fix byteswap for VLEN>256, and add initial CI

.github/workflows/bench-config.h

@@ -0,0 +1,12 @@
+#define HAS_E64 (__riscv_v_elen >= 64)
+#define HAS_F16 1
+#define MAX_MEM (4096*8)
+#define NEXT(c) (c + c/3 + 3)
+#define VALIDATE 1
+#define MIN_REPEATS 2
+#define MAX_REPEATS 2
+
+#define STOP_CYCLES (1024*1024*500)
+#define SCALE_mandelbrot(N) ((N)/10)
+#define SCALE_mergelines(N) ((N)/10)
+#define mandelbrot_ITER 100

.github/workflows/rv32-config.mk

@@ -0,0 +1,3 @@
+WARN=-Wall -Wextra -Wno-unused-function -Wno-unused-parameter
+CC=clang-17
+CFLAGS=--target=riscv32 -march=rv32gc_zve32f_zfh_zba_zbb_zbs -O3 ${WARN} -nostdlib -fno-builtin -nodefaultlibs -ffreestanding

.github/workflows/rv32-run.sh

@@ -0,0 +1,9 @@
+#!/bin/sh
+
+qemu-riscv32-static -cpu rv32,zve32f=on,vext_spec=v1.0,vlen=128,rvv_ta_all_1s=on,rvv_ma_all_1s=on,zfh=true,x-zvfh=true $@ && \
+qemu-riscv32-static -cpu rv32,zve32f=on,vext_spec=v1.0,vlen=256,rvv_ta_all_1s=on,rvv_ma_all_1s=on,zfh=true,x-zvfh=true $@ && \
+qemu-riscv32-static -cpu rv32,zve32f=on,vext_spec=v1.0,vlen=512,rvv_ta_all_1s=on,rvv_ma_all_1s=on,zfh=true,x-zvfh=true $@ && \
+qemu-riscv32-static -cpu rv32,zve32f=on,vext_spec=v1.0,vlen=1024,rvv_ta_all_1s=on,rvv_ma_all_1s=on,zfh=true,x-zvfh=true $@ &&\
+qemu-riscv32-static -cpu rv32,zve32f=on,vext_spec=v1.0,vlen=128,rvv_ta_all_1s=off,rvv_ma_all_1s=off,zfh=true,x-zvfh=true $@ && \
+qemu-riscv32-static -cpu rv32,zve32f=on,vext_spec=v1.0,vlen=1024,rvv_ta_all_1s=off,rvv_ma_all_1s=off,zfh=true,x-zvfh=true $@
+

.github/workflows/rv64-config.mk

@@ -0,0 +1,3 @@
+WARN=-Wall -Wextra -Wno-unused-function -Wno-unused-parameter
+CC=clang-17
+CFLAGS=--target=riscv64 -march=rv64gcv_zfh_zba_zbb_zbs -O3 ${WARN} -nostdlib -fno-builtin -nodefaultlibs -ffreestanding

.github/workflows/rv64-run.sh

@@ -0,0 +1,8 @@
+#!/bin/sh
+
+qemu-riscv64-static -cpu rv64,v=on,vext_spec=v1.0,vlen=128,rvv_ta_all_1s=on,rvv_ma_all_1s=on,zfh=true,x-zvfh=true $@ && \
+qemu-riscv64-static -cpu rv64,v=on,vext_spec=v1.0,vlen=256,rvv_ta_all_1s=on,rvv_ma_all_1s=on,zfh=true,x-zvfh=true $@ && \
+qemu-riscv64-static -cpu rv64,v=on,vext_spec=v1.0,vlen=512,rvv_ta_all_1s=on,rvv_ma_all_1s=on,zfh=true,x-zvfh=true $@ && \
+qemu-riscv64-static -cpu rv64,v=on,vext_spec=v1.0,vlen=1024,rvv_ta_all_1s=on,rvv_ma_all_1s=on,zfh=true,x-zvfh=true $@ && \
+qemu-riscv64-static -cpu rv64,v=on,vext_spec=v1.0,vlen=128,rvv_ta_all_1s=off,rvv_ma_all_1s=off,zfh=true,x-zvfh=true $@ && \
+qemu-riscv64-static -cpu rv64,v=on,vext_spec=v1.0,vlen=1024,rvv_ta_all_1s=off,rvv_ma_all_1s=off,zfh=true,x-zvfh=true $@

