Add dilithium 257 NTT, iNTT, basemul

* Includes addition to the parser that allows register ranges (e.g., for vldm)

example.py

@@ -1547,6 +1547,85 @@ def core(self, slothy):
 
         slothy.optimize(start="pointwise_montgomery_acc_start", end="pointwise_montgomery_acc_end")
 
+class fnt_257_dilithium(Example):
+    def __init__(self, var="", arch=Arch_Armv7M, target=Target_CortexM7, timeout=None):
+        name = "fnt_257_dilithium"
+        infile = name
+        funcname = "__asm_fnt_257"
+
+        if var != "":
+            name += f"_{var}"
+            infile += f"_{var}"
+        name += f"_{target_label_dict[target]}"
+
+        super().__init__(infile, name, rename=True, arch=arch, target=target, timeout=timeout, funcname=funcname)
+
+    def core(self, slothy):
+        slothy.config.outputs = ["r14", "r12"]
+        slothy.config.inputs_are_outputs = True
+
+        slothy.optimize(start="_fnt_0_1_2_start", end="_fnt_0_1_2_end")
+        slothy.optimize(start="_fnt_3_4_5_6_start", end="_fnt_3_4_5_6_end")
+        slothy.optimize(start="_fnt_to_16_bit_start", end="_fnt_to_16_bit_end")
+
+class ifnt_257_dilithium(Example):
+    def __init__(self, var="", arch=Arch_Armv7M, target=Target_CortexM7, timeout=None):
+        name = "ifnt_257_dilithium"
+        infile = name
+        funcname = "__asm_ifnt_257"
+
+        if var != "":
+            name += f"_{var}"
+            infile += f"_{var}"
+        name += f"_{target_label_dict[target]}"
+
+        super().__init__(infile, name, rename=True, arch=arch, target=target, timeout=timeout, funcname=funcname)
+
+    def core(self, slothy):
+        slothy.config.outputs = ["r14", "s1", "r12"]
+        slothy.config.inputs_are_outputs = True
+
+        slothy.optimize(start="_ifnt_7_6_5_4_start", end="_ifnt_7_6_5_4_end")
+        slothy.optimize(start="_ifnt_0_1_2_start", end="_ifnt_0_1_2_end")
+
+class basemul_257_dilithium(Example):
+    def __init__(self, var="", arch=Arch_Armv7M, target=Target_CortexM7, timeout=None):
+        name = "basemul_257_dilithium"
+        infile = name
+        funcname = "__asm_point_mul_257_16"
+
+        if var != "":
+            name += f"_{var}"
+            infile += f"_{var}"
+        name += f"_{target_label_dict[target]}"
+
+        super().__init__(infile, name, rename=True, arch=arch, target=target, timeout=timeout, funcname=funcname)
+
+    def core(self, slothy):
+        slothy.config.outputs = ["r14", "r12"]
+        slothy.config.inputs_are_outputs = True
+
+        slothy.optimize(start="_point_mul_16_loop_start", end="_point_mul_16_loop_end")
+
+class basemul_257_asymmetric_dilithium(Example):
+    def __init__(self, var="", arch=Arch_Armv7M, target=Target_CortexM7, timeout=None):
+        name = "basemul_257_asymmetric_dilithium"
+        infile = name
+        funcname = "__asm_asymmetric_mul_257_16"
+
+        if var != "":
+            name += f"_{var}"
+            infile += f"_{var}"
+        name += f"_{target_label_dict[target]}"
+
+        super().__init__(infile, name, rename=True, arch=arch, target=target, timeout=timeout, funcname=funcname)
+
+    def core(self, slothy):
+        slothy.config.outputs = ["r14", "r12"]
+        slothy.config.inputs_are_outputs = True
+
+        slothy.optimize(start="_asymmetric_mul_16_loop_start", end="_asymmetric_mul_16_loop_end")
+
 def main():
     examples = [ ExampleDilithium(),
                  ExampleKyber(),
@@ -1697,7 +1776,11 @@ def main():
 
