Skip to content

[RISCV] Fix vp-intrinsics args in cost model tests. NFC #112463

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 2 commits into from
Oct 16, 2024
Merged

[RISCV] Fix vp-intrinsics args in cost model tests. NFC #112463

merged 2 commits into from
Oct 16, 2024

Conversation

@ElvisWang123
Copy link
Contributor

@ElvisWang123 ElvisWang123 commented Oct 16, 2024 •
edited
Loading

This patch contains following changes to fix vp intrinsics tests.

  1. v*float -> v*f32, v*double -> v*f64 and v*half -> v*f16
  2. Fix the order of the vp-intrinsics.

@ElvisWang123
[RISCV] Correct vp intrinsics args in tests. NFC
7082dcb 
This patch contains following changes to fix vp intrinsics tests.
1. v*float -> v*f32, v*double -> v*f64 and v*half -> v*f16
2. llvm.vp.s{u}itofp.v*i*.v*f* -> llvm.vp.s{u}itofp.v*f*.i*
3. llvm.vp.fptos{u}i.v*f*.v*i* -> llvm.vp.fptos{u}i.v*i*.v*f*
@llvmbot llvmbot added the llvm:analysis Includes value tracking, cost tables and constant folding label Oct 16, 2024
@llvmbot
Copy link
Member

llvmbot commented Oct 16, 2024

@llvm/pr-subscribers-llvm-analysis

Author: Elvis Wang (ElvisWang123)

Changes

This patch contains following changes to fix vp intrinsics tests.

  1. v*float -> v*f32, v*double -> v*f64 and v*half -> v*f16
  2. llvm.vp.s{u}itofp.v*\i*.v*f* -> llvm.vp.s{u}itofp.v*f*.i*
  3. llvm.vp.fptos{u}i.v*f*.v*i* -> llvm.vp.fptos{u}i.v*i*.v*f*

Patch is 201.38 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/112463.diff

2 Files Affected:

  • (modified) llvm/test/Analysis/CostModel/RISCV/cast.ll (+644-644)
  • (modified) llvm/test/Analysis/CostModel/RISCV/cmp.ll (+32-32)
diff --git a/llvm/test/Analysis/CostModel/RISCV/cast.ll b/llvm/test/Analysis/CostModel/RISCV/cast.ll
index 853eef6bcb2e87..44367c7031500b 100644
--- a/llvm/test/Analysis/CostModel/RISCV/cast.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/cast.ll
@@ -3071,113 +3071,113 @@ define void @fpext() {
   %v2f16_v2f64 = fpext <2 x half> undef to <2 x double>
   %v2f32_v2f64 = fpext <2 x float> undef to <2 x double>
 
-  %vp_v2f16_v2f32 = call <2 x float> @llvm.vp.fpext.v2half.v2float(<2 x half> undef, <2 x i1> undef, i32 undef)
-  %vp_v2f16_v2f64 = call <2 x double> @llvm.vp.fpext.v2half.v2double(<2 x half> undef, <2 x i1> undef, i32 undef)
-  %vp_v2f32_v2f64 = call <2 x double> @llvm.vp.fpext.v2float.v2double(<2 x float> undef, <2 x i1> undef, i32 undef)
+  %vp_v2f16_v2f32 = call <2 x float> @llvm.vp.fpext.v2f16.v2f32(<2 x half> undef, <2 x i1> undef, i32 undef)
+  %vp_v2f16_v2f64 = call <2 x double> @llvm.vp.fpext.v2f16.v2f64(<2 x half> undef, <2 x i1> undef, i32 undef)
+  %vp_v2f32_v2f64 = call <2 x double> @llvm.vp.fpext.v2f32.v2f64(<2 x float> undef, <2 x i1> undef, i32 undef)
 
   %v4f16_v4f32 = fpext <4 x half> undef to <4 x float>
   %v4f16_v4f64 = fpext <4 x half> undef to <4 x double>
   %v4f32_v4f64 = fpext <4 x float> undef to <4 x double>
 
