[mlir][ArmSME][test] Make use of arm_sme.streaming_vl (NFC) #77322

MacDue · 2024-01-08T14:46:31Z

No description provided.

llvmbot · 2024-01-10T10:29:00Z

@llvm/pr-subscribers-mlir-linalg

@llvm/pr-subscribers-mlir-sme

Author: Benjamin Maxwell (MacDue)

Changes

Full diff: https://github.com/llvm/llvm-project/pull/77322.diff

7 Files Affected:

(modified) mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir (+1-4)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir (+1-3)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir (+1-8)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir (+1-8)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir (+1-3)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir (+2-8)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir (+1-4)

diff --git a/mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir b/mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir
index 81546de6a3466b..6314e6f279952b 100644
--- a/mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir
+++ b/mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir
@@ -20,12 +20,9 @@ func.func @entry() {
 
   %c123_f32 = arith.constant 123.0 : f32
 
-  %min_elts_s = arith.constant 4 : index
-  %vscale = vector.vscale
-
   // "svl" refers to the Streaming Vector Length and "svl_s" the number of
   // 32-bit elements in a vector of SVL bits.
-  %svl_s = arith.muli %min_elts_s, %vscale : index
+  %svl_s = arm_sme.streaming_vl <word>
 
   %tile_init = bufferization.alloc_tensor(%svl_s, %svl_s) : tensor<?x?xf32>
 
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir
index 936163d1cd30d9..064141c349241e 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir
@@ -17,9 +17,7 @@ func.func @entry() {
   %c1_i32 = arith.constant 1 : i32
 
   // Calculate the size of a 32-bit tile, e.g. ZA{n}.s.
-  %vscale = vector.vscale
-  %min_elts_s = arith.constant 4 : index
-  %svl_s = arith.muli %min_elts_s, %vscale : index
+  %svl_s = arm_sme.streaming_vl <word>
   %za_s_size = arith.muli %svl_s, %svl_s : index
 
   // Allocate memory.
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir
index 839aed2e840c90..6e028d5fb83614 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir
@@ -134,12 +134,6 @@ func.func @initialize_memory(%d0 : index, %d1 : index) -> memref<?x?xf32> {
   return %A : memref<?x?xf32>
 }
 
-// This will be made a streaming function by enable-arm-streaming so return SVL.
-func.func @get_svl() -> index {
-  %vscale = vector.vscale
-  return %vscale : index
-}
-
 func.func @entry() {
   %c0 = arith.constant 0 : index
   %c1 = arith.constant 1 : index
@@ -148,8 +142,7 @@ func.func @entry() {
 
   // Allocate enough memory to load a 32-bit tile plus a tiny bit more to test
   // non-zero offsets while remaining inbounds.
-  %svl = call @get_svl() : () -> index
-  %svl_s = arith.muli %c4, %svl : index
+  %svl_s = arm_sme.streaming_vl <word>
   %svl_s_plus_two = arith.addi %svl_s, %c2 : index
 
   %A = call @initialize_memory(%svl_s_plus_two, %svl_s_plus_two) : (index, index) -> memref<?x?xf32>
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir
index 84246606daa8af..cb30fee4e12d72 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir
@@ -96,12 +96,6 @@ func.func @initialize_memory(%d0 : index, %d1 : index) -> memref<?x?xf32> {
   return %A : memref<?x?xf32>
 }
 
-// This will be made a streaming function by enable-arm-streaming so return SVL.
-func.func @get_svl() -> index {
-  %vscale = vector.vscale
-  return %vscale : index
-}
-
 func.func @entry() {
   %c0 = arith.constant 0 : index
   %c2 = arith.constant 2 : index
@@ -111,8 +105,7 @@ func.func @entry() {
   //
   // Allocate enough memory to load a 32-bit tile plus a tiny bit more to test
   // non-zero offsets while remaining inbounds.
-  %svl = call @get_svl() : () -> index
-  %svl_s = arith.muli %c4, %svl : index
+  %svl_s = arm_sme.streaming_vl <word>
   %svl_s_plus_two = arith.addi %svl_s, %c2 : index
   %A = call @initialize_memory(%svl_s_plus_two, %svl_s_plus_two) : (index, index) -> memref<?x?xf32>
 
