Merge pull request #1057 from pq-code-package/compiler_tests

hanno-becker · web-flow · commit 5933458463ec · 2025-05-29T14:22:29.000+01:00
CI: Run compiler tests with -O0, -Os, and -O3
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -154,10 +154,11 @@ jobs:
           cflags: "-DMLKEM_DEBUG -fsanitize=address -fsanitize=undefined -fno-sanitize-recover=all -DMLK_FORCE_AARCH64 -DMLK_CONFIG_FIPS202_BACKEND_FILE=\\\\\\\"fips202/native/aarch64/${{ matrix.backend }}.h\\\\\\\""
           check_namespace: 'false'
   compiler_tests:
-    name: Compiler tests  (${{ matrix.compiler.name }}, ${{ matrix.target.name }})
+    name: Compiler tests  (${{ matrix.compiler.name }}, ${{ matrix.target.name }}, ${{ matrix.cflags }})
     strategy:
       fail-fast: false
       matrix:
+        cflags: [ "-O0", "-Os", "-O3" ]
         target:
          - runner: pqcp-arm64
            name: 'aarch64'
@@ -285,6 +286,7 @@ jobs:
           examples: ${{ matrix.compiler.examples }}
           opt: ${{ matrix.compiler.opt }}
           nix-shell: ${{ matrix.compiler.shell }}
+          cflags: "${{ matrix.cflags }}"
       - name: native build+functest (C90)
         if: ${{ matrix.compiler.darwin || matrix.target.runner != 'macos-latest' }}
         uses: ./.github/actions/multi-functest
@@ -297,7 +299,7 @@ jobs:
           examples: ${{ matrix.compiler.examples }}
           opt: ${{ matrix.compiler.opt }}
           nix-shell: ${{ matrix.compiler.shell }}
-          cflags: "-std=c90"
+          cflags: "-std=c90 ${{ matrix.cflags }}"
       - name: native build+functest (C99)
         if: ${{ matrix.compiler.darwin || matrix.target.runner != 'macos-latest' }}
         uses: ./.github/actions/multi-functest
@@ -310,7 +312,7 @@ jobs:
           examples: ${{ matrix.compiler.examples }}
           opt: ${{ matrix.compiler.opt }}
           nix-shell: ${{ matrix.compiler.shell }}
-          cflags: "-std=c99"
+          cflags: "-std=c99 ${{ matrix.cflags }}"
       - name: native build+functest (C11)
         if: ${{ matrix.compiler.darwin || matrix.target.runner != 'macos-latest' }}
         uses: ./.github/actions/multi-functest
@@ -323,7 +325,7 @@ jobs:
           examples: ${{ matrix.compiler.examples }}
           opt: ${{ matrix.compiler.opt }}
           nix-shell: ${{ matrix.compiler.shell }}
-          cflags: "-std=c11"
+          cflags: "-std=c11 ${{ matrix.cflags }}"
       - name: native build+functest (C17)
         if: ${{ (matrix.compiler.darwin || matrix.target.runner != 'macos-latest') &&
                 matrix.compiler.c17 }}
@@ -337,7 +339,7 @@ jobs:
           examples: ${{ matrix.compiler.examples }}
           opt: ${{ matrix.compiler.opt }}
           nix-shell: ${{ matrix.compiler.shell }}
-          cflags: "-std=c17"
+          cflags: "-std=c17 ${{ matrix.cflags }}"
       - name: native build+functest (C23)
         if: ${{ (matrix.compiler.darwin || matrix.target.runner != 'macos-latest') &&
                 matrix.compiler.c23 }}
@@ -351,7 +353,7 @@ jobs:
           examples: ${{ matrix.compiler.examples }}
           opt: ${{ matrix.compiler.opt }}
           nix-shell: ${{ matrix.compiler.shell }}
-          cflags: "-std=c23"
+          cflags: "-std=c23 ${{ matrix.cflags }}"
   config_variations:
     name: Non-standard configurations
     strategy:
diff --git a/dev/aarch64_clean/meta.h b/dev/aarch64_clean/meta.h
@@ -51,26 +51,32 @@ static MLK_INLINE void mlk_poly_mulcache_compute_native(
                                 mlk_aarch64_zetas_mulcache_twisted_native);
 }
 
+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) || MLKEM_K == 2
 static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k2_native(
     int16_t r[MLKEM_N], const int16_t a[2 * MLKEM_N],
     const int16_t b[2 * MLKEM_N], const int16_t b_cache[2 * (MLKEM_N / 2)])
 {
   mlk_polyvec_basemul_acc_montgomery_cached_asm_k2(r, a, b, b_cache);
 }
+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED || MLKEM_K == 2 */
 