-  %vp_v4f16_v4f32 = call <4 x float> @llvm.vp.fpext.v4half.v4float(<4 x half> undef, <4 x i1> undef, i32 undef)
-  %vp_v4f16_v4f64 = call <4 x double> @llvm.vp.fpext.v4half.v4double(<4 x half> undef, <4 x i1> undef, i32 undef)
-  %vp_v4f32_v4f64 = call <4 x double> @llvm.vp.fpext.v4float.v4double(<4 x float> undef, <4 x i1> undef, i32 undef)
+  %vp_v4f16_v4f32 = call <4 x float> @llvm.vp.fpext.v4f16.v4f32(<4 x half> undef, <4 x i1> undef, i32 undef)
+  %vp_v4f16_v4f64 = call <4 x double> @llvm.vp.fpext.v4f16.v4f64(<4 x half> undef, <4 x i1> undef, i32 undef)
+  %vp_v4f32_v4f64 = call <4 x double> @llvm.vp.fpext.v4f32.v4f64(<4 x float> undef, <4 x i1> undef, i32 undef)
 
   %v8f16_v8f32 = fpext <8 x half> undef to <8 x float>
   %v8f16_v8f64 = fpext <8 x half> undef to <8 x double>
   %v8f32_v8f64 = fpext <8 x float> undef to <8 x double>
 
-  %vp_v8f16_v8f32 = call <8 x float> @llvm.vp.fpext.v8half.v8float(<8 x half> undef, <8 x i1> undef, i32 undef)
-  %vp_v8f16_v8f64 = call <8 x double> @llvm.vp.fpext.v8half.v8double(<8 x half> undef, <8 x i1> undef, i32 undef)
-  %vp_v8f32_v8f64 = call <8 x double> @llvm.vp.fpext.v8float.v8double(<8 x float> undef, <8 x i1> undef, i32 undef)
+  %vp_v8f16_v8f32 = call <8 x float> @llvm.vp.fpext.v8f16.v8f32(<8 x half> undef, <8 x i1> undef, i32 undef)
+  %vp_v8f16_v8f64 = call <8 x double> @llvm.vp.fpext.v8f16.v8f64(<8 x half> undef, <8 x i1> undef, i32 undef)
+  %vp_v8f32_v8f64 = call <8 x double> @llvm.vp.fpext.v8f32.v8f64(<8 x float> undef, <8 x i1> undef, i32 undef)
 
   %v16f16_v16f32 = fpext <16 x half> undef to <16 x float>
   %v16f16_v16f64 = fpext <16 x half> undef to <16 x double>
   %v16f32_v16f64 = fpext <16 x float> undef to <16 x double>
 
-  %vp_v16f16_v16f32 = call <16 x float> @llvm.vp.fpext.v16half.v16float(<16 x half> undef, <16 x i1> undef, i32 undef)
-  %vp_v16f16_v16f64 = call <16 x double> @llvm.vp.fpext.v16half.v16double(<16 x half> undef, <16 x i1> undef, i32 undef)
-  %vp_v16f32_v16f64 = call <16 x double> @llvm.vp.fpext.v16float.v16double(<16 x float> undef, <16 x i1> undef, i32 undef)
+  %vp_v16f16_v16f32 = call <16 x float> @llvm.vp.fpext.v16f16.v16f32(<16 x half> undef, <16 x i1> undef, i32 undef)
+  %vp_v16f16_v16f64 = call <16 x double> @llvm.vp.fpext.v16f16.v16f64(<16 x half> undef, <16 x i1> undef, i32 undef)
+  %vp_v16f32_v16f64 = call <16 x double> @llvm.vp.fpext.v16f32.v16f64(<16 x float> undef, <16 x i1> undef, i32 undef)
 
   %v32f16_v32f32 = fpext <32 x half> undef to <32 x float>
   %v32f16_v32f64 = fpext <32 x half> undef to <32 x double>
   %v32f32_v32f64 = fpext <32 x float> undef to <32 x double>
 