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir
index 2751c2d136485e..eee3c56351d81e 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir
@@ -17,9 +17,7 @@ func.func @entry() {
   %c1_i32 = arith.constant 1 : i32
 
   // Calculate the size of a 32-bit tile, e.g. ZA{n}.s.
-  %vscale = vector.vscale
-  %min_elts_s = arith.constant 4 : index
-  %svl_s = arith.muli %min_elts_s, %vscale : index
+  %svl_s = arm_sme.streaming_vl <word>
   %za_s_size = arith.muli %svl_s, %svl_s : index
 
   // Allocate memory.
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir
index 5d2d0a73992f1e..2f151e2ec72fb7 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir
@@ -24,12 +24,9 @@ func.func @za0_d_f64() -> i32 {
   %c1_f64 = arith.constant 1.0 : f64
   %c1_index = arith.constant 1 : index
 
-  %min_elts_d = arith.constant 2 : index
-  %vscale = vector.vscale
-
   // "svl" refers to the Streaming Vector Length and "svl_d" the number of
   // 64-bit elements in a vector of SVL bits.
-  %svl_d = arith.muli %min_elts_d, %vscale : index
+  %svl_d = arm_sme.streaming_vl <double>
 
   // Allocate "mem1" and fill each "row" with row number.
   //
@@ -170,13 +167,10 @@ func.func @load_store_two_za_s_tiles() -> i32 {
   %c1_index = arith.constant 1 : index
   %c2_index = arith.constant 2 : index
 
-  %min_elts_s = arith.constant 4 : index
-  %vscale = vector.vscale
-
   // "svl" refers to the Streaming Vector Length and "svl_s" can mean either:
   // * the number of 32-bit elements in a vector of SVL bits.
   // * the number of tile slices (1d vectors) in a 32-bit element tile.
-  %svl_s = arith.muli %min_elts_s, %vscale : index
+  %svl_s = arm_sme.streaming_vl <word>
 
   // Allocate memory for two 32-bit element tiles.
   %size_of_tile = arith.muli %svl_s, %svl_s : index
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir
index af5fe236e5bd57..073c08bff1c415 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir
@@ -14,12 +14,9 @@ func.func @entry() -> i32 {
   %c1_i8 = arith.constant 1 : i8
   %c1_index = arith.constant 1 : index
 
-  %c16 = arith.constant 16 : index
-  %vscale = vector.vscale
-
   // "svl" refers to the Streaming Vector Length and "svl_b" the number of
   // 8-bit elements in a vector of SVL bits.
-  %svl_b = arith.muli %c16, %vscale : index
+  %svl_b = arm_sme.streaming_vl <byte>
 
   // Allocate memory and fill with ones.
   //

llvmbot · 2024-01-10T10:29:01Z

@llvm/pr-subscribers-mlir-vector

Author: Benjamin Maxwell (MacDue)

Changes

Full diff: https://github.com/llvm/llvm-project/pull/77322.diff

7 Files Affected:

(modified) mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir (+1-4)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir (+1-3)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir (+1-8)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir (+1-8)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir (+1-3)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir (+2-8)
(modified) mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir (+1-4)

diff --git a/mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir b/mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir
index 81546de6a3466b..6314e6f279952b 100644
--- a/mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir
+++ b/mlir/test/Integration/Dialect/Linalg/CPU/ArmSME/fill-2d.mlir
@@ -20,12 +20,9 @@ func.func @entry() {
 
   %c123_f32 = arith.constant 123.0 : f32
 
-  %min_elts_s = arith.constant 4 : index
-  %vscale = vector.vscale
-
   // "svl" refers to the Streaming Vector Length and "svl_s" the number of
   // 32-bit elements in a vector of SVL bits.
-  %svl_s = arith.muli %min_elts_s, %vscale : index
+  %svl_s = arm_sme.streaming_vl <word>
 
