Merge pull request #108 from slothy-optimizer/equiv_test

Add equivalence tester

.github/actions/setup-ubuntu/action.yml

@@ -0,0 +1,33 @@
+name: Setup ubuntu
+description: Setup ubuntu
+
+inputs:
+  packages:
+    description: Space-separated list of additional packages to install
+    required: false
+    default: 'llvm llvm-runtime'
+
+runs:
+  using: composite
+  steps:
+    - name: Update package repository
+      shell: bash
+      run: |
+        sudo apt-get update
+    - name: Install base packages
+      shell: bash
+      run: |
+        sudo apt-get install python3-venv python3-pip make -y
+    - name: Install additional packages
+      if: ${{ inputs.packages != ''}}
+      shell: bash
+      run: |
+        sudo apt-get install ${{ inputs.packages }} -y
+    - name: Setup Python venv
+      shell: bash
+      run: |
+        python3 -m venv venv
+        source venv/bin/activate
+        python3 -m pip install -r requirements.txt
+        deactivate
+        echo "$(pwd)/venv/bin/" >> "$GITHUB_PATH"

.github/workflows/test_basic.yaml

@@ -12,16 +12,12 @@ jobs:
             github.event.pull_request.user.login == 'mkannwischer'
             }}
     runs-on: ubuntu-latest
-    strategy: 
-      matrix:      
+    strategy:
+      matrix:
         target: [slothy.targets.arm_v7m.cortex_m7,slothy.targets.arm_v81m.cortex_m55r1, slothy.targets.arm_v81m.cortex_m85r1, slothy.targets.aarch64.cortex_a55, slothy.targets.aarch64.cortex_a72_frontend, slothy.targets.aarch64.apple_m1_firestorm_experimental, slothy.targets.aarch64.apple_m1_icestorm_experimental]
     steps:
     - uses: actions/checkout@v3
-    - name: Install python dependencies
-      run: |
-        python3 -m venv venv
-        ./venv/bin/python3 -m pip install -r requirements.txt
-        echo BASH_ENV="./venv/bin/activate" >> $GITHUB_ENV
+    - uses: ./.github/actions/setup-ubuntu
     - name: Run examples
       run: |
         python3 example.py --dry-run --only-target=${{ matrix.target }}
@@ -34,11 +30,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v3
-    - name: Install python dependencies
-      run: |
-        python3 -m venv venv
-        ./venv/bin/python3 -m pip install -r requirements.txt
-        echo BASH_ENV="./venv/bin/activate" >> $GITHUB_ENV
+    - uses: ./.github/actions/setup-ubuntu
     - name: Run tutorial
       run: |
         (cd tutorial && ./tutorial_all.sh)
@@ -51,11 +43,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v3
-    - name: Install python dependencies
-      run: |
-        python3 -m venv venv
-        ./venv/bin/python3 -m pip install -r requirements.txt
-        echo BASH_ENV="./venv/bin/activate" >> $GITHUB_ENV
+    - uses: ./.github/actions/setup-ubuntu
     - name: Run examples
       run: |
         python3 example.py --examples simple0,simple1,simple0_loop,simple1_loop
@@ -68,11 +56,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v3
-    - name: Install python dependencies
-      run: |
-        python3 -m venv venv
-        ./venv/bin/python3 -m pip install -r requirements.txt
-        echo BASH_ENV="./venv/bin/activate" >> $GITHUB_ENV
+    - uses: ./.github/actions/setup-ubuntu
     - name: Run examples
       run: |
         python3 example.py --examples ntt_kyber_1_23_45_67_m55,ntt_dilithium_12_34_56_78_m55 --timeout=300
@@ -85,11 +69,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v3
-    - name: Install python dependencies
-      run: |
-        python3 -m venv venv
-        ./venv/bin/python3 -m pip install -r requirements.txt
-        echo BASH_ENV="./venv/bin/activate" >> $GITHUB_ENV
+    - uses: ./.github/actions/setup-ubuntu
     - name: Run examples
       run: |
         python3 example.py --examples ntt_kyber_123_4567_a55,ntt_dilithium_123_45678_a55 --timeout=300
@@ -102,11 +82,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v3
-    - name: Install python dependencies
-      run: |
-        python3 -m venv venv
-        ./venv/bin/python3 -m pip install -r requirements.txt
-        echo BASH_ENV="./venv/bin/activate" >> $GITHUB_ENV
+    - uses: ./.github/actions/setup-ubuntu
     - name: Run examples
       run: |
         (cd paper/scripts && NO_LOG=Y ./slothy_sqmag.sh)
@@ -119,11 +95,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v3
-    - name: Install python dependencies
-      run: |
-        python3 -m venv venv
-        ./venv/bin/python3 -m pip install -r requirements.txt
-        echo BASH_ENV="./venv/bin/activate" >> $GITHUB_ENV
+    - uses: ./.github/actions/setup-ubuntu
     - name: Run examples
       run: |
         (cd paper/scripts && NO_LOG=Y ./slothy_fft.sh)

example.py

@@ -656,6 +656,24 @@ def core(self,slothy):
         slothy.config.inputs_are_outputs = True
         slothy.optimize(start="start", end="end")
 