-  %vp_v32f16_v32f32 = call <32 x float> @llvm.vp.fpext.v32half.v32float(<32 x half> undef, <32 x i1> undef, i32 undef)
-  %vp_v32f16_v32f64 = call <32 x double> @llvm.vp.fpext.v32half.v32double(<32 x half> undef, <32 x i1> undef, i32 undef)
-  %vp_v32f32_v32f64 = call <32 x double> @llvm.vp.fpext.v32float.v32double(<32 x float> undef, <32 x i1> undef, i32 undef)
+  %vp_v32f16_v32f32 = call <32 x float> @llvm.vp.fpext.v32f16.v32f32(<32 x half> undef, <32 x i1> undef, i32 undef)
+  %vp_v32f16_v32f64 = call <32 x double> @llvm.vp.fpext.v32f16.v32f64(<32 x half> undef, <32 x i1> undef, i32 undef)
+  %vp_v32f32_v32f64 = call <32 x double> @llvm.vp.fpext.v32f32.v32f64(<32 x float> undef, <32 x i1> undef, i32 undef)
 
   %v64f16_v64f32 = fpext <64 x half> undef to <64 x float>
   %v64f16_v64f64 = fpext <64 x half> undef to <64 x double>
   %v64f32_v64f64 = fpext <64 x float> undef to <64 x double>
 
-  %vp_v64f16_v64f32 = call <64 x float> @llvm.vp.fpext.v64half.v64float(<64 x half> undef, <64 x i1> undef, i32 undef)
-  %vp_v64f16_v64f64 = call <64 x double> @llvm.vp.fpext.v64half.v64double(<64 x half> undef, <64 x i1> undef, i32 undef)
-  %vp_v64f32_v64f64 = call <64 x double> @llvm.vp.fpext.v64float.v64double(<64 x float> undef, <64 x i1> undef, i32 undef)
+  %vp_v64f16_v64f32 = call <64 x float> @llvm.vp.fpext.v64f16.v64f32(<64 x half> undef, <64 x i1> undef, i32 undef)
+  %vp_v64f16_v64f64 = call <64 x double> @llvm.vp.fpext.v64f16.v64f64(<64 x half> undef, <64 x i1> undef, i32 undef)
+  %vp_v64f32_v64f64 = call <64 x double> @llvm.vp.fpext.v64f32.v64f64(<64 x float> undef, <64 x i1> undef, i32 undef)
 
   %v128f16_v128f32 = fpext <128 x half> undef to <128 x float>
   %v128f16_v128f64 = fpext <128 x half> undef to <128 x double>
   %v128f32_v128f64 = fpext <128 x float> undef to <128 x double>
 
-  %vp_v128f16_v128f32 = call <128 x float> @llvm.vp.fpext.v128half.v128float(<128 x half> undef, <128 x i1> undef, i32 undef)
-  %vp_v128f16_v128f64 = call <128 x double> @llvm.vp.fpext.v128half.v128double(<128 x half> undef, <128 x i1> undef, i32 undef)
-  %vp_v128f32_v128f64 = call <128 x double> @llvm.vp.fpext.v128float.v128double(<128 x float> undef, <128 x i1> undef, i32 undef)
+  %vp_v128f16_v128f32 = call <128 x float> @llvm.vp.fpext.v128f16.v128f32(<128 x half> undef, <128 x i1> undef, i32 undef)
+  %vp_v128f16_v128f64 = call <128 x double> @llvm.vp.fpext.v128f16.v128f64(<128 x half> undef, <128 x i1> undef, i32 undef)
+  %vp_v128f32_v128f64 = call <128 x double> @llvm.vp.fpext.v128f32.v128f64(<128 x float> undef, <128 x i1> undef, i32 undef)
 
   %nxv1f16_nxv1f32 = fpext <vscale x 1 x half> undef to <vscale x 1 x float>
   %nxv1f16_nxv1f64 = fpext <vscale x 1 x half> undef to <vscale x 1 x double>
   %nxv1f32_nxv1f64 = fpext <vscale x 1 x float> undef to <vscale x 1 x double>
 