   %tile_init = bufferization.alloc_tensor(%svl_s, %svl_s) : tensor<?x?xf32>
 
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir
index 936163d1cd30d9..064141c349241e 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-load-vertical.mlir
@@ -17,9 +17,7 @@ func.func @entry() {
   %c1_i32 = arith.constant 1 : i32
 
   // Calculate the size of a 32-bit tile, e.g. ZA{n}.s.
-  %vscale = vector.vscale
-  %min_elts_s = arith.constant 4 : index
-  %svl_s = arith.muli %min_elts_s, %vscale : index
+  %svl_s = arm_sme.streaming_vl <word>
   %za_s_size = arith.muli %svl_s, %svl_s : index
 
   // Allocate memory.
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir
index 839aed2e840c90..6e028d5fb83614 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-read-2d.mlir
@@ -134,12 +134,6 @@ func.func @initialize_memory(%d0 : index, %d1 : index) -> memref<?x?xf32> {
   return %A : memref<?x?xf32>
 }
 
-// This will be made a streaming function by enable-arm-streaming so return SVL.
-func.func @get_svl() -> index {
-  %vscale = vector.vscale
-  return %vscale : index
-}
-
 func.func @entry() {
   %c0 = arith.constant 0 : index
   %c1 = arith.constant 1 : index
@@ -148,8 +142,7 @@ func.func @entry() {
 
   // Allocate enough memory to load a 32-bit tile plus a tiny bit more to test
   // non-zero offsets while remaining inbounds.
-  %svl = call @get_svl() : () -> index
-  %svl_s = arith.muli %c4, %svl : index
+  %svl_s = arm_sme.streaming_vl <word>
   %svl_s_plus_two = arith.addi %svl_s, %c2 : index
 
   %A = call @initialize_memory(%svl_s_plus_two, %svl_s_plus_two) : (index, index) -> memref<?x?xf32>
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir
index 84246606daa8af..cb30fee4e12d72 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transfer-write-2d.mlir
@@ -96,12 +96,6 @@ func.func @initialize_memory(%d0 : index, %d1 : index) -> memref<?x?xf32> {
   return %A : memref<?x?xf32>
 }
 
-// This will be made a streaming function by enable-arm-streaming so return SVL.
-func.func @get_svl() -> index {
-  %vscale = vector.vscale
-  return %vscale : index
-}
-
 func.func @entry() {
   %c0 = arith.constant 0 : index
   %c2 = arith.constant 2 : index
@@ -111,8 +105,7 @@ func.func @entry() {
   //
   // Allocate enough memory to load a 32-bit tile plus a tiny bit more to test
   // non-zero offsets while remaining inbounds.
-  %svl = call @get_svl() : () -> index
-  %svl_s = arith.muli %c4, %svl : index
+  %svl_s = arm_sme.streaming_vl <word>
   %svl_s_plus_two = arith.addi %svl_s, %c2 : index
   %A = call @initialize_memory(%svl_s_plus_two, %svl_s_plus_two) : (index, index) -> memref<?x?xf32>
 
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir
index 2751c2d136485e..eee3c56351d81e 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/test-transpose.mlir
@@ -17,9 +17,7 @@ func.func @entry() {
   %c1_i32 = arith.constant 1 : i32
 
   // Calculate the size of a 32-bit tile, e.g. ZA{n}.s.
-  %vscale = vector.vscale
-  %min_elts_s = arith.constant 4 : index
-  %svl_s = arith.muli %min_elts_s, %vscale : index
+  %svl_s = arm_sme.streaming_vl <word>
   %za_s_size = arith.muli %svl_s, %svl_s : index
 
   // Allocate memory.
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir
index 5d2d0a73992f1e..2f151e2ec72fb7 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-load-store.mlir
@@ -24,12 +24,9 @@ func.func @za0_d_f64() -> i32 {
   %c1_f64 = arith.constant 1.0 : f64
   %c1_index = arith.constant 1 : index
 
-  %min_elts_d = arith.constant 2 : index
-  %vscale = vector.vscale
-
   // "svl" refers to the Streaming Vector Length and "svl_d" the number of
   // 64-bit elements in a vector of SVL bits.
-  %svl_d = arith.muli %min_elts_d, %vscale : index
+  %svl_d = arm_sme.streaming_vl <double>
 
