		#endif // defined(__linux__)

		unsigned __riscv_ifunc_select(struct riscv_hwprobe *ReqireKeys,

[compiler-rt][RISCV] Implement __riscv_ifunc_select #85790

Are you sure you want to change the base?

[compiler-rt][RISCV] Implement __riscv_ifunc_select #85790

Conversation

BeMg commented Mar 19, 2024 • edited

BeMg commented Mar 19, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

BeMg commented Mar 30, 2024

github-actions bot commented Apr 1, 2024 • edited

wangpc-pp commented Apr 1, 2024

BeMg commented Apr 8, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

BeMg commented Apr 22, 2024

BeMg commented Mar 19, 2024 •

edited

github-actions bot commented Apr 1, 2024 •

edited