-  %vp_nxv1f16_nxv1f32 = call <vscale x 1 x float> @llvm.vp.fpext.nxv1half.nxv1float(<vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-  %vp_nxv1f16_nxv1f64 = call <vscale x 1 x double> @llvm.vp.fpext.nxv1half.nxv1double(<vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-  %vp_nxv1f32_nxv1f64 = call <vscale x 1 x double> @llvm.vp.fpext.nxv1float.nxv1double(<vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
+  %vp_nxv1f16_nxv1f32 = call <vscale x 1 x float> @llvm.vp.fpext.nxv1f16.nxv1f32(<vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
+  %vp_nxv1f16_nxv1f64 = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f16.nxv1f64(<vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
+  %vp_nxv1f32_nxv1f64 = call <vscale x 1 x double> @llvm.vp.fpext.nxv1f32.nxv1f64(<vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
 
   %nxv2f16_nxv2f32 = fpext <vscale x 2 x half> undef to <vscale x 2 x float>
   %nxv2f16_nxv2f64 = fpext <vscale x 2 x half> undef to <vscale x 2 x double>
   %nxv2f32_nxv2f64 = fpext <vscale x 2 x float> undef to <vscale x 2 x double>
 
-  %vp_nxv2f16_nxv2f32 = call <vscale x 2 x float> @llvm.vp.fpext.nxv2half.nxv2float(<vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-  %vp_nxv2f16_nxv2f64 = call <vscale x 2 x double> @llvm.vp.fpext.nxv2half.nxv2double(<vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-  %vp_nxv2f32_nxv2f64 = call <vscale x 2 x double> @llvm.vp.fpext.nxv2float.nxv2double(<vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
+  %vp_nxv2f16_nxv2f32 = call <vscale x 2 x float> @llvm.vp.fpext.nxv2f16.nxv2f32(<vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
+  %vp_nxv2f16_nxv2f64 = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f16.nxv2f64(<vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
+  %vp_nxv2f32_nxv2f64 = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f32.nxv2f64(<vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
 
   %nxv4f16_nxv4f32 = fpext <vscale x 4 x half> undef to <vscale x 4 x float>
   %nxv4f16_nxv4f64 = fpext <vscale x 4 x half> undef to <vscale x 4 x double>
   %nxv4f32_nxv4f64 = fpext <vscale x 4 x float> undef to <vscale x 4 x double>
 
-  %vp_nxv4f16_nxv4f32 = call <vscale x 4 x float> @llvm.vp.fpext.nxv4half.nxv4float(<vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-  %vp_nxv4f16_nxv4f64 = call <vscale x 4 x double> @llvm.vp.fpext.nxv4half.nxv4double(<vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-  %vp_nxv4f32_nxv4f64 = call <vscale x 4 x double> @llvm.vp.fpext.nxv4float.nxv4double(<vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
+  %vp_nxv4f16_nxv4f32 = call <vscale x 4 x float> @llvm.vp.fpext.nxv4f16.nxv4f32(<vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
+  %vp_nxv4f16_nxv4f64 = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f16.nxv4f64(<vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
+  %vp_nxv4f32_nxv4f64 = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f32.nxv4f64(<vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
 
   %nxv8f16_nxv8f32 = fpext <vscale x 8 x half> undef to <vscale x 8 x float>
   %nxv8f16_nxv8f64 = fpext <vscale x 8 x half> undef to <vscale x 8 x double>
   %nxv8f32_nxv8f64 = fpext <vscale x 8 x float> undef to <vscale x 8 x double>
 
-  %vp_nxv8f16_nxv8f32 = call <vscale x 8 x float> @llvm.vp.fpext.nxv8half.nxv8float(<vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-  %vp_nxv8f16_nxv8f64 = call <vscale x 8 x double> @llvm.vp.fpext.nxv8half.nxv8double(<vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-  %vp_nxv8f32_nxv8f64 = call <vscale x 8 x double> @llvm.vp.fpext.nxv8float.nxv8double(<vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
+  %vp_nxv8f16_nxv8f32 = call <vscale x 8 x float> @llvm.vp.fpext.nxv8f16.nxv8f32(<vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
+  %vp_nxv8f16_nxv8f64 = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f16.nxv8f64(<vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
+  %vp_nxv8f32_nxv8f64 = call <vscale x 8 x double> @llvm.vp.fpext.nxv8f32.nxv8f64(<vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
 