.github/workflows/validate-bench.yml

@@ -0,0 +1,33 @@
+name: Validate bench
+
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+    branches: [ main ]
+
+jobs:
+  Tests:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+    - name: Install packages
+      run: |
+        git submodule update --init --recursive
+        sudo sed 's/jammy/devel/g' -i /etc/apt/sources.list
+        sudo apt-get update -q -y
+        sudo apt-get install -y make qemu-user-static clang-17 lld-17
+    - name: Validate RV64
+      run: |
+        cp .github/workflows/rv64-config.mk ./config.mk
+        cp .github/workflows/rv64-run.sh ./run.sh
+        cp .github/workflows/bench-config.h ./bench/config.h
+        make -C bench run -j$(nproc)
+        make -C bench clean
+    - name: Validate RV32
+      run: |
+        cp .github/workflows/rv32-config.mk ./config.mk
+        cp .github/workflows/rv32-run.sh ./run.sh
+        cp .github/workflows/bench-config.h ./bench/config.h
+        make -C bench run -j$(nproc)
+        make -C bench clean

bench/Makefile

@@ -13,5 +13,5 @@ clean:
 	rm -f ${EXECS}
 
 run: all
-	for i in ${EXECS}; do ../run.sh ./$$i; done
+	for i in ${EXECS}; do ../run.sh ./$$i || { printf "\n\n\033[0;31mFAILED\033[0m\n\n"; exit 1; } ; done
 

bench/bench.h

@@ -137,7 +137,7 @@ bench_run(Bench *benches, size_t nBenches)
 			for (size_t n = 1; n < N; n = BENCH_NEXT(n)) {
 				ux si = 0, s0 = 0;
 
-#if MAX_REPEATS > 4
+#if VALIDATE
 				if (i != b->impls) {
 					URand seed = randState;
 					(void)b->func(i->func, n);

bench/byteswap.S

@@ -1,15 +1,10 @@
-/*
- * TODO: This currently only works for VLEN<=256.
- *       I think rvv 1.0 should only vrgatherei16.vv here in the future.
- */
-
-#ifdef MX
-
+#if HAS_RVV_1_0
+#if MX_N == 4 || MX_N == 2 || MX_N == 1
 
 # a0 = ptr, a1 = len
-.global MX(byteswap32_rvv_gather_)
-MX(byteswap32_rvv_gather_):
-	vsetvli t0, x0, e8, MX(), ta, ma
+.global MX(byteswap32_rvv_gatherei16_)
+MX(byteswap32_rvv_gatherei16_):
+	vsetvli t0, x0, e16, MX2(), ta, ma
 	vid.v v0
 	vand.vi v8, v0, 3
 	vrsub.vi v8, v8, 3
@@ -21,7 +16,7 @@ MX(byteswap32_rvv_gather_):
 	vle32.v v8, (a0)
 	slli t1, t0, 2
 	vsetvli x0, t1, e8, MX(), ta, ma
-	vrgather.vv v16, v8, v0
+	vrgatherei16.vv v16, v8, v0
 	vsetvli x0, t0, e32, MX(), ta, ma
 	vse32.v v16, (a0)
 	sub a1, a1, t0
@@ -32,10 +27,10 @@ MX(byteswap32_rvv_gather_):
 
