[RISCV] Update floating point load latency in SiFive7 scheduling model

[mshockwave]

The latency of floating point loads in SiFive7 should be the same as their integer counterparts.

[llvmbot]

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

Author: Min-Yih Hsu (mshockwave)

Changes

The latency of floating point loads in SiFive7 should be the same as their integer counterparts.

---

Full diff: https://github.com/llvm/llvm-project/pull/159462.diff

2 Files Affected:

• (modified) llvm/lib/Target/RISCV/RISCVSchedSiFive7.td (+8-10)
• (added) llvm/test/tools/llvm-mca/RISCV/SiFive7/scalar-load-store.s (+82)

diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index ddb6427ff3366..d1d5a27da40f0 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -362,16 +362,14 @@ multiclass SiFive7WriteResBase<int VLEN,
   def : WriteRes<WriteFST64, [PipeA]>;
 
   let Latency = 3 in {
-  def : WriteRes<WriteLDB, [PipeA]>;
-  def : WriteRes<WriteLDH, [PipeA]>;
-  def : WriteRes<WriteLDW, [PipeA]>;
-  def : WriteRes<WriteLDD, [PipeA]>;
-  }
-
-  let Latency = 2 in {
-  def : WriteRes<WriteFLD16, [PipeA]>;
-  def : WriteRes<WriteFLD32, [PipeA]>;
-  def : WriteRes<WriteFLD64, [PipeA]>;
+    def : WriteRes<WriteLDB, [PipeA]>;
+    def : WriteRes<WriteLDH, [PipeA]>;
+    def : WriteRes<WriteLDW, [PipeA]>;
+    def : WriteRes<WriteLDD, [PipeA]>;
+
+    def : WriteRes<WriteFLD16, [PipeA]>;
+    def : WriteRes<WriteFLD32, [PipeA]>;
+    def : WriteRes<WriteFLD64, [PipeA]>;
   }
 
   // Atomic memory
diff --git a/llvm/test/tools/llvm-mca/RISCV/SiFive7/scalar-load-store.s b/llvm/test/tools/llvm-mca/RISCV/SiFive7/scalar-load-store.s
new file mode 100644
index 0000000000000..01fe46244b558
--- /dev/null
+++ b/llvm/test/tools/llvm-mca/RISCV/SiFive7/scalar-load-store.s
@@ -0,0 +1,82 @@
+# NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py
+# RUN: llvm-mca -mtriple=riscv64 -mcpu=sifive-x280 -instruction-tables=full -iterations=1 < %s \
+# RUN:   | FileCheck %s
+
+lh t0, 0(sp)
+flh fa0, 0(sp)
+lw t2, 0(sp)
+flw fa2, 0(sp)
+ld t4, 0(sp)
+fld fa4, 0(sp)
+
+sh t1, 0(sp)
+fsh fa1, 0(sp)
+sw t3, 0(sp)
+fsw fa3, 0(sp)
+sd t5, 0(sp)
+fsd fa5, 0(sp)
+
+# CHECK:      Resources:
+# CHECK-NEXT: [0]   - VLEN512SiFive7FDiv:1
+# CHECK-NEXT: [1]   - VLEN512SiFive7IDiv:1
+# CHECK-NEXT: [2]   - VLEN512SiFive7PipeA:1
+# CHECK-NEXT: [3]   - VLEN512SiFive7PipeAB:2 VLEN512SiFive7PipeA, VLEN512SiFive7PipeB
+# CHECK-NEXT: [4]   - VLEN512SiFive7PipeB:1
+# CHECK-NEXT: [5]   - VLEN512SiFive7VA:1
+# CHECK-NEXT: [6]   - VLEN512SiFive7VCQ:1
+# CHECK-NEXT: [7]   - VLEN512SiFive7VL:1
+# CHECK-NEXT: [8]   - VLEN512SiFive7VS:1
+
+# CHECK:      Instruction Info:
+# CHECK-NEXT: [1]: #uOps
+# CHECK-NEXT: [2]: Latency
+# CHECK-NEXT: [3]: RThroughput
+# CHECK-NEXT: [4]: MayLoad
+# CHECK-NEXT: [5]: MayStore
+# CHECK-NEXT: [6]: HasSideEffects (U)
+# CHECK-NEXT: [7]: Bypass Latency
+# CHECK-NEXT: [8]: Resources (<Name> | <Name>[<ReleaseAtCycle>] | <Name>[<AcquireAtCycle>,<ReleaseAtCycle])
+# CHECK-NEXT: [9]: LLVM Opcode Name
+
+# CHECK:      [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]                                        [9]                        Instructions:
+# CHECK-NEXT:  1      3     1.00    *                    3     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   LH                         lh	t0, 0(sp)
+# CHECK-NEXT:  1      3     1.00    *                    3     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   FLH                        flh	fa0, 0(sp)
+# CHECK-NEXT:  1      3     1.00    *                    3     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   C_LWSP                     lw	t2, 0(sp)
+# CHECK-NEXT:  1      3     1.00    *                    3     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   FLW                        flw	fa2, 0(sp)
+# CHECK-NEXT:  1      3     1.00    *                    3     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   C_LDSP                     ld	t4, 0(sp)
+# CHECK-NEXT:  1      3     1.00    *                    3     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   C_FLDSP                    fld	fa4, 0(sp)
+# CHECK-NEXT:  1      1     1.00           *             1     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   SH                         sh	t1, 0(sp)
+# CHECK-NEXT:  1      1     1.00           *             1     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   FSH                        fsh	fa1, 0(sp)
+# CHECK-NEXT:  1      1     1.00           *             1     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   C_SWSP                     sw	t3, 0(sp)
+# CHECK-NEXT:  1      1     1.00           *             1     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   FSW                        fsw	fa3, 0(sp)
+# CHECK-NEXT:  1      1     1.00           *             1     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   C_SDSP                     sd	t5, 0(sp)
+# CHECK-NEXT:  1      1     1.00           *             1     VLEN512SiFive7PipeA,VLEN512SiFive7PipeAB   C_FSDSP                    fsd	fa5, 0(sp)
+
+# CHECK:      Resources:
+# CHECK-NEXT: [0]   - VLEN512SiFive7FDiv
+# CHECK-NEXT: [1]   - VLEN512SiFive7IDiv
+# CHECK-NEXT: [2]   - VLEN512SiFive7PipeA
+# CHECK-NEXT: [3]   - VLEN512SiFive7PipeB
+# CHECK-NEXT: [4]   - VLEN512SiFive7VA
+# CHECK-NEXT: [5]   - VLEN512SiFive7VCQ
+# CHECK-NEXT: [6]   - VLEN512SiFive7VL
+# CHECK-NEXT: [7]   - VLEN512SiFive7VS
+
+# CHECK:      Resource pressure per iteration:
+# CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]
+# CHECK-NEXT:  -      -     12.00   -      -      -      -      -
+
+# CHECK:      Resource pressure by instruction:
+# CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    Instructions:
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     lh	t0, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     flh	fa0, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     lw	t2, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     flw	fa2, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     ld	t4, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     fld	fa4, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     sh	t1, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     fsh	fa1, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     sw	t3, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     fsw	fa3, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     sd	t5, 0(sp)
+# CHECK-NEXT:  -      -     1.00    -      -      -      -      -     fsd	fa5, 0(sp)

[wangpc-pp]

LGTM.