   %nxv16f16_nxv16f32 = fpext <vscale x 16 x half> undef to <vscale x 16 x float>
   %nxv16f16_nxv16f64 = fpext <vscale x 16 x half> undef to <vscale x 16 x double>
   %nxv16f32_nxv16f64 = fpext <vscale x 16 x float> undef to <vscale x 16 x double>
 
-  %vp_nxv16f16_nxv16f32 = call <vscale x 16 x float> @llvm.vp.fpext.nxv16half.nxv16float(<vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
-  %vp_nxv16f16_nxv16f64 = call <vscale x 16 x double> @llvm.vp.fpext.nxv16half.nxv16double(<vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
-  %vp_nxv16f32_nxv16f64 = call <vscale x 16 x double> @llvm.vp.fpext.nxv16float.nxv16double(<vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
+  %vp_nxv16f16_nxv16f32 = call <vscale x 16 x float> @llvm.vp.fpext.nxv16f16.nxv16f32(<vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+  %vp_nxv16f16_nxv16f64 = call <vscale x 16 x double> @llvm.vp.fpext.nxv16f16.nxv16f64(<vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+  %vp_nxv16f32_nxv16f64 = call <vscale x 16 x double> @llvm.vp.fpext.nxv16f32.nxv16f64(<vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
 
   %nxv32f16_nxv32f32 = fpext <vscale x 32 x half> undef to <vscale x 32 x float>
   %nxv32f16_nxv32f64 = fpext <vscale x 32 x half> undef to <vscale x 32 x double>
   %nxv32f32_nxv32f64 = fpext <vscale x 32 x float> undef to <vscale x 32 x double>
 
-  %vp_nxv32f16_nxv32f32 = call <vscale x 32 x float> @llvm.vp.fpext.nxv32half.nxv32float(<vscale x 32 x half> undef, <vscale x 32 x i1> undef, i32 undef)
-  %vp_nxv32f16_nxv32f64 = call <vscale x 32 x double> @llvm.vp.fpext.nxv32half.nxv32double(<vscale x 32 x half> undef, <vscale x 32 x i1> undef, i32 undef)
-  %vp_nxv32f32_nxv32f64 = call <vscale x 32 x double> @llvm.vp.fpext.nxv32float.nxv32double(<vscale x 32 x float> undef, <vscale x 32 x i1> undef, i32 undef)
+  %vp_nxv32f16_nxv32f32 = call <vscale x 32 x float> @llvm.vp.fpext.nxv32f16.nxv32f32(<vscale x 32 x half> undef, <vscale x 32 x i1> undef, i32 undef)
+  %vp_nxv32f16_nxv32f64 = call <vscale x 32 x double> @llvm.vp.fpext.nxv32f16.nxv32f64(<vscale x 32 x half> undef, <vscale x 32 x i1> undef, i32 undef)
+  %vp_nxv32f32_nxv32f64 = call <vscale x 32 x double> @llvm.vp.fpext.nxv32f32.nxv32f64(<vscale x 32 x float> undef, <vscale x 32 x i1> undef, i32 undef)
 
   %nxv64f16_nxv64f32 = fpext <vscale x 64 x half> undef to <vscale x 64 x float>
   %nxv64f16_nxv64f64 = fpext <vscale x 64 x half> undef to <vscale x 64 x double>
   %nxv64f32_nxv64f64 = fpext <vscale x 64 x float> undef to <vscale x 64 x double>
 