   // Allocate "mem1" and fill each "row" with row number.
   //
@@ -170,13 +167,10 @@ func.func @load_store_two_za_s_tiles() -> i32 {
   %c1_index = arith.constant 1 : index
   %c2_index = arith.constant 2 : index
 
-  %min_elts_s = arith.constant 4 : index
-  %vscale = vector.vscale
-
   // "svl" refers to the Streaming Vector Length and "svl_s" can mean either:
   // * the number of 32-bit elements in a vector of SVL bits.
   // * the number of tile slices (1d vectors) in a 32-bit element tile.
-  %svl_s = arith.muli %min_elts_s, %vscale : index
+  %svl_s = arm_sme.streaming_vl <word>
 
   // Allocate memory for two 32-bit element tiles.
   %size_of_tile = arith.muli %svl_s, %svl_s : index
diff --git a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir
index af5fe236e5bd57..073c08bff1c415 100644
--- a/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir
+++ b/mlir/test/Integration/Dialect/Vector/CPU/ArmSME/vector-ops.mlir
@@ -14,12 +14,9 @@ func.func @entry() -> i32 {
   %c1_i8 = arith.constant 1 : i8
   %c1_index = arith.constant 1 : index
 
-  %c16 = arith.constant 16 : index
-  %vscale = vector.vscale
-
   // "svl" refers to the Streaming Vector Length and "svl_b" the number of
   // 8-bit elements in a vector of SVL bits.
-  %svl_b = arith.muli %c16, %vscale : index
+  %svl_b = arm_sme.streaming_vl <byte>
 
   // Allocate memory and fill with ones.
   //

c-rhodes

Nice! LGTM cheers

banach-space

LGTM, thanks for the clean-up!

This does make me think that we should introduce some tests comparing SVE and SSVE.

[mlir][ArmSME][test] Make use of arm_sme.streaming_vl (NFC)

e82c327

MacDue force-pushed the streaming_vl_nfc branch from 67170f7 to e82c327 Compare January 10, 2024 10:27

MacDue requested a review from c-rhodes January 10, 2024 10:28

MacDue marked this pull request as ready for review January 10, 2024 10:28

MacDue requested review from banach-space, dcaballe and nicolasvasilache as code owners January 10, 2024 10:28

llvmbot added mlir:linalg mlir:vectorops mlir mlir:sme mlir:vector labels Jan 10, 2024

c-rhodes approved these changes Jan 10, 2024

View reviewed changes

banach-space approved these changes Jan 10, 2024

View reviewed changes

MacDue merged commit dc97457 into llvm:main Jan 11, 2024

MacDue deleted the streaming_vl_nfc branch January 11, 2024 10:24

justinfargnoli pushed a commit to justinfargnoli/llvm-project that referenced this pull request Jan 28, 2024

[mlir][ArmSME][test] Make use of arm_sme.streaming_vl (NFC) (llvm#77322)

32533d1

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[mlir][ArmSME][test] Make use of arm_sme.streaming_vl (NFC) #77322

[mlir][ArmSME][test] Make use of arm_sme.streaming_vl (NFC) #77322

MacDue commented Jan 8, 2024 •

edited

Loading

llvmbot commented Jan 10, 2024 •

edited

Loading

llvmbot commented Jan 10, 2024

c-rhodes left a comment

banach-space left a comment

[mlir][ArmSME][test] Make use of arm_sme.streaming_vl (NFC) #77322

[mlir][ArmSME][test] Make use of arm_sme.streaming_vl (NFC) #77322

Conversation

MacDue commented Jan 8, 2024 • edited Loading

llvmbot commented Jan 10, 2024 • edited Loading

llvmbot commented Jan 10, 2024

c-rhodes left a comment

Choose a reason for hiding this comment

banach-space left a comment

Choose a reason for hiding this comment

MacDue commented Jan 8, 2024 •

edited

Loading

llvmbot commented Jan 10, 2024 •

edited

Loading