[llvm] d42f63b - [AMDGPU] Use non-compressed exports in a test. NFC.

Jay Foad via llvm-commits llvm-commits at lists.llvm.org
Tue Mar 23 04:18:19 PDT 2021


Author: Jay Foad
Date: 2021-03-23T11:18:12Z
New Revision: d42f63beebc59253d9d6001d0e43109330d0de8c

URL: https://github.com/llvm/llvm-project/commit/d42f63beebc59253d9d6001d0e43109330d0de8c
DIFF: https://github.com/llvm/llvm-project/commit/d42f63beebc59253d9d6001d0e43109330d0de8c.diff

LOG: [AMDGPU] Use non-compressed exports in a test. NFC.

I don't think there's any need for this test to use compressed exports.
Using normal exports seems a bit more straightforwards and avoids a tiny
bit of bitcasting.

Differential Revision: https://reviews.llvm.org/D99167

Added: 
    

Modified: 
    llvm/test/CodeGen/AMDGPU/skip-if-dead.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/AMDGPU/skip-if-dead.ll b/llvm/test/CodeGen/AMDGPU/skip-if-dead.ll
index 690fe5a7e683..bc84fe8adab6 100644
--- a/llvm/test/CodeGen/AMDGPU/skip-if-dead.ll
+++ b/llvm/test/CodeGen/AMDGPU/skip-if-dead.ll
@@ -1171,7 +1171,7 @@ define amdgpu_ps void @cbranch_kill(i32 inreg %0, float %val0, float %val1) {
 ; SI-NEXT:    v_mul_f32_e32 v2, v0, v1
 ; SI-NEXT:  ; %bb.5: ; %export
 ; SI-NEXT:    s_or_b64 exec, exec, s[0:1]
-; SI-NEXT:    exp mrt0 v2, v2, v2, v2 done compr vm
+; SI-NEXT:    exp mrt0 v2, v2, v2, v2 done vm
 ; SI-NEXT:    s_endpgm
 ; SI-NEXT:  BB14_6:
 ; SI-NEXT:    s_mov_b64 exec, 0
@@ -1208,7 +1208,7 @@ define amdgpu_ps void @cbranch_kill(i32 inreg %0, float %val0, float %val1) {
 ; GFX10-WAVE64-NEXT:    v_mul_f32_e32 v2, v0, v1
 ; GFX10-WAVE64-NEXT:  ; %bb.5: ; %export
 ; GFX10-WAVE64-NEXT:    s_or_b64 exec, exec, s[0:1]
-; GFX10-WAVE64-NEXT:    exp mrt0 v2, v2, v2, v2 done compr vm
+; GFX10-WAVE64-NEXT:    exp mrt0 v2, v2, v2, v2 done vm
 ; GFX10-WAVE64-NEXT:    s_endpgm
 ; GFX10-WAVE64-NEXT:  BB14_6:
 ; GFX10-WAVE64-NEXT:    s_mov_b64 exec, 0
@@ -1245,7 +1245,7 @@ define amdgpu_ps void @cbranch_kill(i32 inreg %0, float %val0, float %val1) {
 ; GFX10-WAVE32-NEXT:    v_mul_f32_e32 v2, v0, v1
 ; GFX10-WAVE32-NEXT:  ; %bb.5: ; %export
 ; GFX10-WAVE32-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX10-WAVE32-NEXT:    exp mrt0 v2, v2, v2, v2 done compr vm
+; GFX10-WAVE32-NEXT:    exp mrt0 v2, v2, v2, v2 done vm
 ; GFX10-WAVE32-NEXT:    s_endpgm
 ; GFX10-WAVE32-NEXT:  BB14_6:
 ; GFX10-WAVE32-NEXT:    s_mov_b32 exec_lo, 0
@@ -1266,8 +1266,7 @@ live:
 
 export:
   %proxy = phi float [ undef, %kill ], [ %scale, %live ]
-  %proxyh = bitcast float %proxy to <2 x half>
-  call void @llvm.amdgcn.exp.compr.v2f16(i32 immarg 0, i32 immarg 15, <2 x half> %proxyh, <2 x half> %proxyh, i1 immarg true, i1 immarg true) #3
+  call void @llvm.amdgcn.exp.f32(i32 immarg 0, i32 immarg 15, float %proxy, float %proxy, float %proxy, float %proxy, i1 immarg true, i1 immarg true) #3
   ret void
 }
 