-  %vp_nxv64f16_nxv64f32 = call <vscale x 64 x float> @llvm.vp.fpext.nxv64half.nxv64float(<vscale x 64 x half> undef, <vscale x 64 x i1> undef, i32 undef)
-  %vp_nxv64f16_nxv64f64 = call <vscale x 64 x double> @llvm.vp.fpext.nxv64half.nxv64double(<vscale x 64 x half> undef, <vscale x 64 x i1> undef, i32 undef)
-  %vp_nxv64f32_nxv64f64 = call <vscale x 64 x double> @llvm.vp.fpext.nxv64float.nxv64double(<vscale x 64 x float> undef, <vscale x 64 x i1> undef, i32 undef)
+  %vp_nxv64f16_nxv64f32 = call <vscale x 64 x float> @llvm.vp.fpext.nxv64f16.nxv64f32(<vscale x 64 x half> undef, <vscale x 64 x i1> undef, i32 undef)
+  %vp_nxv64f16_nxv64f64 = call <vscale x 64 x double> @llvm.vp.fpext.nxv64f16.nxv64f64(<vscale x 64 x half> undef, <vscale x 64 x i1> undef, i32 undef)
+  %vp_nxv64f32_nxv64f64 = call <vscale x 64 x double> @llvm.vp.fpext.nxv64f32.nxv64f64(<vscale x 64 x float> undef, <vscale x 64 x i1> undef, i32 undef)
 
   ret void
 }
@@ -3274,113 +3274,113 @@ define void @fptrunc() {
   %v2f64_v2f16 = fptrunc <2 x double> undef to <2 x half>
   %v2f64_v2f32 = fptrunc <2 x double> undef to <2 x float>
 
-  %vp_v2f32_v2f16 = call <2 x half> @llvm.vp.fptrunc.v2float.v2half(<2 x float> undef, <2 x i1> undef, i32 undef)
-  %vp_v2f64_v2f16 = call <2 x half> @llvm.vp.fptrunc.v2double.v2half(<2 x double> undef, <2 x i1> undef, i32 undef)
-  %vp_v2f64_v2f32 = call <2 x float> @llvm.vp.fptrunc.v2double.v2float(<2 x double> undef, <2 x i1> undef, i32 undef)
+  %vp_v2f32_v2f16 = call <2 x half> @llvm.vp.fptrunc.v2f32.v2f16(<2 x float> undef, <2 x i1> undef, i32 undef)
+  %vp_v2f64_v2f16 = call <2 x half> @llvm.vp.fptrunc.v2f64.v2f16(<2 x double> undef, <2 x i1> undef, i32 undef)
+  %vp_v2f64_v2f32 = call <2 x float> @llvm.vp.fptrunc.v2f64.v2f32(<2 x double> undef, <2 x i1> undef, i32 undef)
 
   %v4f32_v4f16 = fptrunc <4 x float> undef to <4 x half>
   %v4f64_v4f16 = fptrunc <4 x double> undef to <4 x half>
   %v4f64_v4f32 = fptrunc <4 x double> undef to <4 x float>
 
-  %vp_v4f32_v4f16 = call <4 x half> @llvm.vp.fptrunc.v4float.v4half(<4 x float> undef, <4 x i1> undef, i32 undef)
-  %vp_v4f64_v4f16 = call <4 x half> @llvm.vp.fptrunc.v4double.v4half(<4 x double> undef, <4 x i1> undef, i32 undef)
-  %vp_v4f64_v4f32 = call <4 x float> @llvm.vp.fptrunc.v4double.v4float(<4 x double> undef, <4 x i1> undef, i32 undef)
+  %vp_v4f32_v4f16 = call <4 x half> @llvm.vp.fptrunc.v4f32.v4f16(<4 x float> undef, <4 x i1> undef, i32 undef)
+  %vp_v4f64_v4f16 = call <4 x half> @llvm.vp.fptrunc.v4f64.v4f16(<4 x double> undef, <4 x i1> undef, i32 undef)
+  %vp_v4f64_v4f32 = call <4 x float> @llvm.vp.fptrunc.v4f64.v4f32(<4 x double> undef, <4 x i1> undef, i32 undef)
 