+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) || MLKEM_K == 3
 static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k3_native(
     int16_t r[MLKEM_N], const int16_t a[3 * MLKEM_N],
     const int16_t b[3 * MLKEM_N], const int16_t b_cache[3 * (MLKEM_N / 2)])
 {
   mlk_polyvec_basemul_acc_montgomery_cached_asm_k3(r, a, b, b_cache);
 }
+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED || MLKEM_K == 3 */
 
+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) || MLKEM_K == 4
 static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k4_native(
     int16_t r[MLKEM_N], const int16_t a[4 * MLKEM_N],
     const int16_t b[4 * MLKEM_N], const int16_t b_cache[4 * (MLKEM_N / 2)])
 {
   mlk_polyvec_basemul_acc_montgomery_cached_asm_k4(r, a, b, b_cache);
 }
+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED || MLKEM_K == 4 */
 
 static MLK_INLINE void mlk_poly_tobytes_native(uint8_t r[MLKEM_POLYBYTES],
                                                const int16_t a[MLKEM_N])
diff --git a/dev/aarch64_opt/meta.h b/dev/aarch64_opt/meta.h
@@ -53,26 +53,32 @@ static MLK_INLINE void mlk_poly_mulcache_compute_native(
                                 mlk_aarch64_zetas_mulcache_twisted_native);
 }
 
+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) || MLKEM_K == 2
 static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k2_native(
     int16_t r[MLKEM_N], const int16_t a[2 * MLKEM_N],
     const int16_t b[2 * MLKEM_N], const int16_t b_cache[2 * (MLKEM_N / 2)])
 {
   mlk_polyvec_basemul_acc_montgomery_cached_asm_k2(r, a, b, b_cache);
 }
+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED || MLKEM_K == 2 */
 
+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) || MLKEM_K == 3
 static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k3_native(
     int16_t r[MLKEM_N], const int16_t a[3 * MLKEM_N],
     const int16_t b[3 * MLKEM_N], const int16_t b_cache[3 * (MLKEM_N / 2)])
 {
   mlk_polyvec_basemul_acc_montgomery_cached_asm_k3(r, a, b, b_cache);
 }
+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED || MLKEM_K == 3 */
 
+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) || MLKEM_K == 4
 static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k4_native(
     int16_t r[MLKEM_N], const int16_t a[4 * MLKEM_N],
     const int16_t b[4 * MLKEM_N], const int16_t b_cache[4 * (MLKEM_N / 2)])
 {
   mlk_polyvec_basemul_acc_montgomery_cached_asm_k4(r, a, b, b_cache);
 }
+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED || MLKEM_K == 4 */
 
 static MLK_INLINE void mlk_poly_tobytes_native(uint8_t r[MLKEM_POLYBYTES],
                                                const int16_t a[MLKEM_N])
diff --git a/mlkem/native/aarch64/meta.h b/mlkem/native/aarch64/meta.h
@@ -53,26 +53,32 @@ static MLK_INLINE void mlk_poly_mulcache_compute_native(
                                 mlk_aarch64_zetas_mulcache_twisted_native);
 }
 
+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) || MLKEM_K == 2
 static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k2_native(
     int16_t r[MLKEM_N], const int16_t a[2 * MLKEM_N],
     const int16_t b[2 * MLKEM_N], const int16_t b_cache[2 * (MLKEM_N / 2)])
 {
   mlk_polyvec_basemul_acc_montgomery_cached_asm_k2(r, a, b, b_cache);
 }
+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED || MLKEM_K == 2 */
 
+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) || MLKEM_K == 3
 static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k3_native(
     int16_t r[MLKEM_N], const int16_t a[3 * MLKEM_N],
     const int16_t b[3 * MLKEM_N], const int16_t b_cache[3 * (MLKEM_N / 2)])
 {
   mlk_polyvec_basemul_acc_montgomery_cached_asm_k3(r, a, b, b_cache);
 }
+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED || MLKEM_K == 3 */
 
+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) || MLKEM_K == 4
 static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k4_native(
     int16_t r[MLKEM_N], const int16_t a[4 * MLKEM_N],
     const int16_t b[4 * MLKEM_N], const int16_t b_cache[4 * (MLKEM_N / 2)])
 {
   mlk_polyvec_basemul_acc_montgomery_cached_asm_k4(r, a, b, b_cache);
 }
+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED || MLKEM_K == 4 */
 
 static MLK_INLINE void mlk_poly_tobytes_native(uint8_t r[MLKEM_POLYBYTES],
                                                const int16_t a[MLKEM_N])