@@ -1309,7 +1308,7 @@ define amdgpu_ps void @complex_loop(i32 inreg %cmpa, i32 %cmpb, i32 %cmpc) {
 ; SI-NEXT:  BB15_6: ; %Flow
 ; SI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; SI-NEXT:  BB15_7: ; %._crit_edge
-; SI-NEXT:    exp mrt0 v2, v2, v0, v0 done compr vm
+; SI-NEXT:    exp mrt0 v2, v2, v0, v0 done vm
 ; SI-NEXT:    s_endpgm
 ; SI-NEXT:  BB15_8:
 ; SI-NEXT:    s_mov_b64 exec, 0
@@ -1352,7 +1351,7 @@ define amdgpu_ps void @complex_loop(i32 inreg %cmpa, i32 %cmpb, i32 %cmpc) {
 ; GFX10-WAVE64-NEXT:  BB15_6: ; %Flow
 ; GFX10-WAVE64-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX10-WAVE64-NEXT:  BB15_7: ; %._crit_edge
-; GFX10-WAVE64-NEXT:    exp mrt0 v2, v2, v0, v0 done compr vm
+; GFX10-WAVE64-NEXT:    exp mrt0 v2, v2, v0, v0 done vm
 ; GFX10-WAVE64-NEXT:    s_endpgm
 ; GFX10-WAVE64-NEXT:  BB15_8:
 ; GFX10-WAVE64-NEXT:    s_mov_b64 exec, 0
@@ -1395,7 +1394,7 @@ define amdgpu_ps void @complex_loop(i32 inreg %cmpa, i32 %cmpb, i32 %cmpc) {
 ; GFX10-WAVE32-NEXT:  BB15_6: ; %Flow
 ; GFX10-WAVE32-NEXT:    s_or_b32 exec_lo, exec_lo, s0
 ; GFX10-WAVE32-NEXT:  BB15_7: ; %._crit_edge
-; GFX10-WAVE32-NEXT:    exp mrt0 v2, v2, v0, v0 done compr vm
+; GFX10-WAVE32-NEXT:    exp mrt0 v2, v2, v0, v0 done vm
 ; GFX10-WAVE32-NEXT:    s_endpgm
 ; GFX10-WAVE32-NEXT:  BB15_8:
 ; GFX10-WAVE32-NEXT:    s_mov_b32 exec_lo, 0
@@ -1424,8 +1423,8 @@ latch:
 
 ._crit_edge:
   %tmp = phi i32 [ -1, %.entry ], [ %ctr.next, %latch ]
-  %out = bitcast i32 %tmp to <2 x half>
-  call void @llvm.amdgcn.exp.compr.v2f16(i32 immarg 0, i32 immarg 15, <2 x half> %out, <2 x half> undef, i1 immarg true, i1 immarg true)
+  %out = bitcast i32 %tmp to float
+  call void @llvm.amdgcn.exp.f32(i32 immarg 0, i32 immarg 15, float %out, float %out, float undef, float undef, i1 immarg true, i1 immarg true)
   ret void
 }
 
@@ -1479,7 +1478,7 @@ bb.1:
   ret void
 }
 
-declare void @llvm.amdgcn.exp.compr.v2f16(i32 immarg, i32 immarg, <2 x half>, <2 x half>, i1 immarg, i1 immarg) #3
+declare void @llvm.amdgcn.exp.f32(i32 immarg, i32 immarg, float, float, float, float, i1 immarg, i1 immarg) #3
 declare float @llvm.amdgcn.image.sample.l.2darray.f32.f32(i32 immarg, float, float, float, float, <8 x i32>, <4 x i32>, i1 immarg, i32 immarg, i32 immarg) #1
 declare <4 x float> @llvm.amdgcn.image.sample.c.1d.v4f32.f32(i32, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
 declare void @llvm.amdgcn.kill(i1) #0


        


More information about the llvm-commits mailing list