   %v8f32_v8f16 = fptrunc <8 x float> undef to <8 x half>
   %v8f64_v8f16 = fptrunc <8 x double> undef to <8 x half>
   %v8f64_v8f32 = fptrunc <8 x double> undef to <8 x float>
 
-  %vp_v8f32_v8f16 = call <8 x half> @llvm.vp.fptrunc.v8float.v8half(<8 x float> undef, <8 x i1> undef, i32 undef)
-  %vp_v8f64_v8f16 = call <8 x half> @llvm.vp.fptrunc.v8double.v8half(<8 x double> undef, <8 x i1> undef, i32 undef)
-  %vp_v8f64_v8f32 = call <8 x float> @llvm.vp.fptrunc.v8double.v8float(<8 x double> undef, <8 x i1> undef, i32 undef)
+  %vp_v8f32_v8f16 = call <8 x half> @llvm.vp.fptrunc.v8f32.v8f16(<8 x float> undef, <8 x i1> undef, i32 undef)
+  %vp_v8f64_v8f16 = call <8 x half> @llvm.vp.fptrunc.v8f64.v8f16(<8 x double> undef, <8 x i1> undef, i32 undef)
+  %vp_v8f64_v8f32 = call <8 x float> @llvm.vp.fptrunc.v8f64.v8f32(<8 x double> undef, <8 x i1> undef, i32 undef)
 
   %v16f32_v16f16 = fptrunc <16 x float> undef to <16 x half>
   %v16f64_v16f16 = fptrunc <16 x double> undef to <16 x half>
   %v16f64_v16f32 = fptrunc <16 x double> undef to <16 x float>
 
-  %vp_v16f32_v16f16 = call <16 x half> @llvm.vp.fptrunc.v16float.v16half(<16 x float> undef, <16 x i1> undef, i32 undef)
-  %vp_v16f64_v16f16 = call <16 x half> @llvm.vp.fptrunc.v16double.v16half(<16 x double> undef, <16 x i1> undef, i32 undef)
-  %vp_v16f64_v16f32 = call <16 x float> @llvm.vp.fptrunc.v16double.v16float(<16 x double> undef, <16 x i1> undef, i32 undef)
+  %vp_v16f32_v16f16 = call <16 x half> @llvm.vp.fptrunc.v16f32.v16f16(<16 x float> undef, <16 x i1> undef, i32 undef)
+  %vp_v16f64_v16f16 = call <16 x half> @llvm.vp.fptrunc.v16f64.v16f16(<16 x double> undef, <16 x i1> undef, i32 undef)
+  %vp_v16f64_v16f32 = call <16 x float> @llvm.vp.fptrunc.v16f64.v16f32(<16 x double> undef, <16 x i1> undef, i32 undef)
 
   %v32f32_v32f16 = fptrunc <32 x float> undef to <32 x half>
   %v32f64_v32f16 = fptrunc <32 x double> undef to <32 x half>
   %v32f64_v32f32 = fptrunc <32 x double> undef to <32 x float>
 
-  %vp_v32f32_v32f16 = call <32 x half> @llvm.vp.fptrunc.v32float.v32half(<32 x float> undef, <32 x i1> undef, i32 undef)
-  %vp_v32f64_v32f16 = call <32 x half> @llvm.vp.fptrunc.v32double.v32half(<32 x double> undef, <32 x i1> undef, i32 undef)
-  %vp_v32f64_v32f32 = call <32 x float> @llvm.vp.fptrunc.v32double.v32float(<32 x double> undef, <32 x i1> undef, i32 undef)
+  %vp_v32f32_v32f16 = call <32 x half> @llvm.vp.fptrunc.v32f32.v32f16(<32 x float> undef, <32 x i1> undef, i32 undef)
+  %vp_v32f64_v32f16 = call <32 x half> @llvm.vp.fptrunc.v32f64.v32f16(<32 x double> undef, <32 x i1> undef, i32 undef)
+  %vp_v32f64_v32f32 = call <32 x float> @llvm.vp.fptrunc.v32f64.v32f32(<32 x double> undef, <32 x i1> undef, i32 undef)
 
