From 32095b0cbbfbf2a9db382931cacbc400ae975603 Mon Sep 17 00:00:00 2001 From: Chip Kerchner Date: Tue, 1 Oct 2024 09:32:42 -0500 Subject: [PATCH] Remove parameter. --- kernel/power/sbgemv_common.c | 8 ++++---- kernel/power/sbgemv_t_vsx.c | 34 +++++++++++++++++----------------- 2 files changed, 21 insertions(+), 21 deletions(-) diff --git a/kernel/power/sbgemv_common.c b/kernel/power/sbgemv_common.c index ab50f430a..c9438b7e6 100644 --- a/kernel/power/sbgemv_common.c +++ b/kernel/power/sbgemv_common.c @@ -58,9 +58,9 @@ FORCEINLINE vec_f32 vec_load_mult(vec_bf16 *in, vec_f32 *inp, vec_bf16 zero) return vec_mult(inp, in0, zero); } -FORCEINLINE void vec_load_vec2(vec_bf16 *in, BLASLONG i, vec_f32 *v_x0, vec_bf16 zero) +FORCEINLINE void vec_load_vec2(vec_bf16 *in, vec_f32 *v_x0, vec_bf16 zero) { - vec_bf16 inp = (vec_bf16)vec_load_vec(&in[i]); + vec_bf16 inp = (vec_bf16)vec_load_vec(in); v_x0[0] = BF16_HI(inp, zero); v_x0[1] = BF16_LO(inp, zero); @@ -89,9 +89,9 @@ FORCEINLINE vec_f32 vec_loadN_mult(vec_bf16 *in, vec_f32 *inp, BLASLONG n, vec_b return vec_mult(inp, in0, zero); } -FORCEINLINE void vec_loadN_vec2(vec_bf16 *in, BLASLONG i, vec_f32 *v_x0, BLASLONG n, vec_bf16 zero) +FORCEINLINE void vec_loadN_vec2(vec_bf16 *in, vec_f32 *v_x0, BLASLONG n, vec_bf16 zero) { - vec_bf16 inp = vec_loadN(&in[i], n); + vec_bf16 inp = vec_loadN(in, n); v_x0[0] = BF16_HI(inp, zero); v_x0[1] = BF16_LO(inp, zero); diff --git a/kernel/power/sbgemv_t_vsx.c b/kernel/power/sbgemv_t_vsx.c index 075040503..272dccef7 100644 --- a/kernel/power/sbgemv_t_vsx.c +++ b/kernel/power/sbgemv_t_vsx.c @@ -55,14 +55,14 @@ static void BF16GEMV_T_VSX_1(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL BLASLONG i = 0; for (; i < n8; i++) { - vec_load_vec2(v_x, i, inp, zero); + vec_load_vec2(&v_x[i], inp, zero); temp0 += vec_load_mult(&va0[i], inp, zero); } n &= 7; if (n > 4) { - vec_loadN_vec2(v_x, i, inp, n, zero); + vec_loadN_vec2(&v_x[i], inp, n, zero); temp0 += vec_loadN_mult(&va0[i], inp, n, zero); } else if (n) { @@ -92,7 +92,7 @@ static void BF16GEMV_T_VSX_2(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL BLASLONG i = 0; for (; i < n8; i++) { - vec_load_vec2(v_x, i, inp, zero); + vec_load_vec2(&v_x[i], inp, zero); temp0 += vec_load_mult(&va0[i], inp, zero); temp1 += vec_load_mult(&va1[i], inp, zero); @@ -100,7 +100,7 @@ static void BF16GEMV_T_VSX_2(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL n &= 7; if (n > 4) { - vec_loadN_vec2(v_x, i, inp, n, zero); + vec_loadN_vec2(&v_x[i], inp, n, zero); temp0 += vec_loadN_mult(&va0[i], inp, n, zero); temp1 += vec_loadN_mult(&va1[i], inp, n, zero); @@ -139,7 +139,7 @@ static void BF16GEMV_T_VSX_4(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL BLASLONG i = 0; for (; i < n8; i++) { - vec_load_vec2(v_x, i, inp, zero); + vec_load_vec2(&v_x[i], inp, zero); temp0 += vec_load_mult(&va0[i], inp, zero); temp1 += vec_load_mult(&va1[i], inp, zero); @@ -149,7 +149,7 @@ static void BF16GEMV_T_VSX_4(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL n &= 7; if (n > 4) { - vec_loadN_vec2(v_x, i, inp, n, zero); + vec_loadN_vec2(&v_x[i], inp, n, zero); temp0 += vec_loadN_mult(&va0[i], inp, n, zero); temp1 += vec_loadN_mult(&va1[i], inp, n, zero); @@ -220,7 +220,7 @@ static void BF16GEMV_T_VSX_8(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL BLASLONG i = 0; for (; i < n8; i++) { - vec_load_vec2(v_x, i, inp, zero); + vec_load_vec2(&v_x[i], inp, zero); temp0 += vec_load_mult(&va0[i], inp, zero); temp1 += vec_load_mult(&va1[i], inp, zero); @@ -234,7 +234,7 @@ static void BF16GEMV_T_VSX_8(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL n &= 7; if (n > 4) { - vec_loadN_vec2(v_x, i, inp, n, zero); + vec_loadN_vec2(&v_x[i], inp, n, zero); temp0 += vec_loadN_mult(&va0[i], inp, n, zero); temp1 += vec_loadN_mult(&va1[i], inp, n, zero); @@ -257,7 +257,7 @@ static void BF16GEMV_T_VSX_8(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL temp7 += vec_loadNHi_mult(&va7[i], inp[0], n, zero); } - vec_f32 t0, t1, t2, t3; + vec_f32 t0, t1, t2, t3, t10, t11, t12, t13; vec_f32 a = { alpha, alpha, alpha, alpha }; vec_f32 b = { beta, beta, beta, beta }; vec_f32 *v_y = (vec_f32 *) y; @@ -272,14 +272,14 @@ static void BF16GEMV_T_VSX_8(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL temp3 = vec_mergel(t1, t3); temp0 += temp1 + temp2 + temp3; - t0 = vec_mergeh(temp4, temp6); - t1 = vec_mergel(temp4, temp6); - t2 = vec_mergeh(temp5, temp7); - t3 = vec_mergel(temp5, temp7); - temp4 = vec_mergeh(t0, t2); - temp5 = vec_mergel(t0, t2); - temp6 = vec_mergeh(t1, t3); - temp7 = vec_mergel(t1, t3); + t10 = vec_mergeh(temp4, temp6); + t11 = vec_mergel(temp4, temp6); + t12 = vec_mergeh(temp5, temp7); + t13 = vec_mergel(temp5, temp7); + temp4 = vec_mergeh(t10, t12); + temp5 = vec_mergel(t10, t12); + temp6 = vec_mergeh(t11, t13); + temp7 = vec_mergel(t11, t13); temp4 += temp5 + temp6 + temp7; vec_load_pair(inp, v_y);