+class Armv7mExample0Func(Example):
+    def __init__(self, var="", arch=Arch_Armv7M, target=Target_CortexM7):
+        name = "armv7m_simple0_func"
+        infile = name
+
+        if var != "":
+            name += f"_{var}"
+            infile += f"_{var}"
+        name += f"_{target_label_dict[target]}"
+
+        super().__init__(infile, name, rename=True, arch=arch, target=target)
+
+    def core(self,slothy):
+        slothy.config.variable_size=True
+        slothy.config.inputs_are_outputs = True
+        slothy.optimize(start="start", end="end")
+        slothy.global_selftest("my_func", {"r0": 1024 })
+
 class Armv7mLoopSubs(Example):
     def __init__(self, var="", arch=Arch_Armv7M, target=Target_CortexM7):
         name = "loop_subs"
@@ -688,7 +706,7 @@ def core(self,slothy):
         slothy.config.variable_size=True
         slothy.config.outputs = ["r6"]
         slothy.optimize_loop("start")
-        
+
 class Armv7mLoopVmovCmp(Example):
     def __init__(self, var="", arch=Arch_Armv7M, target=Target_CortexM7):
         name = "loop_vmov_cmp"
@@ -720,12 +738,13 @@ def __init__(self, var="", arch=AArch64_Neon, target=Target_CortexA55):
 
     def core(self,slothy):
         slothy.optimize()
-        
+
 class ntt_kyber_123_4567(Example):
     def __init__(self, var="", arch=AArch64_Neon, target=Target_CortexA55, timeout=None):
         name = "ntt_kyber_123_4567"
         infile = name
 
+        self.var = var
         if var != "":
             name += f"_{var}"
             infile += f"_{var}"
@@ -744,6 +763,10 @@ def core(self, slothy):
         slothy.config.constraints.stalls_first_attempt = 64
         slothy.optimize_loop("layer123_start")
         slothy.optimize_loop("layer4567_start")
+        # Build + emulate entire function to test that behaviour has not changed
+        if self.var == "":
+            slothy.global_selftest("ntt_kyber_123_4567",
+                                   {"x0": 1024, "x1": 1024, "x3": 1024, "x4": 1024, "x5": 1024})
 
 class intt_kyber_123_4567(Example):
     def __init__(self, var="", arch=AArch64_Neon, target=Target_CortexA55, timeout=None):
@@ -1226,7 +1249,7 @@ def core(self, slothy):
         slothy.config.constraints.stalls_first_attempt = 110
         slothy.optimize_loop("layer123_start")
 
-        
+
 
 
 class ntt_dilithium_123(Example):
@@ -1349,7 +1372,7 @@ def core(self, slothy):
         slothy.optimize_loop("layer5678_start")
 
         slothy.config = conf.copy()
-        
+
         if self.timeout is not None:
             slothy.config.timeout = self.timeout // 12
 
@@ -1366,7 +1389,7 @@ def core(self, slothy):
         slothy.config.split_heuristic_stepsize = 0.1
         slothy.config.constraints.stalls_first_attempt = 14
         slothy.optimize_loop("layer1234_start")
-            
+
 
 class ntt_dilithium_1234(Example):
     def __init__(self, var="", arch=AArch64_Neon, target=Target_CortexA72):
@@ -1513,6 +1536,7 @@ def main():
 
                 # Armv7m examples
                  Armv7mExample0(),
+                 Armv7mExample0Func(),
 
                 # Loop examples
                  AArch64LoopSubs(),

examples/naive/aarch64/ntt_kyber_123_4567.s

@@ -23,8 +23,13 @@
 /// SOFTWARE.
 ///
 
+// Commented out for simple standalone emulation not
+// requiring correct constant data
+//
+// Should be commented when used.
+//
 // Needed to provide ASM_LOAD directive
-#include <hal_env.h>
+// #include <hal_envh>
 
 .macro mulmodq dst, src, const, idx0, idx1
         sqrdmulh t2.8h,   \src\().8h, \const\().h[\idx1]
@@ -154,7 +159,12 @@
 .data
 .p2align 4
 roots:
-        #include "ntt_kyber_123_45_67_twiddles.s"
+// Commented out for simple standalone emulation not
+// requiring correct constant data
+//
+// Should be commented when used.
+//
+//        #include "ntt_kyber_123_45_67_twiddles.s"
 
         in      .req x0
         inp     .req x1
@@ -223,9 +233,14 @@ ntt_kyber_123_4567:
 _ntt_kyber_123_4567:
         push_stack
 
-        ASM_LOAD(r_ptr0, roots)
-        ASM_LOAD(r_ptr1, roots_l56)
-        ASM_LOAD(xtmp, const_addr)
+// Commented out for simple standalone emulation not
+// requiring correct constant data.
+//
+// Should be commented when used.
+//
+//        ASM_LOAD(r_ptr0, roots)
+//        ASM_LOAD(r_ptr1, roots_l56)
+//        ASM_LOAD(xtmp, const_addr)
 
         ld1 {consts.8h}, [xtmp]
 

examples/naive/armv7m/armv7m_simple0_func.s

@@ -0,0 +1,20 @@
+.syntax unified
+//.cpu cortex-m4 // llvm-mc does not like this...
+//.thumb // unicorn seems to get confused by this...
+
+.align 2
+.global my_func
+// .type my_func, %function // llvm-mc does not like this...
+my_func:
+  push {r4-r11, lr}
+
+start:
+ldr r8, [r0, #4]
+add r8, r2, r8
+eor.w r8, r8, r3
+smlabt r3, r2, r2, r8
+asrs r3, r3, #1
+str r3, [r0, #4]
+end:
+
+  pop {r4-r11, pc}

requirements.txt

@@ -2,3 +2,4 @@ numpy==1.26.4
 ortools==9.7.2996
 pandas==2.1.1
 sympy==1.12
+unicorn==2.1.1