 #if MX_N == 2
 
-.macro byteswap32_rvv_m1_gathers n
-	.global byteswap32_rvv_m1_gathers_m\n
-	byteswap32_rvv_m1_gathers_m\n:
-		vsetvli t0, x0, e8, m1, ta, ma
+.macro byteswap32_rvv_m1_gatherei16s n
+	.global byteswap32_rvv_m1_gatherei16s_m\n
+	byteswap32_rvv_m1_gatherei16s_m\n:
+		vsetvli t0, x0, e16, MX(), ta, ma
 		vid.v v0
 		vand.vi v8, v0, 3
 		vrsub.vi v8, v8, 3
@@ -46,17 +41,17 @@ MX(byteswap32_rvv_gather_):
 		vsetvli t0, a1, e32, m\n, ta, ma
 		vle32.v v8, (a0)
 		vsetvli t1, x0, e8, m1, ta, ma
-		vrgather.vv v16, v8, v0
+		vrgatherei16.vv v16, v8, v0
 	.ifge \n-2
-		vrgather.vv v17, v9, v0
+		vrgatherei16.vv v17, v9, v0
 	.ifge \n-4
-		vrgather.vv v18, v10, v0
-		vrgather.vv v19, v11, v0
+		vrgatherei16.vv v18, v10, v0
+		vrgatherei16.vv v19, v11, v0
 	.ifge \n-8
-		vrgather.vv v20, v12, v0
-		vrgather.vv v21, v13, v0
-		vrgather.vv v22, v14, v0
-		vrgather.vv v23, v15, v0
+		vrgatherei16.vv v20, v12, v0
+		vrgatherei16.vv v21, v13, v0
+		vrgatherei16.vv v22, v14, v0
+		vrgatherei16.vv v23, v15, v0
 	.endif
 	.endif
 	.endif
@@ -69,13 +64,13 @@ MX(byteswap32_rvv_gather_):
 		ret
 .endm
 
-byteswap32_rvv_m1_gathers 2
+byteswap32_rvv_m1_gatherei16s 2
 #endif
 #if MX_N == 4
-byteswap32_rvv_m1_gathers 4
+byteswap32_rvv_m1_gatherei16s 4
 #endif
 #if MX_N == 8
-byteswap32_rvv_m1_gathers 8
+byteswap32_rvv_m1_gatherei16s 8
 #endif
 
-
+#endif

bench/byteswap.c

@@ -40,14 +40,24 @@ byteswap32_SWAR_rev8(uint32_t *ptr, size_t n)
 #endif
 
 
+/* we don't support these on XTheadVector */
+#ifndef __riscv_vector
+#define IMPLS_RVV(f)
+#else
+#define IMPLS_RVV(f) \
+	f(rvv_gatherei16_m1) \
+	f(rvv_gatherei16_m2) \
+	f(rvv_gatherei16_m4) \
+	f(rvv_m1_gatherei16s_m2) \
+	f(rvv_m1_gatherei16s_m4) \
+	f(rvv_m1_gatherei16s_m8)
+#endif
+
 #define IMPLS(f) \
 	f(scalar) \
 	f(scalar_autovec) \
 	REV8(f) \
-	MX(f, rvv_gather) \
-	f(rvv_m1_gathers_m2) \
-	f(rvv_m1_gathers_m4) \
-	f(rvv_m1_gathers_m8) \
+	IMPLS_RVV(f)
 
 typedef void Func(uint32_t *ptr, size_t n);
 

bench/config.h

@@ -15,6 +15,8 @@
 /* stop repeats early afer this many cycles have elapsed */
 #define STOP_CYCLES (1024*1024*500)
 
+/* validate against reference implementation on the first repetition */
+#define VALIDATE 1
 
 /* custom scaling factors for benchmarks, these are used to make sure each
  * benchmark approximately takes the same amount of time. */
@@ -24,3 +26,4 @@
 
 /* benchmark specific configurations */
 #define mandelbrot_ITER 100
+

nolibc.h

@@ -26,15 +26,15 @@ static void print_flush(void);
 #define EXIT_SUCCESS 0
 
 /* customize me */
-static void
-exit(int x) { __asm volatile("unimp\n"); }
-
 static void
 memwrite(void const *ptr, size_t len) { }
 
 // static size_t /* only needed for vector-utf/bench.c */
 // memread(void *ptr, size_t len) { }
 
+static void
+exit(int x) { __asm volatile("unimp\n"); }
+
 int main(void);
 
 void _start(void) {