Original file line number	Diff line number	Diff line change
`@@ -51,26 +51,32 @@ static MLK_INLINE void mlk_poly_mulcache_compute_native(`
`51`	`51`	`mlk_aarch64_zetas_mulcache_twisted_native);`
`52`	`52`	`}`
`53`	`53`
	`54`	`+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) \|\| MLKEM_K == 2`
`54`	`55`	`static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k2_native(`
`55`	`56`	`int16_t r[MLKEM_N], const int16_t a[2 * MLKEM_N],`
`56`	`57`	`const int16_t b[2 * MLKEM_N], const int16_t b_cache[2 * (MLKEM_N / 2)])`
`57`	`58`	`{`
`58`	`59`	`mlk_polyvec_basemul_acc_montgomery_cached_asm_k2(r, a, b, b_cache);`
`59`	`60`	`}`
	`61`	`+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED \|\| MLKEM_K == 2 */`
`60`	`62`
	`63`	`+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) \|\| MLKEM_K == 3`
`61`	`64`	`static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k3_native(`
`62`	`65`	`int16_t r[MLKEM_N], const int16_t a[3 * MLKEM_N],`
`63`	`66`	`const int16_t b[3 * MLKEM_N], const int16_t b_cache[3 * (MLKEM_N / 2)])`
`64`	`67`	`{`
`65`	`68`	`mlk_polyvec_basemul_acc_montgomery_cached_asm_k3(r, a, b, b_cache);`
`66`	`69`	`}`
	`70`	`+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED \|\| MLKEM_K == 3 */`
`67`	`71`
	`72`	`+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) \|\| MLKEM_K == 4`
`68`	`73`	`static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k4_native(`
`69`	`74`	`int16_t r[MLKEM_N], const int16_t a[4 * MLKEM_N],`
`70`	`75`	`const int16_t b[4 * MLKEM_N], const int16_t b_cache[4 * (MLKEM_N / 2)])`
`71`	`76`	`{`
`72`	`77`	`mlk_polyvec_basemul_acc_montgomery_cached_asm_k4(r, a, b, b_cache);`
`73`	`78`	`}`
	`79`	`+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED \|\| MLKEM_K == 4 */`
`74`	`80`
`75`	`81`	`static MLK_INLINE void mlk_poly_tobytes_native(uint8_t r[MLKEM_POLYBYTES],`
`76`	`82`	`const int16_t a[MLKEM_N])`
Original file line number	Diff line number	Diff line change
`@@ -53,26 +53,32 @@ static MLK_INLINE void mlk_poly_mulcache_compute_native(`
`53`	`53`	`mlk_aarch64_zetas_mulcache_twisted_native);`
`54`	`54`	`}`
`55`	`55`
	`56`	`+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) \|\| MLKEM_K == 2`
`56`	`57`	`static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k2_native(`
`57`	`58`	`int16_t r[MLKEM_N], const int16_t a[2 * MLKEM_N],`
`58`	`59`	`const int16_t b[2 * MLKEM_N], const int16_t b_cache[2 * (MLKEM_N / 2)])`
`59`	`60`	`{`
`60`	`61`	`mlk_polyvec_basemul_acc_montgomery_cached_asm_k2(r, a, b, b_cache);`
`61`	`62`	`}`
	`63`	`+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED \|\| MLKEM_K == 2 */`
`62`	`64`
	`65`	`+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) \|\| MLKEM_K == 3`
`63`	`66`	`static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k3_native(`
`64`	`67`	`int16_t r[MLKEM_N], const int16_t a[3 * MLKEM_N],`
`65`	`68`	`const int16_t b[3 * MLKEM_N], const int16_t b_cache[3 * (MLKEM_N / 2)])`
`66`	`69`	`{`
`67`	`70`	`mlk_polyvec_basemul_acc_montgomery_cached_asm_k3(r, a, b, b_cache);`
`68`	`71`	`}`
	`72`	`+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED \|\| MLKEM_K == 3 */`
`69`	`73`
	`74`	`+#if defined(MLK_CONFIG_MULTILEVEL_WITH_SHARED) \|\| MLKEM_K == 4`
`70`	`75`	`static MLK_INLINE void mlk_polyvec_basemul_acc_montgomery_cached_k4_native(`
`71`	`76`	`int16_t r[MLKEM_N], const int16_t a[4 * MLKEM_N],`
`72`	`77`	`const int16_t b[4 * MLKEM_N], const int16_t b_cache[4 * (MLKEM_N / 2)])`
`73`	`78`	`{`
`74`	`79`	`mlk_polyvec_basemul_acc_montgomery_cached_asm_k4(r, a, b, b_cache);`
`75`	`80`	`}`
	`81`	`+#endif /* MLK_CONFIG_MULTILEVEL_WITH_SHARED \|\| MLKEM_K == 4 */`
`76`	`82`
`77`	`83`	`static MLK_INLINE void mlk_poly_tobytes_native(uint8_t r[MLKEM_POLYBYTES],`
`78`	`84`	`const int16_t a[MLKEM_N])`