                  intt_dilithium_123_456_78(),
                  pointwise_montgomery_dilithium(),
-                 pointwise_acc_montgomery_dilithium()
+                 pointwise_acc_montgomery_dilithium(),
+                 fnt_257_dilithium(),
+                 ifnt_257_dilithium(),
+                 basemul_257_dilithium(),
+                 basemul_257_asymmetric_dilithium()
                  ]
 
     all_example_names = [e.name for e in examples]

examples/naive/armv7m/basemul_257_asymmetric_dilithium.s

@@ -0,0 +1,46 @@
+// 2
+.macro barrett_32 a, Qbar, Q, tmp
+    smmulr.w \tmp, \a, \Qbar
+    mls.w \a, \tmp, \Q, \a
+.endm
+
+.syntax unified
+.cpu cortex-m4
+
+.align 2
+.global __asm_asymmetric_mul_257_16
+.type __asm_asymmetric_mul_257_16, %function
+__asm_asymmetric_mul_257_16:
+    push.w {r4-r11, lr}
+
+    .equ width, 4
+
+    add.w r12, r0, #256*width
+    _asymmetric_mul_16_loop:
+    _asymmetric_mul_16_loop_start:
+
+    ldr.w r7, [r1, #width]
+    ldr.w r4, [r1], #2*width
+    ldr.w r8, [r2, #width]
+    ldr.w r5, [r2], #2*width
+    ldr.w r9, [r3, #width]
+    ldr.w r6, [r3], #2*width
+
+    smuad r10, r4, r6
+    smuadx r11, r4, r5
+
+    str.w r11, [r0, #width]
+    str.w r10, [r0], #2*width
+
+    smuad r10, r7, r9
+    smuadx r11, r7, r8
+
+    str.w r11, [r0, #width]
+    str.w r10, [r0], #2*width
+
+    _asymmetric_mul_16_loop_end:
+
+    cmp.w r0, r12
+    bne.w _asymmetric_mul_16_loop
+
+    pop.w {r4-r11, pc}

examples/naive/armv7m/basemul_257_dilithium.s

@@ -0,0 +1,61 @@
+// 2
+.macro barrett_32 a, Qbar, Q, tmp
+    smmulr.w \tmp, \a, \Qbar
+    mls.w \a, \tmp, \Q, \a
+.endm
+
+.syntax unified
+.cpu cortex-m4
+
+.align 2
+.global __asm_point_mul_257_16
+.type __asm_point_mul_257_16, %function
+__asm_point_mul_257_16:
+    push.w {r4-r11, lr}
+
+    ldr.w r14, [sp, #36]
+
+    .equ width, 4
+
+    add.w r12, r14, #64*width
+    _point_mul_16_loop:
+    _point_mul_16_loop_start:
+
+    ldr.w r7, [r1, #2*width]
+    ldr.w r8, [r1, #3*width]
+    ldr.w r9, [r14, #1*width]
+    ldr.w r5, [r1, #1*width]
+    ldr.w r4, [r1], #4*width
+    ldr.w r6, [r14], #2*width
+
+    smultb r10, r4, r6
+    barrett_32 r10, r2, r3, r11
+    pkhbt r4, r4, r10, lsl #16
+
+    neg.w r6, r6
+
+    smultb r10, r5, r6
+    barrett_32 r10, r2, r3, r11
+    pkhbt r5, r5, r10, lsl #16
+
+    str.w r5, [r0, #1*width]
+    str.w r4, [r0], #2*width
+
+    smultb r10, r7, r9
+    barrett_32 r10, r2, r3, r11
+    pkhbt r7, r7, r10, lsl #16
+
+    neg.w r9, r9
+
+    smultb r10, r8, r9
+    barrett_32 r10, r2, r3, r11
+    pkhbt r8, r8, r10, lsl #16
+
+    str.w r8, [r0, #1*width]
+    str.w r7, [r0], #2*width
+
+    _point_mul_16_loop_end:
+    cmp.w r14, r12
+    bne.w _point_mul_16_loop
+
+    pop.w {r4-r11, pc}