[RISCV][VLOPT] Add mask load to isSupported and getOperandInfo

[michaelmaitland]

Add mask store to getOperandInfo since it has the same behavior.

[llvmbot]

@llvm/pr-subscribers-backend-risc-v

Author: Michael Maitland (michaelmaitland)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/122030.diff

2 Files Affected:

• (modified) llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp (+4)
• (modified) llvm/test/CodeGen/RISCV/rvv/vl-opt-op-info.mir (+30)

diff --git a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
index d7ac3afe7b76b2..2acdefcc6d505d 100644
--- a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
+++ b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
@@ -257,6 +257,9 @@ static OperandInfo getOperandInfo(const MachineOperand &MO,
   // Vector Unit-Stride Instructions
   // Vector Strided Instructions
   /// Dest EEW encoded in the instruction and EMUL=(EEW/SEW)*LMUL
+  case RISCV::VLM_V:
+  case RISCV::VSM_V:
+    return OperandInfo(RISCVVType::getEMULEqualsEEWDivSEWTimesLMUL(0, MI), 0);
   case RISCV::VLE8_V:
   case RISCV::VSE8_V:
   case RISCV::VLSE8_V:
@@ -742,6 +745,7 @@ static bool isSupportedInstr(const MachineInstr &MI) {
   switch (RVV->BaseInstr) {
   // Vector Unit-Stride Instructions
   // Vector Strided Instructions
+  case RISCV::VLM_V:
   case RISCV::VLE8_V:
   case RISCV::VLSE8_V:
   case RISCV::VLE16_V:
diff --git a/llvm/test/CodeGen/RISCV/rvv/vl-opt-op-info.mir b/llvm/test/CodeGen/RISCV/rvv/vl-opt-op-info.mir
index a1bbfc8a7d3514..689809f6201054 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vl-opt-op-info.mir
+++ b/llvm/test/CodeGen/RISCV/rvv/vl-opt-op-info.mir
@@ -603,6 +603,36 @@ body: |
     %x:vr = PseudoVADD_VV_MF2 $noreg, %x, $noreg, 1, 3 /* e8 */, 0
 ...
 ---
+name: vlm_v
+body: |
+  bb.0:
+    ; CHECK-LABEL: name: vlm_v
+    ; CHECK: %x:vr = PseudoVLM_V_B8 $noreg, $noreg, 1, 0 /* e8 */, 0 /* tu, mu */
+    ; CHECK-NEXT: %y:vr = PseudoVMAND_MM_B8 $noreg, %x, 1, 0 /* e8 */
+    %x:vr = PseudoVLM_V_B8 $noreg, $noreg, -1, 0, 0
+    %y:vr = PseudoVMAND_MM_B8 $noreg, %x, 1, 0
+...
+---
+name: vlm_v_incompatible_eew
+body: |
+  bb.0:
+    ; CHECK-LABEL: name: vlm_v_incompatible_eew
+    ; CHECK: %x:vr = PseudoVLM_V_B8 $noreg, $noreg, -1, 0 /* e8 */, 0 /* tu, mu */
+    ; CHECK-NEXT: %y:vr = PseudoVADD_VV_M1 $noreg, $noreg, %x, 1, 4 /* e16 */, 0 /* tu, mu */
+    %x:vr = PseudoVLM_V_B8 $noreg, $noreg, -1, 0, 0
+    %y:vr = PseudoVADD_VV_M1 $noreg, $noreg, %x, 1, 4 /* e16 */, 0
+...
+---
+name: vlm_v_incompatible_emul
+body: |
+  bb.0:
+    ; CHECK-LABEL: name: vlm_v_incompatible_emul
+    ; CHECK: %x:vr = PseudoVLM_V_B8 $noreg, $noreg, -1, 0 /* e8 */, 0 /* tu, mu */
+    ; CHECK-NEXT: %y:vr = PseudoVMAND_MM_B16 $noreg, %x, 1, 0 /* e8 */
+    %x:vr = PseudoVLM_V_B8 $noreg, $noreg, -1, 0, 0
+    %y:vr = PseudoVMAND_MM_B16 $noreg, %x, 1, 0
+...
+---
 name: vsseN_v
 body: |
   bb.0:

[topperc]

masked -> mask. "Masked load" would be load of a vector with some elements skipped by the mask. "Mask load" loads a memory location containing a mask.

[michaelmaitland]

masked -> mask. "Masked load" would be load of a vector with some elements skipped by the mask. "Mask load" loads a memory location containing a mask.

done

[topperc]

No tests for VSM?

[michaelmaitland]

Added, although I'm not sure that it makes sense to have an incompatible_eew test does it?

[topperc]

LGTM

[llvm-ci]

LLVM Buildbot has detected a new failure on builder openmp-offload-libc-amdgpu-runtime running on omp-vega20-1 while building llvm at step 7 "Add check check-offload".

Full details are available at: https://lab.llvm.org/buildbot/#/builders/73/builds/11336

Here is the relevant piece of the build log for the reference

Step 7 (Add check check-offload) failure: 1200 seconds without output running [b'ninja', b'-j 32', b'check-offload'], attempting to kill
...
PASS: libomptarget :: x86_64-unknown-linux-gnu-LTO :: offloading/bug53727.cpp (986 of 999)
PASS: libomptarget :: x86_64-unknown-linux-gnu-LTO :: offloading/bug47654.cpp (987 of 999)
PASS: libomptarget :: x86_64-unknown-linux-gnu-LTO :: offloading/bug50022.cpp (988 of 999)
PASS: libomptarget :: x86_64-unknown-linux-gnu-LTO :: offloading/test_libc.cpp (989 of 999)
PASS: libomptarget :: x86_64-unknown-linux-gnu-LTO :: offloading/wtime.c (990 of 999)
PASS: libomptarget :: x86_64-unknown-linux-gnu :: offloading/bug49021.cpp (991 of 999)
PASS: libomptarget :: x86_64-unknown-linux-gnu :: offloading/std_complex_arithmetic.cpp (992 of 999)
PASS: libomptarget :: x86_64-unknown-linux-gnu-LTO :: offloading/complex_reduction.cpp (993 of 999)
PASS: libomptarget :: x86_64-unknown-linux-gnu-LTO :: offloading/bug49021.cpp (994 of 999)
PASS: libomptarget :: x86_64-unknown-linux-gnu-LTO :: offloading/std_complex_arithmetic.cpp (995 of 999)
command timed out: 1200 seconds without output running [b'ninja', b'-j 32', b'check-offload'], attempting to kill
process killed by signal 9
program finished with exit code -1
elapsedTime=1236.299768