   %v64f32_v64f16 = fptrunc <64 x float> undef to <64 x half>
   %v64f64_v64f16 = fptrunc <64 x double> undef to <64 x half>
   %v64f64_v64f32 = fptrunc <64 x double> undef to <64 x float>
 
-  %vp_v64f32_v64f16 = call <64 x half> @llvm.vp.fptrunc.v64float.v64half(<64 x float> undef, <64 x i1> undef, i32 undef)
-  %vp_v64f64_v64f16 = call <64 x half> @llvm.vp.fptrunc.v64double.v64half(<64 x double> undef, <64 x i1> undef, i32 undef)
-  %vp_v64f64_v64f32 = call <64 x float> @llvm.vp.fptrunc.v64double.v64float(<64 x double> undef, <64 x i1> undef, i32 undef)
+  %vp_v64f32_v64f16 = call <64 x half> @llvm.vp.fptrunc.v64f32.v64f16(<64 x float> undef, <64 x i1> undef, i32 undef)
+  %vp_v64f64_v64f16 = call <64 x half> @llvm.vp.fptrunc.v64f64.v64f16(<64 x double> undef, <64 x i1> undef, i32 undef)
+  %vp_v64f64_v64f32 = call <64 x float> @llvm.vp.fptrunc.v64f64.v64f32(<64 x double> undef, <64 x i1> undef, i32 undef)
 
   %v128f32_v128f16 = fptrunc <128 x float> undef to <128 x half>
   %v128f64_v128f16 = fptrunc <128 x double> undef to <128 x half>
   %v128f64_v128f32 = fptrunc <128 x double> undef to <128 x float>
 
-  %vp_v128f32_v128f16 = call <128 x half> @llvm.vp.fptrunc.v128float.v128half(<128 x float> undef, <128 x i1> undef, i32 undef)
-  %vp_v128f64_v128f16 = call <128 x half> @llvm.vp.fptrunc.v128double.v128half(<128 x double> undef, <128 x i1> undef, i32 undef)
-  %vp_v128f64_v128f32 = call <128 x float> @llvm.vp.fptrunc.v128double.v128flo...
[truncated]

topperc
topperc reviewed Oct 16, 2024
View reviewed changes
llvm/test/Analysis/CostModel/RISCV/cast.ll Outdated
%vp_v2f16_v2f32 = call <2 x float> @llvm.vp.fpext.v2half.v2float(<2 x half> undef, <2 x i1> undef, i32 undef)
%vp_v2f16_v2f64 = call <2 x double> @llvm.vp.fpext.v2half.v2double(<2 x half> undef, <2 x i1> undef, i32 undef)
%vp_v2f32_v2f64 = call <2 x double> @llvm.vp.fpext.v2float.v2double(<2 x float> undef, <2 x i1> undef, i32 undef)
%vp_v2f16_v2f32 = call <2 x float> @llvm.vp.fpext.v2f16.v2f32(<2 x half> undef, <2 x i1> undef, i32 undef)
Copy link
Collaborator

@topperc topperc Oct 16, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Are these types in the right order? Isn't the return type usually first?

Copy link
Contributor Author

@ElvisWang123 ElvisWang123 Oct 16, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed, thanks.

topperc
topperc approved these changes Oct 16, 2024
View reviewed changes
Copy link
Collaborator

@topperc topperc left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@ElvisWang123 ElvisWang123 merged commit f364804 into llvm:main Oct 16, 2024
8 checks passed
@ElvisWang123 ElvisWang123 deleted the fix-vp-args branch October 16, 2024 04:57
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@topperc topperc topperc approved these changes

@lukel97 lukel97 Awaiting requested review from lukel97

@arcbbb arcbbb Awaiting requested review from arcbbb

Assignees

No one assigned

Labels

llvm:analysis Includes value tracking, cost tables and constant folding

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@ElvisWang123 @llvmbot @topperc